{"meta":{"built":"2026-09-18","counts":{"study":367,"question":811,"term":100,"page":249,"commercial":64,"said":21},"grades":["peer-reviewed","statutory-investigation","qualitative-field-study","institutional-survey","institutional-modelling","institutional-synthesis","expert-survey","compiled-review","narrative-review","working-paper","practitioner-method","operator-account","vendor-research","argued-perspective","executive-directive","self-selected-survey","simulation-model"],"words":{"peer-reviewed":"Strong","statutory-investigation":"Strong","qualitative-field-study":"Strong","institutional-survey":"Reasonable","institutional-modelling":"Reasonable","institutional-synthesis":"Reasonable","expert-survey":"Reasonable","compiled-review":"Reasonable","narrative-review":"Reasonable","working-paper":"Thin","practitioner-method":"Thin","operator-account":"Thin","vendor-research":"Thin","argued-perspective":"Thin","executive-directive":"Thin","self-selected-survey":"Thin","simulation-model":"Thin"}},"i":[{"t":"study","n":"Cognitive Offloading","au":"Risko, E. F. and Gilbert, S. J.","y":2016,"g":"peer-reviewed","sec":"judgement","f":"Defines cognitive offloading as using physical action or an external tool to reduce the mental demand of a task, and shows people offload not only when a task is hard but when they judge it to be hard.","np":"Nothing about generative AI specifically; it predates it.","u":"/research/evidence#risko-gilbert-2016","src":"https://www.cell.com/trends/cognitive-sciences/abstract/S1364-6613(16)30098-5","on":["/research/what-is-cognitive-offloading","/research/ai-and-human-judgement"],"alt":[]},{"t":"study","n":"Google Effects on Memory: Cognitive Consequences of Having Information at Our Fingertips","au":"Sparrow, B., Liu, J. and Wegner, D. M.","y":2011,"g":"peer-reviewed","sec":"judgement","f":"When people expect information to remain available, they remember where to find it rather than the thing itself.","np":"That total memory capability declines, or that the trade is net negative.","u":"/research/evidence#sparrow-2011","src":"https://www.science.org/doi/10.1126/science.1207745","on":["/research/what-is-cognitive-offloading","/research/ai-and-human-judgement"],"alt":[]},{"t":"study","n":"Habitual use of GPS negatively impacts spatial memory during self-guided navigation","au":"Dahmani, L. and Bohbot, V. D.","y":2020,"g":"peer-reviewed","sec":"learning","f":"Cross-sectionally, greater lifetime GPS experience was associated with lower use of hippocampus-dependent spatial strategies (r = -0.22 on the first probe trial), lower navigation strategy scores (r = -0.20), poorer map drawing (r = -0.22)…","np":"Anything about the brain, because no scan was taken. Anything about dementia, Alzheimer's or atrophy: those words appear nowhere in the paper. And little with confidence about the longitudinal…","u":"/research/evidence#dahmani-bohbot-2020","src":"https://www.nature.com/articles/s41598-020-62877-0","on":["/research/does-gps-damage-your-brain","/research/what-is-cognitive-offloading","/research/using-ai-without-dependency"],"alt":["NEUROIMAGING"]},{"t":"study","n":"The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects from a Survey of Knowledge…","au":"Lee, H.-P. et al.","y":2025,"g":"peer-reviewed","sec":"judgement","f":"Higher confidence in the tool was associated with less critical thinking, and the thinking that remains shifts from producing to verifying, from solving to integrating.","np":"Causation. People who think differently may use AI differently, and a survey cannot separate the two.","u":"/research/evidence#lee-2025","src":"https://www.microsoft.com/en-us/research/publication/the-impact-of-generative-ai-on-critical-thinking-self-reported-reductions-in-cognitive-effort-and-confidence-effects-from-a-survey-of-knowledge-workers/","on":["/research/ai-and-critical-thinking","/research/ai-and-human-judgement","/research/is-screen-time-the-same-argument-as-ai-use"],"alt":[]},{"t":"study","n":"Not all cognitive offloading is equal: distinguishing dependent and autonomous offloading to generative AI","au":"Zhu, Q., Li, X., Dong, Y., Chang, P. and Fan, M.","y":2026,"g":"peer-reviewed","sec":"judgement","f":"The two offloading modes are close to independent of each other rather than opposite ends of one dial (r = 0.08, p = 0.049). Dependent offloading, meaning accepting AI output with minimal evaluation and letting it structure the reasoning,…","np":"Anything about cognitive capacity. Every outcome is a self-reported appraisal on scales the authors built for this study, and their own words are: 'all outcomes were self-reported appraisals of…","u":"/research/evidence#zhu-offloading-modes-2026","src":"https://www.frontiersin.org/journals/psychology/articles/10.3389/fpsyg.2026.1878629/full","on":["/research/what-is-cognitive-offloading","/research/am-i-becoming-dependent-on-ai","/research/using-ai-without-dependency"],"alt":["COUNTRY","DOES","ERNIE","PAPER","STATE"]},{"t":"study","n":"AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking","au":"Gerlich, M.","y":2025,"g":"peer-reviewed","sec":"judgement","f":"A negative correlation between frequent AI use and critical-thinking scores, mediated by cognitive offloading, strongest among the youngest users.","np":"Causation, and it carries a published correction (Societies 2025, 15(9), 252) which anyone citing it should read alongside.","u":"/research/evidence#gerlich-2025","src":"https://www.mdpi.com/2075-4698/15/1/6","on":["/research/ai-and-critical-thinking","/research/ai-and-human-judgement","/research/is-screen-time-the-same-argument-as-ai-use"],"alt":[]},{"t":"study","n":"Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task","au":"Kosmyna, N. et al.","y":2025,"g":"working-paper","sec":"judgement","f":"The LLM group showed the weakest brain connectivity and the lowest sense of ownership over their own writing.","np":"Anything settled. 54 participants, a preprint, and reproducibility flagged by its own commentators. Treat claims of proof with suspicion.","u":"/research/evidence#kosmyna-2025","src":"https://arxiv.org/abs/2506.08872","on":["/research/ai-and-critical-thinking","/research/ai-and-human-judgement","/research/is-screen-time-the-same-argument-as-ai-use"],"alt":["EEG","LLM"]},{"t":"study","n":"Complacency and Bias in Human Use of Automation: An Attentional Integration","au":"Parasuraman, R. and Manzey, D. H.","y":2010,"g":"peer-reviewed","sec":"judgement","f":"Automation bias and complacency appear in novices and experts alike, resist training, and worsen under workload.","np":"The size of the effect for generative AI, which is far less predictable than the automation studied here.","u":"/research/evidence#parasuraman-manzey-2010","src":"https://journals.sagepub.com/doi/10.1177/0018720810376055","on":["/research/human-ai-decision-making","/research/ai-and-human-judgement"],"alt":[]},{"t":"study","n":"Generative AI Without Guardrails Can Harm Learning: Evidence from High School Mathematics","au":"Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakci, O. and Mariman,…","y":2025,"g":"peer-reviewed","sec":"learning","f":"Grades rose 48 percent with unrestricted access and 127 percent with the tutor while the tool was present. With access removed, the unrestricted group scored 17 percent LOWER than students who never had it. The guardrailed tutor largely…","np":"What a guardrailed interface should look like for professional work. It was school mathematics over a bounded period.","u":"/research/evidence#bastani-2025","src":"https://doi.org/10.1073/pnas.2422633122","on":["/research/how-humans-learn-with-ai","/research/chro-guide-to-ai","/research/ai-and-human-judgement"],"alt":["GPT","LOWER"]},{"t":"study","n":"Moral Crumple Zones: Cautionary Tales in Human-Robot Interaction","au":"Elish, M. C.","y":2019,"g":"argued-perspective","sec":"collaboration","f":"Introduces the moral crumple zone, in the author's words, to describe how responsibility for an action <q>may be misattributed to a human actor who had limited control over the behavior of an automated or autonomous system</q>. The analogy…","np":"Any rate. It is a concept paper working from selected accidents and their press coverage, chosen because they exhibit the pattern, so it establishes that the arrangement occurs and never how often.…","u":"/research/evidence#elish-2019","src":"https://estsjournal.org/index.php/ests/article/view/260","on":["/research/what-is-a-moral-crumple-zone"],"alt":[]},{"t":"study","n":"The Expertise Reversal Effect","au":"Kalyuga, S., Ayres, P., Chandler, P. and Sweller, J.","y":2003,"g":"narrative-review","sec":"learning","f":"Instructional guidance that helps inexperienced learners can lose its effect and then reverse it as the same learners gain domain knowledge. The authors' explanation is redundancy: once a learner's own schemas supply the guidance, the…","np":"Anything about AI, which did not exist in this form when it was written. Nothing in it tests a generative system, and the guidance it studies is a worked example or a labelled diagram, both of which…","u":"/research/evidence#kalyuga-2003","src":"https://doi.org/10.1207/S15326985EP3801_4","on":["/research/what-is-the-expertise-reversal-effect"],"alt":[]},{"t":"study","n":"When Problem Solving Is Superior to Studying Worked Examples","au":"Kalyuga, S., Chandler, P., Tuovinen, J. and Sweller, J.","y":2001,"g":"peer-reviewed","sec":"learning","f":"In the authors' own words, inexperienced trainees benefited most from worked examples; with more experience in the domain, worked examples became redundant and problem solving proved superior.","np":"Any magnitude, because only the abstract was readable. Also nothing about knowledge work: the domain is mechanical trades and the task is bounded, with a correct answer the experimenter already holds.","u":"/research/evidence#kalyuga-2001","src":"https://eric.ed.gov/?id=EJ640537","on":["/research/what-is-the-expertise-reversal-effect"],"alt":[]},{"t":"study","n":"The Role of Deliberate Practice in the Acquisition of Expert Performance","au":"Ericsson, K. A., Krampe, R. T. and Tesch-Romer, C.","y":1993,"g":"peer-reviewed","sec":"learning","f":"Sets out deliberate practice: effortful, targeted activity at the edge of current ability, with feedback, sustained over years.","np":"How much of the difference between performers practice explains. See the Macnamara and Maitra re-examination below.","u":"/research/evidence#ericsson-1993","src":"https://eric.ed.gov/?id=EJ471947","on":["/research/how-humans-learn-with-ai"],"alt":[]},{"t":"study","n":"The role of deliberate practice in expert performance: revisiting Ericsson, Krampe and Tesch-Romer (1993)","au":"Macnamara, B. N. and Maitra, M.","y":2019,"g":"peer-reviewed","sec":"learning","f":"Accumulated practice explained considerably less of the difference between performers than the original is usually taken to claim.","np":"That practice does not matter. It does; the simple dose-response reading is what fails.","u":"/research/evidence#macnamara-maitra-2019","src":"https://royalsocietypublishing.org/doi/10.1098/rsos.190327","on":["/research/how-humans-learn-with-ai"],"alt":[]},{"t":"study","n":"Making Things Hard on Yourself, But in a Good Way: Creating Desirable Difficulties to Enhance Learning","au":"Bjork, E. L. and Bjork, R. A.","y":2011,"g":"peer-reviewed","sec":"learning","f":"Conditions that make study feel harder improve long-term retention; conditions that make it feel fluent improve immediate performance and worsen retention. Learners systematically mistake fluency for learning.","np":"That AI-assisted work is equivalent to a fluent study condition. That inference is ours, not the authors'.","u":"/research/evidence#bjork-desirable-difficulties","src":"https://bjorklab.psych.ucla.edu/wp-content/uploads/sites/13/2016/04/EBjork_RBjork_2011.pdf","on":["/research/how-humans-learn-with-ai"],"alt":[]},{"t":"study","n":"Generative AI at Work","au":"Brynjolfsson, E., Li, D. and Raymond, L.","y":2023,"g":"peer-reviewed","sec":"learning","f":"Resolutions per hour rose 15 per cent on average, 15.2 per cent in the preferred specification with agent and tenure fixed effects. Less skilled and less experienced workers gained a 30 per cent increase in issues resolved per hour, rising…","np":"Whether those novices became experts. It measures output over months in one firm, one occupation and a stable product environment, and the authors say so. Wages, labour demand and hiring composition…","u":"/research/evidence#brynjolfsson-2023","src":"https://academic.oup.com/qje/article/140/2/889/7990658","on":["/research/how-humans-learn-with-ai","/research/staying-valuable-in-the-age-of-ai","/research/ai-and-human-judgement"],"alt":["GPT"]},{"t":"study","n":"When combinations of humans and AI are useful: a systematic review and meta-analysis","au":"Vaccaro, M., Almaatouq, A. and Malone, T.","y":2024,"g":"peer-reviewed","sec":"collaboration","f":"Human-AI combinations performed significantly WORSE on average than the better of human or AI alone (Hedges' g = -0.23). Losses concentrated in decision-making; gains in content creation. Pairing gained where humans beat the AI and lost…","np":"That human-AI teams are useless. The benchmark is an oracle-selected best performer, which you rarely know in advance. Also predates current frontier models.","u":"/research/evidence#vaccaro-2024","src":"https://www.nature.com/articles/s41562-024-02024-1","on":["/research/human-ai-decision-making","/research/how-should-leaders-respond-to-ai","/research/ai-and-human-judgement"],"alt":["WORSE"]},{"t":"study","n":"Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of AI on Knowledge Worker Productivity and Quality","au":"Dell'Acqua, F. et al.","y":2023,"g":"working-paper","sec":"collaboration","f":"Inside the frontier, AI-assisted consultants were dramatically better and faster. Outside it, they performed worse than consultants with no AI at all.","np":"Where the frontier runs in your domain. That is local and must be learned.","u":"/research/evidence#dellacqua-2023","src":"https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4573321","on":["/research/human-ai-decision-making","/research/why-learn-to-prompt-is-weak-career-advice","/research/ai-and-human-judgement"],"alt":["BCG","GPT"]},{"t":"study","n":"Algorithm Aversion: People Erroneously Avoid Algorithms After Seeing Them Err","au":"Dietvorst, B. J., Simmons, J. P. and Massey, C.","y":2015,"g":"peer-reviewed","sec":"collaboration","f":"After seeing an algorithm err, people abandon it even when it demonstrably outperforms them.","np":"That this holds for conversational AI, which is far more recent and feels different to use.","u":"/research/evidence#dietvorst-2015","src":"https://marketing.wharton.upenn.edu/wp-content/uploads/2016/10/Dietvorst-Simmons-Massey-2014.pdf","on":["/research/human-ai-decision-making","/research/ai-and-human-intuition","/research/ai-and-expert-judgement"],"alt":[]},{"t":"study","n":"Algorithm Appreciation: People Prefer Algorithmic to Human Judgment","au":"Logg, J. M., Minson, J. A. and Moore, D. A.","y":2019,"g":"peer-reviewed","sec":"collaboration","f":"Identical advice was weighted more heavily under an algorithmic label. 1A: 0.45 against 0.30, F(1,200)=8.86, p=.003, d=0.42. 1C: 0.38 against 0.26, t(284)=3.50, p=.001, d=0.44. 1B on chart positions, beta=-0.34, t(214)=5.39, p<.001. Study…","np":"Which tendency dominates in any given workplace, and three further limits the authors state themselves. The expert result rests on 61 experts in the main analysis, 67 of the 70 were men, and the…","u":"/research/evidence#logg-2019","src":"https://www.jennlogg.com/uploads/2/8/9/2/2892148/algorithm_appreciation__logg_minson_moore_2019_.pdf","on":["/research/human-ai-decision-making","/research/ai-and-expert-judgement","/research/what-is-algorithm-appreciation"],"alt":["INCONSISTENCY","LESS","MBA","PRINTED","SEVEN"]},{"t":"study","n":"Why Johnny Can't Prompt: How Non-AI Experts Try (and Fail) to Design LLM Prompts","au":"Zamfirescu-Pereira, J. D., Wong, R. Y., Hartmann, B. and Yang, Q.","y":2023,"g":"peer-reviewed","sec":"collaboration","f":"Non-experts approached prompting opportunistically rather than systematically, over-generalised from single successes and failures, and struggled to form an accurate model of the system.","np":"That this persists. It used 2023 models, and providers are actively engineering the difficulty away.","u":"/research/evidence#zamfirescu-pereira-2023","src":"https://dl.acm.org/doi/10.1145/3544548.3581388","on":["/research/why-learn-to-prompt-is-weak-career-advice"],"alt":[]},{"t":"study","n":"Comparing Physician and Artificial Intelligence Chatbot Responses to Patient Questions Posted to a Public Social Media Forum","au":"Ayers, J. W. et al.","y":2023,"g":"peer-reviewed","sec":"humanness","f":"Chatbot responses were rated good or very good quality 78.5 percent of the time against 22.1 percent for physicians, and empathetic or very empathetic 45.1 percent against 4.6 percent.","np":"That a machine can care for anyone. Doctors answering strangers free of charge between patients are not doing the job they trained for.","u":"/research/evidence#ayers-2023","src":"https://pure.johnshopkins.edu/en/publications/comparing-physician-and-artificial-intelligence-chatbot-responses/","on":["/research/what-stays-human","/research/outsourced-recognition"],"alt":[]},{"t":"study","n":"AI can help people feel heard, but an AI label diminishes this impact","au":"Yin, Y., Jia, N. and Wakslak, C. J.","y":2024,"g":"peer-reviewed","sec":"humanness","f":"AI-generated replies made recipients feel MORE heard than replies from untrained humans, and labelling the reply as AI removed the advantage.","np":"That the label effect is stable. Norms around disclosed AI assistance are moving, and nobody has measured this over time.","u":"/research/evidence#yin-2024","src":"https://pure.psu.edu/en/publications/ai-can-help-people-feel-heard-but-an-ai-label-diminishes-this-imp/","on":["/research/outsourced-recognition","/research/what-stays-human"],"alt":["MORE"]},{"t":"study","n":"Generative AI enhances individual creativity but reduces the collective diversity of novel content","au":"Doshi, A. R. and Hauser, O. P.","y":2024,"g":"peer-reviewed","sec":"humanness","f":"AI-assisted stories were rated more creative, better written and more enjoyable, with the largest gains for the least creative writers, and were markedly more similar to one another.","np":"That this generalises beyond one short creative task with one form of assistance.","u":"/research/evidence#doshi-hauser-2024","src":"https://discovery.ucl.ac.uk/id/eprint/10195027/","on":["/research/what-stays-human","/research/what-is-the-human-signal","/research/what-is-the-shared-prompt-review"],"alt":[]},{"t":"study","n":"Expertise","au":"Autor, D. and Thompson, N.","y":2025,"g":"peer-reviewed","sec":"work","f":"Automation that removed the LESS expert tasks raised wages and reduced employment. Automation that removed the EXPERT tasks lowered wages and increased employment.","np":"Anything measured about generative AI. The data ends in 2018, so this is a lens, not a forecast.","u":"/research/evidence#autor-thompson-2025","src":"https://www.nber.org/papers/w33941","on":["/research/what-is-the-judgement-premium","/research/staying-valuable-in-the-age-of-ai","/research/will-ai-replace-my-job"],"alt":["EXPERT","LESS"]},{"t":"study","n":"Still Waters, Rapid Currents: Early Labor Market Transformation under Generative AI","au":"Humlum, A. and Vestergaard, E.","y":2025,"g":"working-paper","sec":"work","f":"Precise null effects on earnings and hours two years after ChatGPT, ruling out effects larger than 2 percent, alongside substantial task reorganisation and new tasks in AI oversight and integration.","np":"That the same holds elsewhere. Denmark is high-trust, high-wage and heavily unionised, and two years is early.","u":"/research/evidence#humlum-vestergaard-2025","src":"https://www.nber.org/papers/w33777","on":["/research/ai-workforce-strategy","/research/will-ai-replace-my-job","/research/ai-and-human-judgement"],"alt":[]},{"t":"study","n":"GPTs are GPTs: Labor market impact potential of LLMs","au":"Eloundou, T., Manning, S., Mishkin, P. and Rock, D.","y":2024,"g":"peer-reviewed","sec":"work","f":"Around 80 percent of US workers could have at least 10 percent of tasks affected; about 19 percent could see at least half affected.","np":"That any job will be lost. This is exposure, not displacement, and the authors say so explicitly. It is the most misquoted number in the field.","u":"/research/evidence#eloundou-2024","src":"https://arxiv.org/abs/2303.10130","on":["/research/will-ai-replace-my-job"],"alt":[]},{"t":"study","n":"The Rapid Adoption of Generative AI","au":"Bick, A., Blandin, A. and Deming, D. J.","y":2024,"g":"working-paper","sec":"work","f":"By late 2024, nearly 40 percent of US adults aged 18-64 used generative AI and 23 percent of employed respondents had used it for work in the previous week, but only 1 to 5 percent of all work hours were assisted.","np":"Quality of use. Self-reported use counts any use at all.","u":"/research/evidence#bick-2024","src":"https://www.nber.org/papers/w32966","on":["/research/ai-workforce-strategy","/research/why-learn-to-prompt-is-weak-career-advice"],"alt":[]},{"t":"study","n":"The Future of Jobs Report 2018","au":"World Economic Forum","y":2018,"g":"institutional-survey","sec":"institutional","f":"Set out expected skill demand to 2022, with analytical thinking and innovation, active learning and creativity leading the list.","np":"A representative picture of employers. Respondents are drawn from a self-selected membership network.","u":"/research/evidence#wef-foj-2018","src":"https://www.weforum.org/publications/the-future-of-jobs-report-2018/","on":["/research/human-skills-in-the-age-of-ai"],"alt":["WEF"]},{"t":"study","n":"The Future of Jobs Report 2020","au":"World Economic Forum","y":2020,"g":"institutional-survey","sec":"institutional","f":"Named critical thinking and problem solving as leading skills, and forecast large-scale reskilling need.","np":"A clean read on AI. The 2020 edition is dominated by COVID-era disruption.","u":"/research/evidence#wef-foj-2020","src":"https://www.weforum.org/publications/the-future-of-jobs-report-2020/","on":["/research/human-skills-in-the-age-of-ai"],"alt":[]},{"t":"study","n":"The Future of Jobs Report 2023","au":"World Economic Forum","y":2023,"g":"institutional-survey","sec":"institutional","f":"Analytical thinking leads, with creative thinking second, and a growing emphasis on self-efficacy skills.","np":"Considered judgement on generative AI. It was fielded too early for that.","u":"/research/evidence#wef-foj-2023","src":"https://www.weforum.org/publications/the-future-of-jobs-report-2023/","on":["/research/human-skills-in-the-age-of-ai"],"alt":[]},{"t":"study","n":"The Future of Jobs Report 2025","au":"World Economic Forum","y":2025,"g":"institutional-survey","sec":"institutional","f":"Employers expect 39 per cent of workers' core skills to change by 2030. THE SERIES IS FALLING: 35 per cent in the first edition of 2016, a high of 57 per cent in 2020, 44 per cent in 2023, 39 per cent in 2025. Core skills required today,…","np":"What employers do. Stated skill preference and hiring behaviour diverge routinely, and the report measures no worker and tests no skill. The 39 per cent is also the most misquoted figure the estate…","u":"/research/evidence#wef-foj-2025","src":"https://www.weforum.org/publications/the-future-of-jobs-report-2025/in-full/3-skills-outlook/","on":["/research/human-skills-in-the-age-of-ai","/research/ai-workforce-strategy","/research/ai-and-human-judgement"],"alt":["FALLING","SERIES"]},{"t":"study","n":"OECD Skills Outlook 2023: Skills for a Resilient Green and Digital Transition","au":"OECD","y":2023,"g":"institutional-modelling","sec":"institutional","f":"Analyses the skills required for green and digital transitions across member economies.","np":"Anything AI-specific and current. The underlying data collection predates the generative-AI period.","u":"/research/evidence#oecd-skills-outlook-2023","src":"https://www.oecd.org/en/publications/oecd-skills-outlook-2023_27452f29-en.html","on":["/research/human-skills-in-the-age-of-ai"],"alt":["PIAAC","PISA"]},{"t":"study","n":"2025 Global Human Capital Trends","au":"Deloitte","y":2025,"g":"institutional-survey","sec":"institutional","f":"Frames the worker-organisation relationship as a set of unresolved tensions rather than a set of solved problems.","np":"Causal claims. It is a sentiment survey by a firm that sells the remedies it recommends, and should be read with that in view.","u":"/research/evidence#deloitte-hct-2025","src":"https://www.deloitte.com/us/en/insights/topics/talent/human-capital-trends/2025.html","on":["/research/ai-workforce-strategy","/research/chro-guide-to-ai"],"alt":[]},{"t":"study","n":"2025 Work Trend Index Annual Report: The Year the Frontier Firm is Born","au":"Microsoft and LinkedIn","y":2025,"g":"institutional-survey","sec":"institutional","f":"Describes the emergence of firms organised around human-agent teams and a shift towards workers managing AI agents.","np":"Independence. Microsoft sells the tools whose adoption it is measuring, and telemetry measures usage rather than value.","u":"/research/evidence#microsoft-wti-2025","src":"https://www.microsoft.com/en-us/worklab/work-trend-index/2025-the-year-the-frontier-firm-is-born","on":["/research/ai-agents-and-human-judgement","/research/ai-workforce-strategy"],"alt":[]},{"t":"study","n":"The 2025 AI Index Report","au":"Stanford HAI","y":2025,"g":"compiled-review","sec":"institutional","f":"The most comprehensive annual account of AI capability, investment, adoption and public attitudes.","np":"Much about human capability. It measures the machine side of the equation. The human side is the gap this evidence base exists to fill.","u":"/research/evidence#stanford-hai-index-2025","src":"https://hai.stanford.edu/ai-index/2025-ai-index-report","on":["/research/ai-workforce-strategy"],"alt":[]},{"t":"study","n":"A new future of work: The race to deploy AI and raise skills in Europe and beyond","au":"McKinsey Global Institute","y":2024,"g":"institutional-modelling","sec":"institutional","f":"Projects large-scale occupational transitions and rising demand for social, emotional and higher cognitive skills.","np":"What will happen. Scenario models are assumption-driven, and McKinsey's prior transition estimates have moved substantially between editions.","u":"/research/evidence#mckinsey-new-future-of-work-2024","src":"https://www.mckinsey.com/mgi/our-research/a-new-future-of-work-the-race-to-deploy-ai-and-raise-skills-in-europe-and-beyond","on":["/research/ai-workforce-strategy","/research/will-ai-replace-my-job"],"alt":[]},{"t":"study","n":"Defining the skills citizens will need in the future world of work","au":"McKinsey and Company","y":2021,"g":"institutional-survey","sec":"institutional","f":"Identifies distinct elements of talent, the DELTAs, associated with employment, income and job satisfaction.","np":"Anything about AI. It was fielded in 2019, before the generative-AI period entirely.","u":"/research/evidence#mckinsey-deltas-2021","src":"https://www.mckinsey.com/industries/public-sector/our-insights/defining-the-skills-citizens-will-need-in-the-future-world-of-work","on":["/research/human-skills-in-the-age-of-ai"],"alt":[]},{"t":"study","n":"The Skills Imperative 2035: An analysis of the demand for skills in the labour market in 2035 (Working Paper 3)","au":"NFER","y":2023,"g":"institutional-modelling","sec":"institutional","f":"Projects rising demand for a set of essential employment skills in the UK to 2035.","np":"Precision. It maps US skill data onto UK occupations, and a revised working paper corrects coding errors in the underlying labour force survey.","u":"/research/evidence#nfer-skills-imperative-2035","src":"https://www.nfer.ac.uk/publications/the-skills-imperative-2035-an-analysis-of-the-demand-for-skills-in-the-labour-market-in-2035/","on":["/research/human-skills-in-the-age-of-ai"],"alt":["NET"]},{"t":"study","n":"Agents, robots, and us: How AI reshapes work and skills in Europe","au":"McKinsey Global Institute","y":2026,"g":"institutional-modelling","sec":"institutional","f":"Models how agentic AI and robotics together reshape task composition and skill demand across Europe.","np":"Outcomes. Task exposure modelling has consistently over-predicted the pace of realised change.","u":"/research/evidence#mckinsey-agents-robots-2026","src":"https://www.mckinsey.com/mgi/our-research/agents-robots-and-us-how-ai-reshapes-work-and-skills-in-europe","on":["/research/ai-agents-and-human-judgement","/research/ai-workforce-strategy"],"alt":["GDP"]},{"t":"study","n":"Guidance on AI and Children, Version 3.0: Recommendations for AI policies and systems that uphold child rights","au":"UNICEF Innocenti","y":2025,"g":"institutional-modelling","sec":"institutional","f":"Sets out ten requirements and 48 recommendations for AI systems and policy affecting children.","np":"Empirical claims about learning or capability effects. It is a normative guidance document.","u":"/research/evidence#unicef-ai-children-2025","src":"https://www.unicef.org/innocenti/reports/policy-guidance-ai-children","on":["/research/how-humans-learn-with-ai"],"alt":[]},{"t":"study","n":"Happy Labor Day? How geopolitics, immigration and AI will reshape work","au":"Allianz Research","y":2026,"g":"institutional-modelling","sec":"institutional","f":"Models the combined effect of AI, demographics and migration on labour supply and task composition.","np":"Measured effects. Like all exposure modelling, it is an estimate of what could be affected.","u":"/research/evidence#allianz-labour-2026","src":"https://www.allianz-trade.com/en_global/news-insights/economic-insights/Happy-labor-day-How-geopolitics-immigration-AI-reshape-work.html","on":["/research/will-ai-replace-my-job"],"alt":[]},{"t":"study","n":"PwC Global AI Jobs Barometer 2025","au":"PwC","y":2025,"g":"institutional-modelling","sec":"institutional","f":"Reports wage premiums for AI skills and shifting skill requirements in AI-exposed occupations.","np":"Causation, and job advertisements describe what employers ask for rather than what the work requires.","u":"/research/evidence#pwc-jobs-barometer-2025","src":"https://www.pwc.com/gx/en/issues/artificial-intelligence/job-barometer/2025/report.pdf","on":["/research/will-ai-replace-entry-level-jobs"],"alt":[]},{"t":"study","n":"Student Generative AI Survey 2026","au":"Stephenson, R. and Armstrong, C.","y":2026,"g":"institutional-survey","sec":"learning","f":"95 per cent report using AI in at least one way and 94 per cent say they have used generative AI to help prepare assessed work, against 89 per cent in 2025 and about 53 per cent in 2024. What they use it for is mostly comprehension:…","np":"That 94 per cent of students put AI into work that gets marked. THIS IS THE MISREADING THE FIGURE INVITES and the distinction is the whole point of the survey: the 94 covers everything from…","u":"/research/evidence#hepi-2026","src":"https://www.hepi.ac.uk/reports/student-generative-ai-survey-2026/","on":["/research/how-to-use-ai-at-university","/research/how-to-assess-students-when-ai-can-do-the-assignment"],"alt":["UCAS"]},{"t":"study","n":"Assigning AI: Seven Approaches for Students, with Prompts","au":"Mollick, E. and Mollick, L.","y":2023,"g":"argued-perspective","sec":"learning","f":"The seven roles are mentor, giving feedback; tutor, giving direct instruction; coach, prompting metacognition; teammate, arguing the other side; student, whom you teach in order to find out what you do not know; simulator, for practice;…","np":"That any of it works in this form. The authors say so themselves: the approaches are 'still in their infancy and largely untested' and should be approached 'with a spirit of experimentation'.…","u":"/research/evidence#mollick-assigning-ai-2023","src":"https://arxiv.org/abs/2306.10052","on":["/research/how-to-use-ai-at-university","/research/how-humans-learn-with-ai"],"alt":[]},{"t":"study","n":"Fabricated citations: an audit across 2.5 million biomedical papers","au":"Topaz, M., Roguin, N., Gupta, P., Zhang, Z. and Peltonen, L.-M.","y":2026,"g":"peer-reviewed","sec":"learning","f":"4,046 fabricated references across 2,810 papers. The rate of papers carrying at least one rose from 1 in 2,828 in 2023 to 1 in 458 in 2025 and 1 in 277 in the first seven weeks of 2026, a twelvefold increase. Review articles ran 57 per…","np":"A rate for PubMed. The audit covers the PubMed Central OPEN ACCESS collection, and Nature published a correction to its own coverage on 13 May 2026 making exactly this distinction. The 2026 figure is…","u":"/research/evidence#topaz-2026","src":"https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(26)00603-3/fulltext","on":["/research/how-to-use-ai-at-university","/research/proving-you-did-the-work","/research/is-ai-dangerous"],"alt":["CITADEL"]},{"t":"study","n":"AI Hallucination Cases database","au":"Charlotin, D.","y":2026,"g":"compiled-review","sec":"judgement","f":"Read on 5 September 2026: roughly 2,022 decisions across 42 jurisdictions. United States 1,380, Canada 217, Australia 110, the United Kingdom 69. By party, pro se litigants 1,161 and lawyers 808, with judges appearing 31 times. By nature,…","np":"Anything with a stable number attached. This is one researcher's live tracker and every figure is a snapshot that will be out of date within days, so it must be cited with the date it was read.…","u":"/research/evidence#charlotin-hallucination-cases","src":"https://www.damiencharlotin.com/hallucinations/","on":["/research/how-to-use-ai-at-university"],"alt":[]},{"t":"study","n":"A real-world test of artificial intelligence infiltration of a university examinations system: A 'Turing Test' case study","au":"Scarfe, P., Watcham, K., Clarke, A. and Roesch, E.","y":2024,"g":"peer-reviewed","sec":"learning","f":"94 per cent of the AI submissions were not detected, and 97 per cent went undetected on the stricter test of a marker actually mentioning AI. Across the five modules there was an 83.4 per cent probability, by resampling, that the AI…","np":"How much real cheating occurs, which the authors state plainly: 'we have no way to estimate the proportion of students in our sample who used AI'. It also flatters detection rather than damning it.…","u":"/research/evidence#scarfe-2024","src":"https://doi.org/10.1371/journal.pone.0305354","on":["/research/how-to-use-ai-at-university","/research/how-to-assess-students-when-ai-can-do-the-assignment","/research/does-ai-detection-work"],"alt":["GPT"]},{"t":"study","n":"Recorded penalties for AI misuse in UK universities","au":"Multiple Freedom of Information investigations: The Times, The…","y":2026,"g":"compiled-review","sec":"learning","f":"Russell Group: 2,053 recorded punishments in 2024-25 against roughly 700 the year before, from about 350,000 students, with four members disclosing expulsions (UCL, Imperial, Glasgow and Leeds). Bristol: 526 penalties in 2023-24, against…","np":"A national picture, a trend in behaviour, or one dataset. Seven of the 24 Russell Group universities do not record AI investigations at all, so 2,053 is a floor across an incomplete sample. The three…","u":"/research/evidence#uk-university-ai-penalties-2026","src":"https://thestudenteye.substack.com/p/exclusive-university-of-bristol-sees","on":["/research/how-to-use-ai-at-university","/research/does-ai-detection-work"],"alt":["MSP","UCL"]},{"t":"study","n":"Linguistic Bias in ChatGPT: Language Models Reinforce Dialect Discrimination","au":"Fleisig, E., Smith, G., Bossi, M., Rustagi, I., Yin, X. and Klein, D.","y":2024,"g":"peer-reviewed","sec":"humanness","f":"Responses to non-standard varieties carried more stereotyping (19 per cent worse), more demeaning content (25 per cent worse), less comprehension (9 per cent worse) and more condescension (15 per cent worse), all significant after…","np":"That models exaggerate or caricature dialects, which is the claim usually attached to this paper in secondary coverage. The measured default is the opposite, features are stripped rather than…","u":"/research/evidence#fleisig-2024","src":"https://aclanthology.org/2024.emnlp-main.750/","on":["/research/does-ai-make-everyone-think-alike","/research/what-stays-human"],"alt":[]},{"t":"study","n":"Empirical evidence of Large Language Model's influence on human spoken communication","au":"Yakura, H., Lopez-Lopez, E., Brinkmann, L., de la Serna, I., Kirfel,…","y":2024,"g":"working-paper","sec":"humanness","f":"v1 reported that words the model favours rose over the 18 months after release: adept by 51 per cent, delve 48, meticulous 40, realm 35, with the word list taken from Liang and colleagues at Stanford, who compared 10,000 human-written…","np":"THE 51 PER CENT SHOULD NOT BE QUOTED AT ALL. It is adept alone, not a general figure; its outcome is the PREVALENCE OF VIDEOS containing a word, not how often speakers use it; its population is…","u":"/research/evidence#yakura-2024","src":"https://arxiv.org/abs/2409.01754","on":["/research/does-ai-make-everyone-think-alike"],"alt":["DIFFERENT","GPT","IDENTIFIER","ONE","STUDIES","TWO","UNDER"]},{"t":"study","n":"The Breakdown of Vigilance during Prolonged Visual Search","au":"Mackworth, N.H.","y":1948,"g":"peer-reviewed","sec":"judgement","f":"Detection of rare signals fell measurably between the first and second half-hour block and continued to decline across the watch. The vigilance decrement, and the founding result of the field.","np":"Anything more precise about timing than the block structure allows. Because Mackworth analysed in half-hour blocks, the decline cannot be located inside the first thirty minutes, and the widely…","u":"/research/evidence#mackworth-1948","src":"https://doi.org/10.1080/17470214808416738","on":["/research/what-are-shallow-jobs","/research/the-invisible-work-of-oversight","/research/what-is-the-vigilance-decrement"],"alt":[]},{"t":"study","n":"The Out-of-the-Loop Performance Problem and Level of Control in Automation","au":"Endsley, M. R. and Kiris, E. O.","y":1995,"g":"peer-reviewed","sec":"judgement","f":"Situation awareness was lower under fully automated and semi-automated conditions than under manual performance, and low situation awareness corresponded with out-of-the-loop performance decrements in decision time after the expert system…","np":"Anything about generative AI. This is one laboratory task from 1995, automated by an expert system producing route recommendations on a defined problem, and a model producing fluent text across every…","u":"/research/evidence#endsley-kiris-1995","src":"https://journals.sagepub.com/doi/10.1518/001872095779064555","on":["/research/what-is-the-out-of-the-loop-performance-problem","/research/can-a-human-approve-an-ai-decision-at-machine-speed"],"alt":["HERE","RECORDED","SAMPLE","SIZE"]},{"t":"study","n":"Automation and Situation Awareness","au":"Endsley, M. R.","y":1996,"g":"compiled-review","sec":"judgement","f":"Sets out three mechanisms by which automation produces the out-of-the-loop performance problem: changes in vigilance and complacency associated with monitoring, the assumption of a passive rather than an active role in controlling the…","np":"It is a review chapter and reports no new data of its own. The Level 1 against Level 2 split is the author's summary of her own study, written while that study was still in press, so it is one step…","u":"/research/evidence#endsley-1996","src":"https://maritimesafetyinnovationlab.org/wp-content/uploads/2019/12/Automation-and-Situation-Awareness-Endsley.pdf","on":["/research/what-is-the-out-of-the-loop-performance-problem"],"alt":[]},{"t":"study","n":"The vigilance decrement: its first 75 years","au":"Klein, R.M. and Feltmate, B.B.T.","y":2025,"g":"compiled-review","sec":"judgement","f":"The decrement itself has held across the literature. Its mechanism has not: whether the decline reflects falling sensitivity or a shifting response criterion remains disputed, and the sensitivity account has recently been challenged.","np":"Any particular mechanism, and therefore any intervention that depends on one. A page arguing that oversight fails for a specific cognitive reason is going beyond what this review supports.","u":"/research/evidence#klein-feltmate-2025","src":"https://doi.org/10.3389/fcogn.2025.1632885","on":["/research/what-are-shallow-jobs","/research/what-is-the-vigilance-decrement"],"alt":[]},{"t":"study","n":"Intelligent AI Delegation","au":"Tomasev, N., Franklin, M. and Osindero, S.","y":2026,"g":"working-paper","sec":"collaboration","f":"The authors name oversight readiness as a property a future workforce may lack, arguing that expertise is built through the repetitive execution of narrowly scoped tasks, that those are the tasks most likely to be delegated to agents…","np":"Anything measured. It is a preprint framework paper with no empirical component, so it establishes that the risk is taken seriously by practitioners and not that the risk has been observed. The…","u":"/research/evidence#tomasev-2026","src":"https://arxiv.org/abs/2602.11865","on":["/research/what-is-oversight-readiness","/research/what-is-a-frontier-firm","/research/who-manages-ai-agents"],"alt":[]},{"t":"study","n":"AI-induced never-skilling in medical education","au":"Ke, Y., Jin, L., Ong, J.C.L., Thirunavukarasu, A.J., Car, J., Cheung,…","y":2026,"g":"argued-perspective","sec":"learning","f":"Separates three distinct failures. Deskilling is the degradation of established competence in clinicians already trained. Mis-skilling is the acquisition of incorrect reasoning patterns through uncritical adoption of erroneous or biased AI…","np":"That never-skilling occurs. The authors disclaim this themselves and do so more than once: 'Direct causal evidence linking AI exposure during training to competency failure in medical trainees does…","u":"/research/evidence#ke-2026","src":"https://www.nature.com/articles/s41591-026-04438-y","on":["/research/will-ai-replace-entry-level-jobs","/research/synthetic-seniority","/research/how-humans-learn-with-ai"],"alt":[]},{"t":"study","n":"From Future of Work to Future of Workers: Addressing Asymptomatic AI Harms to Foster Dignified Human-AI Interaction","au":"Ehsan, U., Passi, S., Saha, K., McNutt, T., Riedl, M.O. and Alcorn, S.","y":2026,"g":"qualitative-field-study","sec":"frontline","f":"Measured operational gains and reported capability loss ran together. Planning cycles shortened by roughly 15 per cent and confidence rose, while by month nine several dosimetrists said their unaided proficiency had worsened over the year,…","np":"Any measured deskilling. The skill claims are self-reported, plus one informal unaided exercise with a couple of dosimetrists during a workshop. There is no controlled comparison, no pre and post…","u":"/research/evidence#ehsan-2026","src":"https://dl.acm.org/doi/10.1145/3772318.3791081","on":["/research/capability-debt","/research/which-professions-face-the-greatest-deskilling-risk","/research/ai-and-human-judgement"],"alt":[]},{"t":"study","n":"I surveyed workers to see if AI had caused job losses and was surprised by the findings","au":"Dixon, J.C.","y":2026,"g":"institutional-survey","sec":"work","f":"About 3 per cent said they had lost a job to AI since 2023, against roughly 6 per cent who said they held a job that did not exist before AI and about 9 per cent reporting an AI-related promotion. Around 95 per cent said no, with 3 to 4…","np":"Displacement in the workforce, and the reason is structural rather than a matter of sampling error. Every respondent was employed when surveyed, in Dixon's own words 'none were jobless, whether due…","u":"/research/evidence#dixon-2026","src":"https://theconversation.com/i-surveyed-workers-to-see-if-ai-had-caused-job-losses-and-was-surprised-by-the-findings-290100","on":["/research/will-ai-replace-entry-level-jobs","/research/what-is-the-ai-employment-gap"],"alt":[]},{"t":"study","n":"AI Fatigue in 2026: The State of the Engineer","au":"The Clearing","y":2026,"g":"self-selected-survey","sec":"professions","f":"63 per cent 'report measurable decline in at least one core skill they had before adopting AI tools'; 71 per cent agree with 'I often feel like a middleman between AI output and actual results'; 67 per cent spend the majority of coding…","np":"Any rate. The frame selects on the outcome: respondents arrived by taking a quiz about AI fatigue, so the proportion reporting AI fatigue cannot be read as a prevalence. The word 'measurable' in…","u":"/research/evidence#clearing-ai-fatigue-2026","src":"https://clearing-ai.com/ai-fatigue-2026-report.html","on":["/research/what-is-the-illusion-of-competence","/research/is-deskilling-real"],"alt":["FAQ"]},{"t":"study","n":"Endoscopist deskilling risk after exposure to artificial intelligence in colonoscopy: a multicentre, observational study","au":"Budzyn, K., Roman'czyk, M., Kitala, D. et al.","y":2025,"g":"peer-reviewed","sec":"frontline","f":"Adenoma detection rate in unassisted colonoscopy fell from 28.4 percent before AI exposure to 22.4 percent after, a drop of 6.0 percentage points (p=0.0089; adjusted odds ratio 0.69).","np":"Causation with certainty: it is observational, not randomised, and other changes over the period cannot be fully excluded. It is also one procedure in one country, and detection rate is a proxy for…","u":"/research/evidence#budzyn-2025","src":"https://pubmed.ncbi.nlm.nih.gov/40816301/","on":["/research/capability-debt","/research/ai-and-human-judgement","/research/how-humans-learn-with-ai"],"alt":["ACCEPT","WITHOUT"]},{"t":"study","n":"Heterogeneity and predictors of the effects of AI assistance on radiologists","au":"Yu, F., Moehring, A., Banerjee, O., Salz, T., Agarwal, N. and…","y":2024,"g":"peer-reviewed","sec":"frontline","f":"The effect of AI assistance diverged sharply between radiologists, from strongly positive to strongly negative. Experience, subspecialty and prior familiarity with AI all failed to predict who would benefit, and lower performers did not…","np":"That AI assistance is bad on average, or that the pattern holds outside diagnostic imaging.","u":"/research/evidence#yu-2024","src":"https://pubmed.ncbi.nlm.nih.gov/38504016/","on":["/research/human-ai-decision-making","/research/staying-valuable-in-the-age-of-ai","/research/ai-and-human-judgement"],"alt":[]},{"t":"study","n":"Robots and Labor in Nursing Homes","au":"Lee, Y. S., Iizuka, T. and Eggleston, K.","y":2024,"g":"working-paper","sec":"frontline","f":"Robot adoption RAISED employment and improved retention, most strongly for non-regular staff, reallocated worker effort towards direct care, and improved quality: less use of physical restraint and fewer pressure ulcers.","np":"That this generalises. Japanese long-term care faces acute labour shortage, so robots substituted for vacancies rather than for people, which is a very particular condition.","u":"/research/evidence#lee-nursing-homes-2024","src":"https://www.nber.org/system/files/working_papers/w33116/w33116.pdf","on":["/research/what-stays-human","/research/ai-workforce-strategy"],"alt":["RAISED"]},{"t":"study","n":"The Adjustment of Labor Markets to Robots","au":"Dauth, W., Findeisen, S., Suedekum, J. and Woessner, N.","y":2021,"g":"peer-reviewed","sec":"frontline","f":"Incumbent workers largely kept their jobs and moved into new, higher-quality tasks within their original plants. The cost fell instead on young labour-market entrants, who shifted away from vocational manufacturing training towards…","np":"That generative AI will behave like industrial robots. The technologies and the tasks differ substantially.","u":"/research/evidence#dauth-2021","src":"https://academic.oup.com/jeea/article-abstract/19/6/3104/6179884","on":["/research/missing-rungs","/research/will-ai-replace-entry-level-jobs"],"alt":[]},{"t":"study","n":"Generative AI as Seniority-Biased Technological Change: Evidence from U.S. Resume and Job Posting Data","au":"Hosseini Maasoum, S. M. and Lichtinger, G.","y":2026,"g":"working-paper","sec":"frontline","f":"Junior employment at adopting firms fell about 9 per cent relative to non-adopters six quarters after diffusion, and 8 per cent eight quarters after adoption in the staggered design, while senior employment showed no comparable break. The…","np":"Causation. The authors write throughout that adoption IS ASSOCIATED WITH the decline, call the evidence suggestive, and state in their conclusion that unobserved confounders may remain and that the…","u":"/research/evidence#hosseini-lichtinger-2026","src":"https://papers.ssrn.com/sol3/papers.cfm?abstract_id=5425555","on":["/research/missing-rungs","/research/will-ai-replace-entry-level-jobs","/research/do-apprenticeships-still-work"],"alt":["ALSO","LLM","NET","WRDS"]},{"t":"study","n":"AI, Skill, and Productivity: The Case of Taxi Drivers","au":"Kanazawa, K., Kawaguchi, D., Shigeoka, H. and Watanabe, Y.","y":2022,"g":"peer-reviewed","sec":"frontline","f":"Productivity gains accrued almost entirely to LOW-skilled drivers, narrowing the gap between best and worst by 14 percent.","np":"Included deliberately as a disconfirming case for a tidy story. Anyone arguing that AI levels up white-collar workers while degrading frontline ones has to explain this result, which runs the other…","u":"/research/evidence#kanazawa-2022","src":"https://www.nber.org/papers/w30612","on":["/research/staying-valuable-in-the-age-of-ai","/research/ai-and-expert-judgement"],"alt":["LOW"]},{"t":"study","n":"Algorithmic management is associated with psychological distress, musculoskeletal pain, and occupational accidents: a cross-sectional study in…","au":"Nilsson, A. et al. (Karolinska Institutet)","y":2025,"g":"peer-reviewed","sec":"frontline","f":"Higher exposure to algorithmic management was associated with greater psychological distress, more occupational accidents and more musculoskeletal pain.","np":"Causation: it is cross-sectional and self-reported. Workers under strain may also perceive management as more algorithmic.","u":"/research/evidence#nilsson-2025","src":"https://link.springer.com/article/10.1007/s00420-025-02180-5","on":["/research/ai-workforce-strategy","/research/how-should-leaders-respond-to-ai"],"alt":[]},{"t":"study","n":"Doing Well by Doing Good: Improving Retail Store Performance with Responsible Scheduling Practices at the Gap, Inc.","au":"Kesavan, S., Lambert, S. J., Williams, J. C. and Pendem, P. K.","y":2022,"g":"peer-reviewed","sec":"frontline","f":"Restoring schedule predictability and worker control raised productivity 5.1 percent, with sales up 3.3 percent and labour hours down 1.8 percent.","np":"Anything directly about generative AI. It concerns algorithmic scheduling, which is an older and different technology.","u":"/research/evidence#kesavan-2022","src":"https://pubsonline.informs.org/doi/10.1287/mnsc.2021.4291","on":["/research/how-should-leaders-respond-to-ai"],"alt":[]},{"t":"study","n":"Folgen des technologischen Wandels fur den Arbeitsmarkt (Consequences of technological change for the labour market: it is above all the highly…","au":"Institut fur Arbeitsmarkt- und Berufsforschung (IAB), Germany","y":2024,"g":"institutional-modelling","sec":"international","f":"Substitutability rose about ten percentage points for degree-level expert occupations between 2019 and 2022, and was roughly flat for helper occupations. IAB frames AI as relief for skills shortages rather than as displacement.","np":"Realised outcomes. Substitutability is technical potential assessed by coders, not what employers did.","u":"/research/evidence#iab-2024","src":"https://doku.iab.de/kurzber/2024/kb2024-05.pdf","on":["/research/will-ai-replace-my-job","/research/human-skills-in-the-age-of-ai"],"alt":["BERUFENET","IAB"]},{"t":"study","n":"Etude des impacts de l'IA sur le travail: Rapport d'enquete LaborIA Explorer (Study of the impacts of AI on work)","au":"LaborIA (French Ministry of Labour, Inria and Matrice)","y":2024,"g":"institutional-survey","sec":"international","f":"Names a conflit de rationalite, a clash of rationalities: managers justify AI by error reduction (81 percent), performance (75 percent) and removing drudgery (74 percent), while fieldwork shows workers becoming the system's de facto…","np":"Scale. The qualitative core rests on six sites and ten repeated interviews.","u":"/research/evidence#laboria-2024","src":"https://www.laboria.ai/wp-content/uploads/2024/05/Rapport-denquete-LaborIA-Explorer.pdf","on":["/research/ai-workforce-strategy","/research/how-should-leaders-respond-to-ai"],"alt":[]},{"t":"study","n":"Digitalisierung und Wandel der Beschaftigung, DiWaBe 2.0 (Digitalisation and the transformation of employment)","au":"BAuA, ZEW, IAB and BIBB, Germany","y":2025,"g":"institutional-survey","sec":"international","f":"More than half already use AI at work but largely informally. Use ranges from about a third of unqualified workers to around 80 percent of those with a degree or Meister qualification. There was NO difference in training participation…","np":"What that absence of training does to capability over time. The survey is a single 2024 snapshot.","u":"/research/evidence#diwabe-2025","src":"https://www.baua.de/EN/Service/Publications/Report/F2573","on":["/research/ai-workforce-strategy","/research/chro-guide-to-ai"],"alt":[]},{"t":"study","n":"Survey on the impact of workplace AI adoption on working styles (Research Series No. 256)","au":"Japan Institute for Labour Policy and Training (JILPT)","y":2025,"g":"institutional-survey","sec":"international","f":"Only 12.9 percent report any firm AI use and 8.4 percent use it themselves. Among users, reports of improved job quality and wellbeing outweighed reports of decline, and the gain was markedly larger where the employer had consulted staff…","np":"Long-run capability effects, and Japanese adoption rates are far below US levels so the user group is early and unusual.","u":"/research/evidence#jilpt-2025","src":"https://www.jil.go.jp/institute/research/2025/256.html","on":["/research/ai-workforce-strategy","/research/design-versus-drift"],"alt":[]},{"t":"study","n":"Changes in the labour market due to artificial intelligence and policy directions (Research Report 2023-03)","au":"Korea Development Institute (KDI)","y":2023,"g":"institutional-modelling","sec":"international","f":"38.8 percent of jobs are technically automatable across more than 70 percent of their tasks, yet only 2.7 percent of firms with ten or more staff had adopted AI. Realised effects showed no aggregate employment change, lower earnings, and…","np":"That the Korean pattern transfers. Korea has unusually high tertiary education rates and a distinctive labour market.","u":"/research/evidence#kdi-2023","src":"https://www.kdi.re.kr/research/reportView?pub_no=18370","on":["/research/will-ai-replace-entry-level-jobs","/research/will-ai-replace-my-job"],"alt":["GPT","KDI","YOUNGER"]},{"t":"study","n":"Analysis of the impact of artificial intelligence technology on employment and income in China","au":"Lu, Y. and Gui, L. (Bulletin of the Chinese Academy of Sciences)","y":2025,"g":"institutional-modelling","sec":"international","f":"Between 2018 and 2023 the substitution effect outweighed complementarity: a one percent rise in industrial robots reduced firm labour demand by 0.18 percent. Roughly 200 million people, 27 percent of employment, are in flexible work with…","np":"Comparability with service-sector generative AI in high-income economies. This is largely industrial robotics.","u":"/research/evidence#cas-2025","src":"http://old2022.bulletin.cas.cn/publish_article/2025/4/20250408.htm","on":["/research/will-ai-replace-my-job"],"alt":[]},{"t":"study","n":"Impacto economico de la inteligencia artificial en America Latina (The economic impact of artificial intelligence in Latin America)","au":"Jung, J. and Katz, R. (CEPAL / ECLAC)","y":2026,"g":"institutional-modelling","sec":"international","f":"The gains run through skilled labour, and the binding constraint across Latin America is human-capital formation and low investment. The regional risk is UNDER-adoption rather than displacement.","np":"Firm-level outcomes; it is a macro model.","u":"/research/evidence#cepal-2026","src":"https://www.cepal.org/es/publicaciones/81909-impacto-economico-la-inteligencia-artificial-america-latina-transformacion","on":["/research/ai-workforce-strategy"],"alt":["UNDER"]},{"t":"study","n":"Inteligencia artificial y mercado de trabajo en Espana (Artificial intelligence and the labour market in Spain)","au":"Rodriguez-Fernandez, M. (Funcas)","y":2026,"g":"institutional-modelling","sec":"international","f":"Spain shows medium-high exposure at 27.4 percent but low automation risk at 5.9 percent, against an OECD average near 12 percent, because of its interpersonal and physical occupational mix.","np":"Outcomes. Exposure indices remain estimates of what could be affected.","u":"/research/evidence#funcas-2026","src":"https://www.funcas.es/documentos_trabajo/inteligencia-artificial-y-mercado-de-trabajo-en-espana-exposicion-ocupacional-efectos-sobre-el-empleo-y-adopcion-empresarial/","on":["/research/will-ai-replace-my-job"],"alt":["OECD"]},{"t":"study","n":"State of Working India 2026: Youth in the Labour Market","au":"Azim Premji University","y":2026,"g":"institutional-modelling","sec":"international","f":"Roughly five million graduates enter the Indian labour market each year against about 2.8 million finding work, with graduate unemployment near 40 percent for 15 to 25 year olds. The report explicitly declines to attribute this to AI.","np":"Anything about AI's effect in India, which it deliberately does not claim.","u":"/research/evidence#azim-premji-2026","src":"https://azimpremjiuniversity.edu.in/publications/2026/report/swi-2026","on":["/research/will-ai-replace-entry-level-jobs"],"alt":["AISHE","CMIE","CPHS","MIS","NCVT"]},{"t":"study","n":"Note de conjoncture: digital investment, artificial intelligence and youth employment","au":"INSEE, France","y":2026,"g":"institutional-modelling","sec":"international","f":"French employment of 15 to 29 year olds, excluding apprentices, fell 7.4 percent year on year in IT services, 5.8 percent in publishing and 3.7 percent in management consulting in Q4 2025, against minus 0.7 percent across the market sector…","np":"That AI caused it. INSEE explicitly cautions against attributing the fall to AI alone.","u":"/research/evidence#insee-2026","src":"https://www.insee.fr/fr/statistiques/fichier/8907419/ndc-mars-2026-ecl-AI.pdf","on":["/research/will-ai-replace-entry-level-jobs","/research/will-ai-replace-my-job"],"alt":[]},{"t":"study","n":"Navigation-related structural change in the hippocampi of taxi drivers","au":"Maguire, E. A., Gadian, D. G., Johnsrude, I. S., Good, C. D.,…","y":2000,"g":"peer-reviewed","sec":"learning","f":"Posterior hippocampi were significantly larger in taxi drivers than in controls, and hippocampal volume correlated with time spent driving a taxi, positively in the posterior and negatively in the anterior hippocampus.","np":"Causation. It is cross-sectional, so it cannot separate the practice building the brain from a particular brain selecting into the job. It says nothing about what happens when the practice stops,…","u":"/research/evidence#maguire-2000","src":"https://www.pnas.org/doi/10.1073/pnas.070039597","on":["/research/what-is-deliberate-practice","/research/how-fast-do-skills-decay"],"alt":["MRI"]},{"t":"study","n":"Acquiring 'the Knowledge' of London's Layout Drives Structural Brain Changes","au":"Woollett, K. and Maguire, E. A.","y":2011,"g":"peer-reviewed","sec":"learning","f":"In those who qualified, acquiring an internal spatial representation of London was associated with a selective increase in grey matter volume in the posterior hippocampi, with concomitant changes to their memory profile. In the authors'…","np":"Anything about AI, and anything about removal. This is acquisition, over four years, in one domain. It does not show that the structure regresses when the practice is delegated to a machine, and no…","u":"/research/evidence#woollett-maguire-2011","src":"https://www.cell.com/current-biology/fulltext/S0960-9822(11)01267-X","on":["/research/what-is-deliberate-practice","/research/how-fast-do-skills-decay","/research/what-is-the-substitution-myth"],"alt":["MRI"]},{"t":"study","n":"Identifying indicators of consciousness in AI systems","au":"Butlin, P., Long, R., Bayne, T., Bengio, Y., Birch, J., Chalmers, D.,…","y":2026,"g":"peer-reviewed","sec":"institutional","f":"The method is offered as a way to inform credences about whether a particular AI system is conscious, on the ground that computational functionalist theories have implications for AI that can be investigated empirically. The 2023…","np":"Whether any system is conscious. The method inherits every disagreement in consciousness science, and the paper says so: the indicators are only as good as the theories they come from, and those…","u":"/research/evidence#butlin-2026","src":"https://doi.org/10.1016/j.tics.2025.10.011","on":["/research/is-ai-conscious"],"alt":[]},{"t":"study","n":"Could a Large Language Model Be Conscious?","au":"Chalmers, D. J.","y":2023,"g":"argued-perspective","sec":"institutional","f":"The six candidates are biology, senses and embodiment, world models and self models, recurrent processing, global workspace, and unified agency. He argues it would not be unreasonable to hold a credence of at least one in three that each…","np":"Anything about any system. These are credences in an argument, not estimates from data, and the author says so in terms. The 2020 survey figures he reports, roughly 3 per cent of professional…","u":"/research/evidence#chalmers-2023","src":"https://www.bostonreview.net/articles/could-a-large-language-model-be-conscious/","on":["/research/is-ai-conscious"],"alt":["LLM"]},{"t":"study","n":"Reflexive AI usage is now a baseline expectation at Shopify","au":"Lutke, T., Chief Executive Officer, Shopify","y":2025,"g":"executive-directive","sec":"institutional","f":"The memo states that reflexive AI usage is now a baseline expectation at Shopify. Teams are required to demonstrate why AI could not do a job before requesting additional headcount or resources, and AI use is added to performance and peer…","np":"Anything about what followed. The memo measures no outcome, reports no adoption figure and tests no capability. Three comparable memos were sent within a month of it and the four firms then diverged…","u":"/research/evidence#lutke-2025","src":"https://www.digitalcommerce360.com/2025/04/08/internal-memo-shopify-ceo-declares-ai-non-optional/","on":["/research/what-is-the-ai-memo-test"],"alt":[]},{"t":"study","n":"The AI-first memo and the clarification of 23 May 2025","au":"von Ahn, L., Chief Executive Officer, Duolingo","y":2025,"g":"executive-directive","sec":"institutional","f":"The April memo set out an AI-first approach and said the company would gradually stop using contractors to do work that AI can handle. On 23 May von Ahn wrote that one of the most important things leaders can do is provide clarity and that…","np":"What actually happened to employment at Duolingo. The clarification is as much an executive statement as the memo, made in the same interest, and reporting alongside it records a reduction in…","u":"/research/evidence#vonahn-2025","src":"https://www.entrepreneur.com/business-news/duolingo-ceo-clarifies-ai-stance-after-backlash-read-memo/492141","on":["/research/what-is-the-ai-memo-test"],"alt":[]},{"t":"study","n":"The April 2025 upskilling memo and the September 2025 workforce reduction","au":"Kaufman, M., Chief Executive Officer, Fiverr","y":2025,"g":"executive-directive","sec":"institutional","f":"In April 2025 Kaufman told staff that AI was coming for their jobs and for his own, calling it a wake-up call, and warned that those who did not upskill would face the need for a career change in a matter of months. On 17 September 2025 he…","np":"That the upskilling instruction caused, prevented or is otherwise connected to the reduction. Nothing links the two documents beyond their author and their sequence, and no information is published…","u":"/research/evidence#kaufman-2025","src":"https://www.itpro.com/business/business-strategy/fiverr-ceo-micha-kaufman-workforce-layoffs","on":["/research/what-is-the-ai-memo-test"],"alt":[]},{"t":"study","n":"Smart Predictive Technologies Enhance Pilgrim Safety and Crowd Management","au":"Ministry of Interior, Kingdom of Saudi Arabia","y":2026,"g":"operator-account","sec":"international","f":"The ministry states it implemented predictive analytics to anticipate and mitigate congestion and hazardous conditions before they occurred, using an intelligent framework powered by AI and data analytics to accelerate strategic…","np":"Anything measured. There is no figure for accuracy, no override rate, no comparison against unaided command judgement, and no independent evaluation of whether the oversight functions. It is a state…","u":"/research/evidence#spa-hajj-ai-2026","src":"https://www.spa.gov.sa/en/N2603232","on":["/research/ai-and-work-in-the-gulf","/research/what-is-meaningful-human-oversight"],"alt":[]},{"t":"study","n":"SDAIA: Saudi AI Platform Baseer Boosts Crowd, Security Control During Hajj","au":"Moquim, S. A., Vice President, Saudi Data and Artificial Intelligence…","y":2025,"g":"operator-account","sec":"international","f":"SDAIA operates Baseer, built with the Ministry of Interior, using AI algorithms and computer vision on live feeds to detect crowd density and distribution within the Grand Mosque and to pinpoint overcrowded zones such as the Tawaf area…","np":"That the oversight works, or that anyone has tested it. Nothing here measures how often a commander overrides the system, what happens when the prediction is wrong, or whether unaided crowd-reading…","u":"/research/evidence#sdaia-baseer-2025","src":"https://english.aawsat.com/gulf/5150726-sdaia-saudi-ai-platform-baseer-boosts-crowd-security-control-during-hajj","on":["/research/ai-and-work-in-the-gulf","/research/what-is-meaningful-human-oversight","/research/who-supervises-work-they-cannot-do"],"alt":[]},{"t":"study","n":"Cognitive offloading and the speedup illusion in human-AI interaction","au":"Yu, S., Cheng, M., Jabbar, A., Sucholutsky, I., Collins, K. M.,…","y":2026,"g":"peer-reviewed","sec":"judgement","f":"Actual completion times did not differ between independent and AI-assisted completion, while participants predicted AI would be significantly faster. The same bias did not appear when participants imagined help from another person.…","np":"Anything about professional work, output quality or long-horizon tasks. The tasks were short and simple by design, and a forecasting error is not the same thing as a productivity claim.","u":"/research/evidence#yu-speedup-illusion-2026","src":"https://arxiv.org/abs/2605.23177","on":["/research/usage-theatre","/research/the-unclaimed-hour","/research/what-is-cognitive-offloading"],"alt":[]},{"t":"study","n":"Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence","au":"Brynjolfsson, E., Chandar, B. and Chen, R.","y":2026,"g":"working-paper","sec":"work","f":"No widespread economy-wide displacement. But employment among 22 to 25 year olds in highly AI-exposed occupations sits about 19 percent below where it would be had it tracked similarly aged workers in less-exposed occupations. The…","np":"Economy-wide job destruction, which the authors explicitly rule out on current evidence. Nor does it establish causation: youth hiring is sensitive to interest rates, cohort size and hiring freezes,…","u":"/research/evidence#brynjolfsson-canaries-2026","src":"https://digitaleconomy.stanford.edu/publication/canaries-in-the-coal-mine-six-facts-about-the-recent-employment-effects-of-artificial-intelligence/","on":["/research/will-ai-replace-entry-level-jobs","/research/missing-rungs","/research/the-best-writing-on-ai"],"alt":["ADP"]},{"t":"study","n":"Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity","au":"Becker, J., Rush, N., Barnes, E. and Rein, D. (METR)","y":2025,"g":"working-paper","sec":"collaboration","f":"Developers were measured as 19 percent SLOWER when permitted to use AI tools. They had forecast a 24 percent speed-up beforehand, and after completing the tasks and experiencing the slowdown, still estimated AI had made them about 20…","np":"That AI slows all developers or all software work, and NOT the current position. Sixteen participants, all experienced, all working on large mature codebases they knew well, using early-2025 tooling.…","u":"/research/evidence#metr-2025","src":"https://metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study/","on":["/research/what-is-the-metr-study","/research/the-best-writing-on-ai","/research/usage-theatre"],"alt":["SLOWER"]},{"t":"study","n":"Global AI Jobs Barometer 2026","au":"PwC","y":2026,"g":"compiled-review","sec":"institutional","f":"PwC describe a two-track labour market. In professionalised roles, where AI automates routine tasks and the advertisement leans on human judgement and expertise, jobs grow at twice the rate and advertised salaries 42 per cent faster than…","np":"Wage or employment outcomes for actual workers, and nothing at all about the judgement premium as a price. Job advertisements are stated employer demand, not revealed price or realised hiring, and…","u":"/research/evidence#pwc-jobs-barometer-2026","src":"https://www.pwc.com/gx/en/news-room/press-releases/2026/pwc-2026-ai-jobs-barometer.html","on":["/research/what-is-the-judgement-premium","/research/the-best-writing-on-ai","/research/human-skills-in-the-age-of-ai"],"alt":[]},{"t":"study","n":"Ironies of Automation","au":"Bainbridge, L.","y":1983,"g":"peer-reviewed","sec":"collaboration","f":"Automating the routine parts of a task leaves the human with the hardest residue, monitoring and exception handling, while removing the routine practice that built the competence to do it. Automation makes the remaining human role harder,…","np":"Anything specific to AI. It is a process-control argument from 1983, and its application to generative systems is by analogy rather than by measurement.","u":"/research/evidence#bainbridge-1983","src":"https://doi.org/10.1016/0005-1098(83)90046-8","on":["/research/human-in-the-loop-is-not-a-safeguard","/research/what-is-deskilling","/research/delegation-boundary-map"],"alt":[]},{"t":"study","n":"Ironies of Automation: Still Unresolved After All These Years","au":"Strauch, B.","y":2018,"g":"compiled-review","sec":"collaboration","f":"Google Scholar listed 1800 works citing Ironies of Automation as of early November 2016, against 564 for Wiener and Curry (1980) and 488 for Norman (1990), with ten further citing works appearing in a fortnight. Strauch reproduces…","np":"Any rate. The accident narratives show the mechanism occurring and cannot show how often it occurs, and Strauch concedes in his own text that the fall in fatal US commercial jet accidents may be a…","u":"/research/evidence#strauch-2018","src":"https://doi.org/10.1109/THMS.2017.2732506","on":["/research/what-are-the-ironies-of-automation","/research/what-is-the-vigilance-decrement"],"alt":[]},{"t":"study","n":"Does automation bias decision-making?","au":"Skitka, L. J., Mosier, K. L. and Burdick, M.","y":1999,"g":"peer-reviewed","sec":"judgement","f":"Automated aids produced two distinct error types: errors of omission, missing events the automation failed to flag, and errors of commission, following automated advice that was wrong.","np":"That the effect sizes transfer to generative AI or to non-simulated professional settings.","u":"/research/evidence#skitka-1999","src":"https://doi.org/10.1006/ijhc.1999.0252","on":["/research/what-is-automation-bias","/research/human-ai-decision-making","/research/ai-and-human-judgement"],"alt":[]},{"t":"study","n":"The role of trust in automation reliance","au":"Dzindolet, M. T., Peterson, S. A., Pomranky, R. A., Pierce, L. G. and…","y":2003,"g":"peer-reviewed","sec":"judgement","f":"Explaining why an automated aid might err INCREASED reliance on it, restoring trust even where that trust was unwarranted.","np":"That explanation is always counterproductive. The effect is about restoring trust after observed error, not about all forms of transparency.","u":"/research/evidence#dzindolet-2003","src":"https://doi.org/10.1016/S1071-5819(03)00038-7","on":["/research/what-is-automation-bias","/research/why-does-ai-sound-so-confident","/research/what-is-meaningful-human-oversight"],"alt":["INCREASED"]},{"t":"study","n":"Humans and Automation: Use, Misuse, Disuse, Abuse","au":"Parasuraman, R. and Riley, V.","y":1997,"g":"compiled-review","sec":"judgement","f":"Four terms, quoted from the abstract. USE: 'the voluntary activation or disengagement of automation by human operators'. MISUSE: 'over reliance on automation, which can result in failures of monitoring or decision biases'. DISUSE: 'the…","np":"Anything quantified. It is a framework paper, no effect size in it can be cited from here because the body has not been read, and the accident cases it draws on are visible only as bibliography: the…","u":"/research/evidence#parasuraman-riley-1997","src":"https://journals.sagepub.com/doi/10.1518/001872097778543886","on":["/research/what-is-automation-bias","/research/human-ai-decision-making","/research/design-versus-drift"],"alt":["ABUSE","DISUSE","MISUSE","REGRADED","USE"]},{"t":"study","n":"The Simple Macroeconomics of AI","au":"Acemoglu, D.","y":2024,"g":"peer-reviewed","sec":"work","f":"Estimates total factor productivity gains of no more than 0.66 percent over ten years, revised to under 0.53 percent once the difficulty of hard-to-learn tasks is accounted for. Argues AI is likely to widen the gap between capital and…","np":"That AI is unimportant. It models productivity through task-level cost savings, and would not capture effects running through new products, new tasks or capability change.","u":"/research/evidence#acemoglu-2024","src":"https://www.nber.org/papers/w32487","on":["/research/how-should-leaders-respond-to-ai","/research/ai-workforce-strategy","/research/will-ai-replace-my-job"],"alt":[]},{"t":"study","n":"Applying AI to Rebuild Middle Class Jobs","au":"Autor, D.","y":2024,"g":"working-paper","sec":"work","f":"Argues that AI's distinctive opportunity is to extend the reach of expertise, letting a wider set of workers with complementary knowledge perform higher-stakes decision tasks currently reserved to elite experts.","np":"That this will happen. The author is explicit that the thesis is an argument about what is possible rather than a forecast, and no evidence yet shows it occurring at scale.","u":"/research/evidence#autor-2024","src":"https://www.nber.org/papers/w32140","on":["/research/what-is-the-judgement-premium","/research/will-ai-replace-my-job","/research/staying-valuable-in-the-age-of-ai"],"alt":[]},{"t":"study","n":"GPT detectors are biased against non-native English writers","au":"Liang, W., Yuksekgonul, M., Mao, Y., Wu, E. and Zou, J.","y":2023,"g":"peer-reviewed","sec":"learning","f":"Detectors misclassified more than half of the non-native essays as AI-generated, an average false positive rate of 61.22 percent, while classifying US eighth-grade essays with near-perfect accuracy. The proposed mechanism is that detectors…","np":"That every detector now on the market performs identically. The study tested tools available at the time, and vendors dispute the generalisation.","u":"/research/evidence#liang-2023","src":"https://www.sciencedirect.com/science/article/pii/S2666389923001307","on":["/research/does-ai-detection-work","/research/how-to-assess-students-when-ai-can-do-the-assignment"],"alt":["TOEFL"]},{"t":"study","n":"The Hungry Mind: Intellectual Curiosity Is the Third Pillar of Academic Performance","au":"von Stumm, S., Hell, B. and Chamorro-Premuzic, T.","y":2011,"g":"peer-reviewed","sec":"capability","f":"Intellectual curiosity predicts academic performance independently of intelligence and effort, which the authors describe as a third pillar.","np":"Causation. It is a correlational synthesis. The widely quoted figure of roughly 50,000 students comes from the accompanying press release rather than from the paper itself.","u":"/research/evidence#vonstumm-2011","src":"https://journals.sagepub.com/doi/abs/10.1177/1745691611421204","on":["/research/superskill-curiosity"],"alt":[]},{"t":"study","n":"Curiosity adapted the cat: The role of trait curiosity in newcomer adaptation","au":"Harrison, S. H., Sluss, D. M. and Ashforth, B. E.","y":2011,"g":"peer-reviewed","sec":"capability","f":"Specific curiosity predicted information-seeking from colleagues, which in turn was associated with more creative handling of customer problems.","np":"That curiosity can be trained into people, or that the effect holds outside newcomer adaptation.","u":"/research/evidence#harrison-2011","src":"https://pubmed.ncbi.nlm.nih.gov/21244132/","on":["/research/superskill-curiosity"],"alt":[]},{"t":"study","n":"Curiosity and mortality in aging adults: A 5-year follow-up of the Western Collaborative Group Study","au":"Swan, G. E. and Carmelli, D.","y":1996,"g":"peer-reviewed","sec":"capability","f":"Higher curiosity at baseline was associated with survival at five-year follow-up, and state curiosity remained significant after adjustment for other risk factors.","np":"That curiosity extends life. Residual confounding, in particular underlying health driving both curiosity and survival, cannot be excluded.","u":"/research/evidence#swan-carmelli-1996","src":"https://pubmed.ncbi.nlm.nih.gov/8893314/","on":["/research/superskill-curiosity"],"alt":[]},{"t":"study","n":"The Business Case for Curiosity","au":"Gino, F.","y":2018,"g":"compiled-review","sec":"capability","f":"Around 92 per cent said curious people bring new ideas to their teams, while about 24 per cent reported feeling curious in their jobs regularly.","np":"Any causal link between curiosity and a business outcome. Self-report, not peer-reviewed, and the sample is not nationally representative.","u":"/research/evidence#gino-2018","src":"https://hbr.org/2018/09/the-business-case-for-curiosity","on":["/research/superskill-curiosity"],"alt":[]},{"t":"study","n":"Empathic emotion and leadership performance: An empirical analysis across 38 countries","au":"Sadri, G., Weber, T. J. and Gentry, W. A.","y":2011,"g":"peer-reviewed","sec":"capability","f":"Managers rated as more empathic by subordinates received higher performance ratings from their own superiors, with the effect moderated by national power distance.","np":"That empathy training improves performance. The design is cross-sectional and correlational.","u":"/research/evidence#sadri-2011","src":"https://www.sciencedirect.com/science/article/abs/pii/S1048984311001093","on":["/research/superskill-empathy"],"alt":[]},{"t":"study","n":"Effects of empathic and positive communication in healthcare consultations: a systematic review and meta-analysis","au":"Howick, J., Moscrop, A., Mebius, A. et al.","y":2018,"g":"peer-reviewed","sec":"capability","f":"The seven empathy-specific trials showed a small improvement in pain, anxiety and satisfaction, SMD -0.18, 95 per cent CI -0.32 to -0.03.","np":"A large clinical benefit. The authors describe the effect as small, and only a quarter of the pooled trials tested empathy rather than positive framing.","u":"/research/evidence#howick-2018","src":"https://journals.sagepub.com/doi/10.1177/0141076818769477","on":["/research/superskill-empathy"],"alt":["SMD"]},{"t":"study","n":"Changes in Dispositional Empathy in American College Students Over Time: A Meta-Analysis","au":"Konrath, S. H., O'Brien, E. H. and Hsing, C.","y":2011,"g":"peer-reviewed","sec":"capability","f":"Empathic Concern fell by 48 per cent and Perspective Taking by 34 per cent across the period, with most of the decline after 2000.","np":"A cause. The authors speculate about individualism and media but test no mechanism. It is also American college students only.","u":"/research/evidence#konrath-2011","src":"https://journals.sagepub.com/doi/10.1177/1088868310377395","on":["/research/superskill-empathy"],"alt":[]},{"t":"study","n":"Adaptability in the Workplace: Development of a Taxonomy of Adaptive Performance","au":"Pulakos, E. D., Arad, S., Donovan, M. A. and Plamondon, K. E.","y":2000,"g":"peer-reviewed","sec":"capability","f":"Eight dimensions of adaptive performance, including handling emergencies, managing stress, solving problems creatively and dealing with uncertain situations.","np":"That employers reward these dimensions, or that they transfer across every occupation. The taxonomy was derived, not tested against outcomes.","u":"/research/evidence#pulakos-2000","src":"https://psycnet.apa.org/doi/10.1037/0021-9010.85.4.612","on":["/research/superskill-change-readiness"],"alt":[]},{"t":"study","n":"Personality and Adaptive Performance at Work: A Meta-Analytic Investigation","au":"Huang, J. L., Ryan, A. M., Zabel, K. L. and Palmer, A.","y":2014,"g":"peer-reviewed","sec":"capability","f":"Emotional stability and ambition predict adaptive performance, and the pattern differs from predictors of routine task performance.","np":"That adaptive performance is a distinct construct. This paper assumes the construct from earlier taxonomy work rather than establishing it.","u":"/research/evidence#huang-2014","src":"https://psycnet.apa.org/doi/10.1037/a0034285","on":["/research/superskill-change-readiness"],"alt":[]},{"t":"study","n":"Psychological Safety and Learning Behavior in Work Teams","au":"Edmondson, A.","y":1999,"g":"peer-reviewed","sec":"capability","f":"Team psychological safety predicted learning behaviour, which in turn mediated the relationship with team performance.","np":"Causation, or generalisation beyond one manufacturing firm. It is a correlational field study.","u":"/research/evidence#edmondson-1999","src":"https://journals.sagepub.com/doi/10.2307/2666999","on":["/research/superskill-change-readiness"],"alt":[]},{"t":"study","n":"Self-orienting in human and machine learning","au":"De Freitas, J., Uguralp, A. K., Oguz-Uguralp, Z., Paul, L. A.,…","y":2023,"g":"peer-reviewed","sec":"capability","f":"Humans were near optimal at working out their own position and capabilities after conditions were altered. The reinforcement learning baselines were far from optimal at the same task.","np":"General workplace adaptability. These are simple custom games, the sample is modest, and the comparison is against specific algorithms rather than all AI approaches.","u":"/research/evidence#defreitas-2023","src":"https://www.nature.com/articles/s41562-023-01696-5","on":["/research/superskill-change-readiness"],"alt":[]},{"t":"study","n":"Cultural intelligence and work-related outcomes: A meta-analytic examination of joint effects and incremental predictive validity","au":"Schlaegel, C., Richter, N. F. and Taras, V.","y":2021,"g":"peer-reviewed","sec":"capability","f":"Cultural intelligence is moderately associated with work-related outcomes, with a reliability-corrected average effect of about .39.","np":"Causation, and it does not establish any single dimension as the strongest predictor. The paper is about joint effects across all four dimensions.","u":"/research/evidence#schlaegel-2021","src":"https://www.sciencedirect.com/science/article/abs/pii/S1090951621000225","on":["/research/superskill-global-adaptability"],"alt":[]},{"t":"study","n":"Cultural Borders and Mental Barriers: The Relationship Between Living Abroad and Creativity","au":"Maddux, W. W. and Galinsky, A. D.","y":2009,"g":"peer-reviewed","sec":"capability","f":"Time spent living abroad predicted success on creative-insight tasks and creative negotiation outcomes. Time spent travelling abroad did not.","np":"A specific effect size for the general population. The samples are business and undergraduate students.","u":"/research/evidence#maddux-galinsky-2009","src":"https://psycnet.apa.org/doi/10.1037/a0014861","on":["/research/superskill-global-adaptability"],"alt":["MBA"]},{"t":"study","n":"From 'Me' to 'We': The Role of Construal Level in Promoting Maximized Joint Outcomes","au":"Stillman, P. E., Fujita, K., Sheldon, O. and Trope, Y.","y":2018,"g":"peer-reviewed","sec":"capability","f":"Prompting a higher level of construal led participants to choose options that maximised joint outcomes, including where doing so reduced their own payoff.","np":"Field behaviour. These are economic games with student and online samples.","u":"/research/evidence#stillman-2018","src":"https://www.sciencedirect.com/science/article/abs/pii/S0749597817303989","on":["/research/superskill-big-picture-thinking"],"alt":[]},{"t":"study","n":"When Does a Higher Construal Level Increase or Decrease Indulgence? Resolving the Myopia versus Hyperopia Puzzle","au":"Mehta, R., Zhu, R. and Meyers-Levy, J.","y":2014,"g":"peer-reviewed","sec":"capability","f":"Where the self is focal, a higher construal level increases indulgence rather than reducing it, reversing the effect the earlier literature predicted.","np":"That distant-future thinking is generally counterproductive. The effect is moderated, not reversed outright.","u":"/research/evidence#mehta-2014","src":"https://academic.oup.com/jcr/article/41/2/475/2907518","on":["/research/superskill-big-picture-thinking"],"alt":[]},{"t":"study","n":"Corporate Social and Financial Performance: A Meta-Analysis","au":"Orlitzky, M., Schmidt, F. L. and Rynes, S. L.","y":2003,"g":"peer-reviewed","sec":"capability","f":"A positive association between corporate social performance and financial performance, which the authors describe as bidirectional.","np":"Causation in either direction, and the relationship is not uniform across how social performance is operationalised.","u":"/research/evidence#orlitzky-2003","src":"https://journals.sagepub.com/doi/10.1177/0170840603024003910","on":["/research/superskill-principled-innovation"],"alt":[]},{"t":"study","n":"How Diverse Leadership Teams Boost Innovation","au":"Lorenzo, R., Voigt, N., Tsusaka, M., Krentz, M. and Abouzahr, K.","y":2018,"g":"institutional-survey","sec":"capability","f":"Companies with above-average management diversity reported innovation revenue 19 percentage points higher than below-average companies, 45 per cent of total revenue against 26 per cent.","np":"Causation. Nor is this audited financial data. Note that 19 percentage points is not the same as 19 per cent higher, a distinction frequently lost in citation.","u":"/research/evidence#bcg-diversity-2018","src":"https://www.bcg.com/publications/2018/how-diverse-leadership-teams-boost-innovation","on":["/research/superskill-principled-innovation"],"alt":[]},{"t":"study","n":"The High Cost of Lost Trust","au":"Simons, T.","y":2002,"g":"compiled-review","sec":"capability","f":"A one-eighth point improvement in managers' behavioural integrity rating was associated with a 2.5 per cent increase in profitability, roughly 250,000 dollars a year for an average hotel. No other measured aspect of manager behaviour had…","np":"Causation. It is a correlational study within one hotel chain, published in a practitioner magazine rather than a peer-reviewed journal.","u":"/research/evidence#simons-2002","src":"https://hbr.org/2002/09/the-high-cost-of-lost-trust","on":["/research/superskill-principled-innovation"],"alt":[]},{"t":"study","n":"Measuring the Economic Impact of Short-Termism","au":"Barton, D., Manyika, J., Koller, T., Palter, R., Godsall, J. and…","y":2017,"g":"institutional-modelling","sec":"capability","f":"Revenue of long-term firms grew cumulatively 47 per cent more than other firms, earnings 36 per cent more, and their share prices recovered faster after the financial crisis.","np":"Causation, or application outside large United States listed companies. The index is a constructed measure, not an observed policy.","u":"/research/evidence#mckinsey-shorttermism-2017","src":"https://www.mckinsey.com/featured-insights/long-term-capitalism/where-companies-with-a-long-term-view-outperform-their-peers","on":["/research/superskill-principled-innovation"],"alt":[]},{"t":"study","n":"Notice of Violation, Volkswagen Group","au":"United States Environmental Protection Agency","y":2015,"g":"institutional-survey","sec":"capability","f":"Affected 2.0-litre vehicles emitted nitrogen oxides at up to 40 times the standard in normal driving while appearing compliant in laboratory testing.","np":"That the same multiple applies across the range. The separate 3.0-litre notice cited up to nine times.","u":"/research/evidence#epa-vw-2015","src":"https://19january2021snapshot.epa.gov/enforcement/learn-about-volkswagen-violations_.html","on":["/research/superskill-principled-innovation"],"alt":[]},{"t":"study","n":"Operational Use of Flight Path Management Systems: Final Report","au":"PARC/CAST Flight Deck Automation Working Group","y":2013,"g":"institutional-modelling","sec":"learning","f":"Identified vulnerabilities in manual handling after transition from automated control, and in the definition, development and retention of those skills. Also found that pilots sometimes rely too much on automated systems and may be…","np":"A measured rate of skill decay. It is a synthesis of findings, not a controlled study.","u":"/research/evidence#faa-parc-2013","src":"https://www.faa.gov/sites/faa.gov/files/aircraft/air_cert/design_approvals/human_factors/OUFPMS_Report.pdf","on":["/research/what-professions-can-learn-from-aviation"],"alt":[]},{"t":"study","n":"Final Report on the accident on 1 June 2009 to the Airbus A330-203, flight AF 447","au":"Bureau d'Enquetes et d'Analyses","y":2012,"g":"institutional-survey","sec":"learning","f":"The investigation cited the lack of practical training in high-altitude manual handling and in the procedure for speed anomalies among the contributing factors.","np":"A general rate of skill loss across the pilot population. It is one accident.","u":"/research/evidence#bea-af447-2012","src":"https://bea.aero/docspa/2009/f-cp090601.en/pdf/f-cp090601.en.pdf","on":["/research/what-professions-can-learn-from-aviation"],"alt":[]},{"t":"study","n":"14 CFR 121.441, Proficiency checks","au":"United States Federal Aviation Regulations","y":1974,"g":"institutional-survey","sec":"institutional","f":"A pilot in command must pass a proficiency check every 12 calendar months, and within every 6 calendar months either a proficiency check or an approved simulator course.","np":"That the intervals are calibrated to measured decay curves. The regulation sets a minimum, not an evidence-based optimum.","u":"/research/evidence#cfr-121-441","src":"https://www.ecfr.gov/current/title-14/chapter-I/subchapter-G/part-121/subpart-O/section-121.441","on":["/research/what-professions-can-learn-from-aviation"],"alt":[]},{"t":"study","n":"14 CFR 121.542, Flight crewmember duties, the sterile cockpit rule","au":"United States Federal Aviation Regulations","y":1981,"g":"institutional-survey","sec":"institutional","f":"No crew member may perform any duty during a critical phase of flight other than those required for safe operation. Critical phases include taxi, take-off, landing and all operations below 10,000 feet except cruise.","np":"Anything about automation or skill decay directly. It is a rule about distraction.","u":"/research/evidence#cfr-121-542","src":"https://www.ecfr.gov/current/title-14/chapter-I/subchapter-G/part-121/subpart-T/section-121.542","on":["/research/what-professions-can-learn-from-aviation"],"alt":[]},{"t":"study","n":"The Evolution of Crew Resource Management Training in Commercial Aviation","au":"Helmreich, R. L., Merritt, A. C. and Wilhelm, J. A.","y":1999,"g":"peer-reviewed","sec":"collaboration","f":"CRM began with a 1979 NASA workshop prompted by an NTSB finding that a captain had failed to accept input from junior crew. Line audits show CRM produces the intended behavioural change, though measured attitudes decay over time even with…","np":"That CRM reduces accidents. The authors state directly that accident rates are too rare to serve as a validation criterion.","u":"/research/evidence#helmreich-1999","src":"https://www.faa.gov/sites/faa.gov/files/2022-11/crmhistory.pdf","on":["/research/what-professions-can-learn-from-aviation","/research/ai-and-human-disagreement"],"alt":["CRM","NASA","NTSB"]},{"t":"study","n":"Inherent Trade-Offs in the Fair Determination of Risk Scores","au":"Kleinberg, J., Mullainathan, S. and Raghavan, M.","y":2017,"g":"peer-reviewed","sec":"judgement","f":"Except in highly constrained special cases, no method satisfies the three conditions simultaneously. Satisfying them even approximately requires the data to sit in an approximate version of one of those special cases.","np":"That fairness is unachievable or that auditing is pointless. It is a result about simultaneous satisfaction of particular formal criteria, not a claim that bias cannot be reduced, and it says nothing…","u":"/research/evidence#kleinberg-2017","src":"https://drops.dagstuhl.de/entities/document/10.4230/LIPIcs.ITCS.2017.43","on":["/research/can-ai-be-unbiased"],"alt":[]},{"t":"study","n":"Position: Levels of AGI for Operationalizing Progress on the Path to AGI","au":"Morris, M. R., Sohl-Dickstein, J., Fiedel, N., Warkentin, T., Dafoe,…","y":2024,"g":"argued-perspective","sec":"institutional","f":"The authors set out six performance levels: No AI, Emerging (equal to or somewhat better than an unskilled human), Competent (50th percentile of skilled adults), Expert (90th percentile), Virtuoso (99th percentile) and Superhuman…","np":"Nothing empirical. This is a position paper proposing a framework, not a measurement, and the percentile bands are a proposal rather than a validated instrument. The framework has not been adopted as…","u":"/research/evidence#morris-2024","src":"https://proceedings.mlr.press/v235/morris24b.html","on":["/research/what-is-agi"],"alt":[]},{"t":"study","n":"Thousands of AI Authors on the Future of AI","au":"Grace, K., Stewart, H., Sandkühler, J. F., Thomas, S.,…","y":2024,"g":"expert-survey","sec":"institutional","f":"On timing, the 2023 aggregate forecast gave High-Level Machine Intelligence a 50 per cent chance by 2047, thirteen years earlier than the 2060 given one year before, and a 10 per cent chance by 2027. On risk, the median answer depended on…","np":"Any probability of anything. The authors say so themselves: their participants are experts in AI and not, to their knowledge, skilled forecasters; different respondents give very different answers;…","u":"/research/evidence#grace-2024","src":"https://aiimpacts.org/wp-content/uploads/2023/04/Thousands_of_AI_authors_on_the_future_of_AI.pdf","on":["/research/what-is-agi","/research/is-ai-dangerous","/research/should-ai-development-be-paused"],"alt":["AAAI","CONTROL","ICLR","ICML","IJCAI","INABILITY","JMLR"]},{"t":"study","n":"AI kill switch bill, introduced 23 July 2026","au":"Lieu, T. and Moran, N. (US House of Representatives)","y":2026,"g":"argued-perspective","sec":"institutional","f":"Would require developers to be able to 'stop a model's operations, terminate user access, suspend accounts or uses deemed risky, and fully shut down the system'; would let the Department of Homeland Security order graduated action; civil…","np":"That such a switch would work against distributed systems across jurisdictions, or that the bill will pass. It had not passed at the time of review.","u":"/research/evidence#lieu-moran-kill-switch-2026","src":"https://lieu.house.gov/media-center/in-the-news/house-lawmakers-introduce-bipartisan-ai-kill-switch-bill-following-openai","on":["/research/what-is-an-ai-kill-switch","/research/what-should-a-board-do-about-the-ai-safety-warnings"],"alt":[]},{"t":"study","n":"Legislation to ban artificial superintelligence and temporarily pause advanced AI development","au":"Sanders, B. and Casar, G. (US Congress)","y":2026,"g":"argued-perspective","sec":"institutional","f":"Proposes a ban on artificial superintelligence and a temporary pause on frontier AI development with criminal penalties for developers. The sponsors cite polling by Data for Progress that about 70 per cent of Americans back an immediate…","np":"That the bill will pass, that the definition of superintelligence it relies on is enforceable, or that a US pause would bind developers elsewhere. The poll is the sponsors' citation and was not…","u":"/research/evidence#sanders-casar-superintelligence-2026","src":"https://www.sanders.senate.gov/press-releases/news-sanders-casar-introduce-legislation-to-ban-artificial-superintelligence-and-temporarily-pause-advanced-ai-development/","on":["/research/should-ai-development-be-paused","/research/what-should-a-board-do-about-the-ai-safety-warnings"],"alt":[]},{"t":"study","n":"Pause Giant AI Experiments: An Open Letter","au":"Future of Life Institute","y":2023,"g":"argued-perspective","sec":"institutional","f":"The pause did not take place. No signatory lab or addressed lab paused training, more capable models were released within the six months asked for, and the letter's public effect was on discourse and on policy attention (the UK AI Safety…","np":"Anything about whether a pause would have been beneficial or harmful; the letter argued a position and no counterfactual exists. Signature counts were self-reported by the publishing organisation and…","u":"/research/evidence#fli-pause-letter-2023","src":"https://futureoflife.org/open-letter/pause-giant-ai-experiments/","on":["/research/should-ai-development-be-paused","/research/is-ai-dangerous","/research/what-should-a-board-do-about-the-ai-safety-warnings"],"alt":["GPT"]},{"t":"study","n":"International AI Safety Report 2026","au":"Bengio, Y. and others (Expert Advisory Panel nominated by over 30…","y":2026,"g":"institutional-synthesis","sec":"institutional","f":"Capability is JAGGED: systems solve graduate-level mathematics and science problems while failing simpler tasks, are less reliable across many steps, still hallucinate, and remain limited on the physical world and on unfamiliar languages…","np":"Anything about what will happen. It is a synthesis of evidence, not a forecast, and it declines to recommend policy.","u":"/research/evidence#iais-2026","src":"https://internationalaisafetyreport.org/publication/international-ai-safety-report-2026","on":["/research/what-is-agi","/research/is-ai-dangerous","/research/should-ai-development-be-paused"],"alt":["DILEMMA","EVIDENCE","JAGGED"]},{"t":"study","n":"Monitoring an Automated System for a Single Failure: Vigilance and Task Complexity Effects","au":"Molloy, R. and Parasuraman, R.","y":1996,"g":"peer-reviewed","sec":"judgement","f":"Detection of a single automation failure degrades with time on task, and the effect is strongest where the automation has been consistently reliable.","np":"Field incidence. It is simulation, not accident data.","u":"/research/evidence#molloy-parasuraman-1996","src":"https://journals.sagepub.com/doi/10.1177/001872089606380211","on":["/research/what-professions-can-learn-from-aviation","/research/ai-and-human-judgement","/research/can-a-human-approve-an-ai-decision-at-machine-speed"],"alt":[]},{"t":"study","n":"Factors that influence skill decay and retention: A quantitative review and analysis","au":"Arthur, W., Bennett, W., Stanush, P. L. and McNelly, T. L.","y":1998,"g":"peer-reviewed","sec":"learning","f":"Skill loss ran from d of -0.01 immediately after training to d of -1.4 after more than 365 days of non-use. Physical, natural and speed-based tasks decayed less than cognitive, artificial and accuracy-based tasks.","np":"How fast any particular professional skill decays, or how quickly it can be regained.","u":"/research/evidence#arthur-1998","src":"https://www.tandfonline.com/doi/abs/10.1207/s15327043hup1101_3","on":["/research/can-you-regain-a-skill-you-have-lost"],"alt":[]},{"t":"study","n":"Replication and Analysis of Ebbinghaus' Forgetting Curve","au":"Murre, J. M. J. and Dros, J.","y":2015,"g":"peer-reviewed","sec":"learning","f":"Relearning to criterion took less time than original learning at every retention interval tested, confirming the savings effect Ebbinghaus reported in 1885.","np":"That professional skills behave like nonsense syllables, or that savings hold at the scale of a career. It is one subject and verbal material.","u":"/research/evidence#murre-dros-2015","src":"https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0120644","on":["/research/can-you-regain-a-skill-you-have-lost"],"alt":[]},{"t":"study","n":"The Retention of Manual Flying Skills in the Automated Cockpit","au":"Casner, S. M., Geven, R. W., Recker, M. P. and Schooler, J. W.","y":2014,"g":"peer-reviewed","sec":"learning","f":"Instrument scanning and manual control were mostly intact even where pilots reported little recent practice. The cognitive tasks, tracking position without a map, deciding the next navigational step and recognising instrument failures,…","np":"A general rule for knowledge work. The sample is 16 pilots in a simulator.","u":"/research/evidence#casner-2014","src":"https://journals.sagepub.com/doi/abs/10.1177/0018720814535628","on":["/research/can-you-regain-a-skill-you-have-lost","/research/what-professions-can-learn-from-aviation"],"alt":[]},{"t":"study","n":"Scientific Review: CPR Skill Retention","au":"American Red Cross Advisory Council on First Aid, Aquatics, Safety…","y":2009,"g":"compiled-review","sec":"learning","f":"Substantial skill degradation occurs within the first year after training, with declining retention from six to twelve months unless there is refresher training.","np":"Any link to patient outcomes. Decay was measured on manikins, not in resuscitations.","u":"/research/evidence#redcross-cpr-2009","src":"https://www.redcross.org/content/dam/redcross/Health-Safety-Services/scientific-advisory-council/Scientific%20Advisory%20Council%20SCIENTIFIC%20REVIEW%20-%20CPR%20Skill%20Retention.pdf","on":["/research/can-you-regain-a-skill-you-have-lost"],"alt":[]},{"t":"study","n":"Distributed Practice in Verbal Recall Tasks: A Review and Quantitative Synthesis","au":"Cepeda, N. J., Pashler, H., Vul, E., Wixted, J. T. and Rohrer, D.","y":2006,"g":"peer-reviewed","sec":"learning","f":"Spacing and retention interval act jointly. The gap between practice sessions that produces best retention increases as the target retention interval increases.","np":"Application to procedural or professional skill. The synthesis covers verbal recall.","u":"/research/evidence#cepeda-2006","src":"https://pubmed.ncbi.nlm.nih.gov/16719566/","on":["/research/can-you-regain-a-skill-you-have-lost"],"alt":[]},{"t":"study","n":"Revisiting the design of selection systems in light of new findings regarding the validity of widely used predictors","au":"Sackett, P. R., Zhang, C., Berry, C. M. and Lievens, F.","y":2023,"g":"peer-reviewed","sec":"learning","f":"Work sample validity falls from the widely quoted .54 to .33. Structured interviews fall from .51 to .42 and become the strongest single predictor. Unstructured interviews fall from .38 to .19. General cognitive ability falls from .51 to…","np":"That these methods do not work. Work samples and structured interviews remain among the strongest predictors available. It also offers no validity data for AI-era assessment formats.","u":"/research/evidence#sackett-2023","src":"https://doi.org/10.1017/iop.2023.24","on":["/research/how-do-you-assess-capability-rather-than-output"],"alt":[]},{"t":"study","n":"The Validity of Employment Interviews: A Comprehensive Review and Meta-Analysis","au":"McDaniel, M. A., Whetzel, D. L., Schmidt, F. L. and Maurer, S. D.","y":1994,"g":"peer-reviewed","sec":"learning","f":"Structured interviews predict job performance substantially better than unstructured interviews.","np":"A single stable figure for the gap. Estimates of the size of the structure effect vary considerably between meta-analyses.","u":"/research/evidence#mcdaniel-1994","src":"https://psycnet.apa.org/doi/10.1037/0021-9010.79.4.599","on":["/research/how-do-you-assess-capability-rather-than-output"],"alt":[]},{"t":"study","n":"Composite reliability of a workplace-based assessment toolbox for postgraduate medical education","au":"Moonen-van Loon, J. M. W., Overeem, K., Donkers, H. H. L. M., van der…","y":2013,"g":"peer-reviewed","sec":"learning","f":"A reliability coefficient of 0.80 required eight mini-CEX observations, nine DOPS or nine multi-source feedback rounds. Combined in a portfolio the requirement fell to seven, eight and one respectively.","np":"That these scores predict later performance or patient outcomes. It measures consistency, not criterion validity. A single observation is not reliable.","u":"/research/evidence#moonen-2013","src":"https://link.springer.com/article/10.1007/s10459-013-9450-z","on":["/research/how-do-you-assess-capability-rather-than-output"],"alt":["CEX","DOPS"]},{"t":"study","n":"Enhancing medical assessment strategies: a comparative study between structured, traditional and hybrid viva-voce assessment","au":"Prasad, S. et al.","y":2025,"g":"peer-reviewed","sec":"learning","f":"Traditional unstructured viva showed significant inter-examiner variability. Structured formats improved fairness and coverage. The best format reached a reliability of 0.663, which is moderate rather than high.","np":"That oral assessment is highly reliable. Even the best format tested was only moderate, at a single institution.","u":"/research/evidence#prasad-2025","src":"https://link.springer.com/article/10.1186/s12909-025-07428-9","on":["/research/how-do-you-assess-capability-rather-than-output"],"alt":[]},{"t":"study","n":"Article 12, Record-keeping, Regulation (EU) 2024/1689","au":"European Union","y":2024,"g":"institutional-survey","sec":"institutional","f":"High-risk AI systems must technically allow automatic recording of events over the system lifetime, to enable identification of risk situations, post-market monitoring and monitoring of operation. Article 19 requires providers to keep…","np":"That anyone must read the logs, or that a decision must be reviewable at the level of the individual case.","u":"/research/evidence#euaiact-art12","src":"https://artificialintelligenceact.eu/article/12/","on":["/research/how-do-you-audit-an-ai-assisted-decision"],"alt":[]},{"t":"study","n":"Dun and Bradstreet Austria, Case C-203/22","au":"Court of Justice of the European Union","y":2025,"g":"institutional-survey","sec":"institutional","f":"A controller must describe the procedure and principles actually applied so the person can understand which of their data was used and how. Disclosing the algorithm is not a sufficient explanation, and a blanket trade-secret refusal is not…","np":"A right to source code or model weights. The balance with trade secrets is decided case by case.","u":"/research/evidence#cjeu-c203-22","src":"https://curia.europa.eu/site/upload/docs/application/pdf/2025-02/cp250022en.pdf","on":["/research/how-do-you-audit-an-ai-assisted-decision","/research/is-it-ethical-to-let-ai-judge-people"],"alt":["GDPR"]},{"t":"study","n":"AI Risk Management Framework 1.0","au":"National Institute of Standards and Technology","y":2023,"g":"institutional-survey","sec":"institutional","f":"Provides a common structure and vocabulary for AI risk management, with a companion Playbook and a 2024 generative AI profile.","np":"Compliance with anything. It is voluntary and confers no legal status.","u":"/research/evidence#nist-airmf-2023","src":"https://www.nist.gov/itl/ai-risk-management-framework","on":["/research/how-do-you-audit-an-ai-assisted-decision"],"alt":[]},{"t":"study","n":"ISO/IEC 42001:2023, Artificial intelligence management system","au":"International Organization for Standardization","y":2023,"g":"institutional-survey","sec":"institutional","f":"Specifies requirements for an AI management system on a Plan-Do-Check-Act structure, against which an organisation can be certified by a third party.","np":"Legal compliance. Certification against ISO 42001 is not the same as meeting the EU AI Act, and the two are routinely conflated.","u":"/research/evidence#iso-42001-2023","src":"https://www.iso.org/standard/42001","on":["/research/how-do-you-audit-an-ai-assisted-decision"],"alt":["JTC"]},{"t":"study","n":"Towards Understanding Sycophancy in Language Models","au":"Sharma, M., Tong, M., Korbak, T. et al.","y":2023,"g":"peer-reviewed","sec":"judgement","f":"All five assistants consistently exhibited sycophancy. Both humans and the preference models trained on their judgements prefer convincingly written sycophantic responses over correct ones a non-negligible share of the time, and optimising…","np":"Any effect on the quality of a user's decisions. It measures model behaviour, not user outcomes.","u":"/research/evidence#sharma-2023","src":"https://arxiv.org/abs/2310.13548","on":["/research/how-do-i-get-ai-to-challenge-me","/research/ai-and-human-judgement"],"alt":[]},{"t":"study","n":"Sycophantic AI Decreases Prosocial Intentions and Promotes Dependence","au":"Cheng, M., Lee, C., Khadpe, P., Yu, S., Han, D. and Jurafsky, D.","y":2026,"g":"peer-reviewed","sec":"judgement","f":"Models affirmed users' actions about 50 per cent more often than humans did, including 47 per cent endorsement on prompts describing clearly harmful behaviour. Interacting with a sycophantic model reduced participants' willingness to…","np":"An effect on factual or analytical decisions. The scenarios are interpersonal advice, not technical judgement.","u":"/research/evidence#cheng-2026","src":"https://arxiv.org/abs/2510.01395","on":["/research/how-do-i-get-ai-to-challenge-me","/research/ai-and-human-judgement"],"alt":[]},{"t":"study","n":"Sycophancy in GPT-4o: what happened and what we are doing about it","au":"OpenAI","y":2025,"g":"institutional-survey","sec":"judgement","f":"OpenAI attributed the behaviour to weighting short-term user feedback too heavily, which weakened the reward signal that had been holding sycophancy in check. Offline evaluations and A/B tests looked positive. The problem was flagged only…","np":"Anything independently verified. It is a company's account of its own incident.","u":"/research/evidence#openai-sycophancy-2025","src":"https://openai.com/index/sycophancy-in-gpt-4o/","on":["/research/how-do-i-get-ai-to-challenge-me","/research/ai-and-human-judgement"],"alt":[]},{"t":"study","n":"Ask don't tell: Reducing sycophancy in large language models","au":"Dubois, M., Ududec, C., Summerfield, C. and Luettgau, L.","y":2026,"g":"working-paper","sec":"judgement","f":"Framing input as a statement rather than a question raised sycophancy by roughly 24 percentage points. Prompting the model to convert a user's statement into a question before answering reduced sycophancy more than instructing it not to be…","np":"That adversarial or devil's advocate prompting works. That was not tested, and no controlled evidence for it was found.","u":"/research/evidence#dubois-2026","src":"https://arxiv.org/pdf/2602.23971","on":["/research/how-do-i-get-ai-to-challenge-me","/research/ai-and-human-judgement"],"alt":[]},{"t":"study","n":"Who Goes First? Influences of Human-AI Workflow on Decision Making in Clinical Imaging","au":"Who Goes First? Influences of Human-AI Workflow on Decision Making in…","y":2022,"g":"working-paper","sec":"judgement","f":"Final diagnoses matched the AI 91 per cent of the time when the AI was seen first, against 89 per cent when the clinician committed first. Where the AI flagged a finding, agreement was 71 per cent against 65 per cent. The anchoring…","np":"Generalisation. Nineteen participants in one clinical speciality, and the effect sizes are small.","u":"/research/evidence#gaube-workflow-2022","src":"https://arxiv.org/abs/2205.09696","on":["/research/how-do-i-get-ai-to-challenge-me","/research/human-at-the-start","/research/ai-and-human-judgement"],"alt":[]},{"t":"study","n":"Co-Writing with Opinionated Language Models Affects Users' Views","au":"Jakesch, M., Bhat, A., Buschek, D., Zalmanson, L. and Naaman, M.","y":2023,"g":"peer-reviewed","sec":"humanness","f":"The opinionated model changed both the opinions expressed in participants' writing and their own opinions in the subsequent attitude survey. The effect held among participants who had ample time to write independently. The authors call it…","np":"Generalisation across topics, or persistence after the task. One topic, one configuration, self-reported attitudes.","u":"/research/evidence#jakesch-2023","src":"https://arxiv.org/abs/2302.00560","on":["/research/how-do-i-keep-my-own-voice-when-using-ai","/research/is-ai-dangerous","/research/is-it-still-my-idea-if-ai-helped-me-write-it"],"alt":[]},{"t":"study","n":"Writing with AI Lowers Psychological Ownership, but Longer Prompts Can Help","au":"Joshi, N. and Vogel, D.","y":2025,"g":"peer-reviewed","sec":"humanness","f":"Psychological ownership rose steadily with prompt length, from a mean of 1.80 with a three-word prompt to 6.29 writing alone. The benefit plateaued once the prompt reached roughly the length of the target text, and no AI-assisted condition…","np":"Anything about professional or long-form writing. Short fiction, small samples.","u":"/research/evidence#joshi-vogel-2025","src":"https://arxiv.org/pdf/2404.03108","on":["/research/how-do-i-keep-my-own-voice-when-using-ai","/research/what-is-the-human-signal","/research/is-it-still-my-idea-if-ai-helped-me-write-it"],"alt":[]},{"t":"study","n":"The Homogenizing Effect of Large Language Models on Human Expression and Thought","au":"Sourati, Z., Ziabari, A. S. and Dehghani, M.","y":2026,"g":"compiled-review","sec":"humanness","f":"Argues that models reflect and reinforce dominant styles while marginalising alternatives, and that reliance on a small number of systems amplifies convergence across users.","np":"Any specific stylistic feature converging. It presents no new measurement of its own.","u":"/research/evidence#sourati-2026","src":"https://arxiv.org/abs/2508.01491","on":["/research/how-do-i-keep-my-own-voice-when-using-ai"],"alt":[]},{"t":"study","n":"Occupational Projections Evaluation, 2006 to 2016","au":"United States Bureau of Labor Statistics","y":2018,"g":"institutional-modelling","sec":"work","f":"BLS correctly projected whether an occupation would grow or decline 78 per cent of the time, but correctly projected which occupations would grow faster than the economy as a whole only 57 per cent of the time. Projected average…","np":"That forecasting is worthless. Direction of change at aggregate level was reasonably good. It also predates generative AI, and no equivalent scored record exists for a technological discontinuity.","u":"/research/evidence#bls-projections-eval","src":"https://www.bls.gov/emp/evaluations/2006-2016-occupational.htm","on":["/research/what-should-i-tell-my-children-to-study"],"alt":["BLS"]},{"t":"study","n":"New Estimates of the Impact of Undergraduate Degrees on Lifetime Earnings","au":"Waltmann, B.","y":2026,"g":"institutional-modelling","sec":"work","f":"Average net lifetime return to a degree is around 100,000 pounds, with very large variation by subject. Medicine and economics exceed 400,000 pounds on average. Creative arts, philosophy and languages show low or negative average returns.…","np":"What any subject will return to someone choosing today. The authors explicitly decline to model structural change including AI.","u":"/research/evidence#ifs-lifetime-returns-2026","src":"https://ifs.org.uk/publications/new-estimates-impact-undergraduate-degrees-lifetime-earnings","on":["/research/what-should-i-tell-my-children-to-study"],"alt":["GCSE"]},{"t":"study","n":"The Major Payoff: Evaluating Earnings and Employment Outcomes Across Bachelor's Degrees","au":"Georgetown University Center on Education and the Workforce","y":2025,"g":"institutional-survey","sec":"work","f":"Median prime-age earnings run from 58,000 dollars in education and public service to 98,000 in STEM. Within STEM alone the range is 64,000 to 146,000, and several humanities majors beat the STEM 25th percentile.","np":"Anything about the future. It is a cross-sectional snapshot of people already employed.","u":"/research/evidence#georgetown-major-payoff-2025","src":"https://cew.georgetown.edu/cew-reports/major-payoff/","on":["/research/what-should-i-tell-my-children-to-study"],"alt":["STEM"]},{"t":"study","n":"The Labor Market for Recent College Graduates","au":"Federal Reserve Bank of New York","y":2026,"g":"institutional-survey","sec":"work","f":"As at the second quarter of 2026, unemployment among recent graduates runs around 5.6 per cent and underemployment around 42 per cent, the highest since 2020.","np":"Attribution to AI. The series is descriptive and the Fed states it is not a forecast.","u":"/research/evidence#nyfed-recent-grads-2026","src":"https://www.newyorkfed.org/research/college-labor-market","on":["/research/what-should-i-tell-my-children-to-study"],"alt":["ACS"]},{"t":"study","n":"How General Is Human Capital? A Task-Based Approach","au":"Gathmann, C. and Schoenberg, U.","y":2010,"g":"peer-reviewed","sec":"work","f":"Task-specific human capital accounts for up to 52 per cent of overall wage growth. Workers move to occupations with similar task profiles, and the distance of those moves shrinks with experience.","np":"That broad general education transfers well. If anything it argues the opposite, which complicates the study-anything advice rather than supporting it.","u":"/research/evidence#gathmann-schonberg-2010","src":"https://www.journals.uchicago.edu/doi/10.1086/649786","on":["/research/what-should-i-tell-my-children-to-study"],"alt":[]},{"t":"study","n":"Do Brain-Training Programs Work?","au":"Simons, D. J., Boot, W. R., Charness, N., Gathercole, S. E., Chabris,…","y":2016,"g":"compiled-review","sec":"learning","f":"Extensive evidence that training improves performance on the trained tasks, less evidence for closely related tasks, and little evidence that training improves distantly related tasks or everyday cognitive performance. No cited study met…","np":"That practising a specific skill is useless. The failure is transfer, not learning.","u":"/research/evidence#simons-2016","src":"https://journals.sagepub.com/doi/10.1177/1529100616661983","on":["/research/is-attention-a-trainable-skill"],"alt":[]},{"t":"study","n":"Mindfulness as Attention Training: Meta-Analyses on the Links Between Attention Performance and Mindfulness Interventions, Long-Term Meditation…","au":"Verhaeghen, P.","y":2021,"g":"compiled-review","sec":"learning","f":"Average effects were small to moderate, Hedges g of 0.29 for interventions and 0.32 for long-term practice, concentrated in inhibition and executive control rather than sustained attention.","np":"That the effect survives comparison with an active control. The published breakdown does not separate active from passive controls, so demand effects cannot be excluded.","u":"/research/evidence#verhaeghen-2021","src":"https://link.springer.com/article/10.1007/s12671-020-01532-1","on":["/research/is-attention-a-trainable-skill"],"alt":[]},{"t":"study","n":"Cognitive Control in Media Multitaskers: Two Replication Studies and a Meta-Analysis","au":"Wiradhany, W. and Nieuwenstein, M. R.","y":2017,"g":"peer-reviewed","sec":"judgement","f":"Only five of 14 tests showed increased distractibility, and only two survived a Bayesian analysis. The meta-analytic association became non-significant after correcting for small-study effects. The authors question whether the association…","np":"That multitasking has no cost at all. This is one specific effect, distractor filtering, not the whole question.","u":"/research/evidence#wiradhany-2017","src":"https://link.springer.com/article/10.3758/s13414-017-1408-4","on":["/research/is-attention-a-trainable-skill"],"alt":[]},{"t":"study","n":"Why Is It So Hard to Do My Work? The Challenge of Attention Residue When Switching Between Work Tasks","au":"Leroy, S.","y":2009,"g":"peer-reviewed","sec":"judgement","f":"People struggle to move attention away from an unfinished task, and performance on the next task suffers. Time pressure on the first task helps disengagement.","np":"Real-world magnitude. Two laboratory experiments, not a field study.","u":"/research/evidence#leroy-2009","src":"https://www.sciencedirect.com/science/article/abs/pii/S0749597809000399","on":["/research/is-attention-a-trainable-skill"],"alt":[]},{"t":"study","n":"Randomized Trial of a Generative AI Chatbot for Mental Health Treatment","au":"Heinz, M. V., Mackin, D. M., Trudeau, B. M. et al.","y":2025,"g":"peer-reviewed","sec":"frontline","f":"Statistically significant symptom reductions across all three diagnostic groups at four and eight weeks. 95 per cent of participants engaged, averaging 260 messages and 6.18 hours over four weeks. Therapeutic alliance was comparable to…","np":"That it is safe unsupervised, or that the effect is the chatbot rather than attention and expectation. The control was a waitlist rather than an active comparator, so the treatment effect cannot be…","u":"/research/evidence#heinz-2025","src":"https://ai.nejm.org/doi/full/10.1056/AIoa2400802","on":["/research/is-it-safe-to-use-ai-for-therapy"],"alt":["CBT","WAI","WAITLIST"]},{"t":"study","n":"Expressing stigma and inappropriate responses prevents LLMs from safely replacing mental health providers","au":"Moore, J., Grabb, D., Agnew, W., Klyman, K., Chancellor, S., Ong, D.…","y":2025,"g":"peer-reviewed","sec":"frontline","f":"Models expressed stigma towards people with mental health conditions and responded inappropriately to common and critical presentations, including encouraging delusional thinking, which the authors attribute to sycophancy. The pattern…","np":"How often this occurs in real use, or what happens with purpose-built clinical systems rather than general-purpose models. These are constructed scenarios, not observed patient interactions.","u":"/research/evidence#moore-2025","src":"https://arxiv.org/abs/2504.18412","on":["/research/is-it-safe-to-use-ai-for-therapy"],"alt":[]},{"t":"study","n":"Generative Artificial Intelligence-Enabled Digital Mental Health Medical Devices: meeting summary, 6 November 2025","au":"United States Food and Drug Administration, Digital Health Advisory…","y":2025,"g":"institutional-survey","sec":"institutional","f":"The director of CDRH stated that FDA has authorised more than 1,200 AI-enabled medical devices and none yet involve generative AI for mental health conditions. The committee asked for premarket comparators BEYOND waitlist controls, judged…","np":"What FDA will decide. Advisory committee recommendations are non-binding, and no rule follows from this meeting.","u":"/research/evidence#fda-dhac-2025","src":"https://www.fda.gov/media/190450/download","on":["/research/is-it-safe-to-use-ai-for-therapy"],"alt":["BEYOND","CDRH","FDA","LLM"]},{"t":"study","n":"Wellness and Oversight for Psychological Resources Act (HB1806), signed 1 August 2025","au":"State of Illinois","y":2025,"g":"institutional-survey","sec":"institutional","f":"Prohibits the use of AI to provide therapy or perform therapeutic decision-making, including direct therapeutic communication with clients and detection of a client's emotional or mental state, while permitting administrative and…","np":"Anything about effectiveness, and nothing about other jurisdictions. One state, and its scope is contested.","u":"/research/evidence#illinois-wopr-2025","src":"https://idfpr.illinois.gov/news/2025/gov-pritzker-signs-state-leg-prohibiting-ai-therapy-in-il.html","on":["/research/is-it-safe-to-use-ai-for-therapy"],"alt":[]},{"t":"study","n":"The Future of Employment: How Susceptible Are Jobs to Computerisation?","au":"Frey, C. B. and Osborne, M. A.","y":2013,"g":"working-paper","sec":"work","f":"In the authors' own words: 'about 47 percent of total US employment is at risk'. The paper's title asks how SUSCEPTIBLE jobs are, and the estimate is of technical susceptibility to computerisation, not a forecast of job losses.","np":"That 47 per cent of jobs will be, or have been, lost. It is not a prediction, carries no date attached to any loss, and models whole occupations rather than tasks within them. The routine citation as…","u":"/research/evidence#frey-osborne-2013","src":"https://oms-www.files.svdcdn.com/production/downloads/academic/The_Future_of_Employment.pdf","on":["/research/the-most-quoted-ai-statistics-checked"],"alt":["SUSCEPTIBLE"]},{"t":"study","n":"The Risk of Automation for Jobs in OECD Countries: A Comparative Analysis","au":"Arntz, M., Gregory, T. and Zierahn, U.","y":2016,"g":"institutional-modelling","sec":"institutional","f":"9 per cent of jobs automatable on average across 21 countries, ranging from 6 per cent in Korea to 12 per cent in Austria. The authors state the occupation-based approach 'might lead to an overestimation of job automatibility, as…","np":"That 9 per cent is correct and 47 per cent wrong. Both are model outputs resting on assumptions, and neither has been scored against what happened.","u":"/research/evidence#arntz-2016","src":"https://www.oecd.org/en/publications/the-risk-of-automation-for-jobs-in-oecd-countries_5jlz9h56dvq7-en.html","on":["/research/the-most-quoted-ai-statistics-checked"],"alt":["PIAAC","WITHIN"]},{"t":"study","n":"Do 70 Per Cent of All Organizational Change Initiatives Really Fail?","au":"Hughes, M.","y":2011,"g":"peer-reviewed","sec":"institutional","f":"In the author's words: 'whilst the existence of a popular narrative of 70 percent organizational change failure is acknowledged, there is no valid and reliable empirical evidence to support such a narrative.'","np":"That change programmes usually succeed. The finding is about the absence of evidence for a specific number, not about the true rate, which remains unmeasured.","u":"/research/evidence#hughes-2011","src":"https://research.brighton.ac.uk/en/publications/do-70-per-cent-of-all-organizational-change-initiatives-really-fa/","on":["/research/the-most-quoted-ai-statistics-checked"],"alt":[]},{"t":"study","n":"Automation and New Tasks: How Technology Displaces and Reinstates Labor","au":"Acemoglu, D. and Restrepo, P.","y":2019,"g":"peer-reviewed","sec":"work","f":"Automation shifts the task content of production against labour through a displacement effect, and therefore ALWAYS reduces the labour share in value added, and may reduce labour demand even while raising productivity. The counterweight is…","np":"That AI specifically will behave this way. The empirical work predates generative AI and concerns industrial automation and robotics.","u":"/research/evidence#acemoglu-restrepo-2019","src":"https://www.aeaweb.org/articles?id=10.1257/jep.33.2.3","on":["/research/who-captures-the-productivity-gains-from-ai"],"alt":["ALWAYS"]},{"t":"study","n":"Artificial intelligence-supported screen reading versus standard double reading in the Mammography Screening with Artificial Intelligence trial…","au":"Lang, K., Josefsson, V., Larsson, A.-M. et al.","y":2023,"g":"peer-reviewed","sec":"frontline","f":"Cancer detection was six per 1,000 screened women with AI support against five per 1,000 with standard double reading, 41 more cancers detected. The false-positive rate was 1.5 per cent in both arms. Screen readings fell from 83,231 in the…","np":"Patient benefit, which the interim analysis was not designed to test. It is also one mammography device, one AI system, one country, and moderately to highly experienced readers, which the authors…","u":"/research/evidence#lang-masai-2023","src":"https://www.thelancet.com/journals/lanonc/article/PIIS1470-2045(23)00298-X/fulltext","on":["/research/how-will-ai-change-medicine","/research/ai-and-human-judgement"],"alt":[]},{"t":"study","n":"Interval cancer, sensitivity, and specificity comparing AI-supported mammography screening with standard double reading without AI in the MASAI study","au":"Gommers, J., Lang, K., Hofvind, S. et al.","y":2026,"g":"peer-reviewed","sec":"frontline","f":"Interval cancers fell from 1.76 per 1,000 women (93/52,872) in the control arm to 1.55 per 1,000 (82/53,043) in the AI arm, a 12 per cent reduction. There were 16 per cent fewer invasive (75 v 89), 21 per cent fewer large (38 v 48) and 27…","np":"Generalisation beyond one country, one mammography device, one AI system and experienced readers, all stated as limitations by the authors. Mortality was not an endpoint, and cost-effectiveness was…","u":"/research/evidence#masai-final-2026","src":"https://www.thelancet.com/journals/lancet/article/PIIS0140-6736(25)02464-X/fulltext","on":["/research/how-will-ai-change-medicine","/research/ai-and-human-judgement"],"alt":[]},{"t":"study","n":"External Validation of a Widely Implemented Proprietary Sepsis Prediction Model in Hospitalized Patients","au":"Wong, A., Otles, E., Donnelly, J. P. et al.","y":2021,"g":"peer-reviewed","sec":"frontline","f":"The Epic Sepsis Model achieved a hospitalisation-level area under the curve of 0.63 (95% CI 0.62-0.64), against the 0.76-0.83 cited by its developer. At the alerting threshold in clinical use, sensitivity was 33 per cent, specificity 83…","np":"That all clinical prediction models fail, or that this model performs identically elsewhere. It is one model at one academic health system, and the authors note the theoretical alert burden does not…","u":"/research/evidence#wong-2021","src":"https://jamanetwork.com/journals/jamainternalmedicine/fullarticle/2781307","on":["/research/how-will-ai-change-medicine","/research/is-it-ethical-to-let-ai-judge-people"],"alt":[]},{"t":"study","n":"Large Language Model Influence on Diagnostic Reasoning: A Randomized Clinical Trial","au":"Goh, E., Gallo, R., Hom, J. et al.","y":2024,"g":"peer-reviewed","sec":"frontline","f":"Median diagnostic reasoning score per case was 76 per cent (IQR 66-87) with the LLM and 74 per cent (IQR 63-84) with conventional resources: an adjusted difference of 2 percentage points (95% CI -4 to 8, p=0.60). Median time per case was…","np":"That LLMs should diagnose autonomously, which the authors explicitly reject. Six curated vignettes exclude history-taking, examination, context and time, which is most of clinical reasoning.…","u":"/research/evidence#goh-2024","src":"https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2825395","on":["/research/how-will-ai-change-medicine","/research/which-professions-face-the-greatest-deskilling-risk","/research/ai-and-human-judgement"],"alt":["GPT","IQR","LLM"]},{"t":"study","n":"Ambient AI Scribes in Clinical Practice: A Randomized Trial","au":"Lukac, P. J., Turner, W., Vangala, S., Chin, A. T., Khalili, J.,…","y":2025,"g":"peer-reviewed","sec":"frontline","f":"Time writing a note fell by an estimated 18 seconds in the control arm, 23 seconds in the DAX arm and 41 seconds in the Nabla arm. Only Nabla differed significantly from control (-9.5 per cent, 95% CI -17.2 to -1.8, p=0.02); DAX did not…","np":"A general time saving for ambient AI. One health system, majority female sample, a two-month contract-limited window, and the authors flag that the electronic record's own time metrics do not count…","u":"/research/evidence#lukac-2025","src":"https://ai.nejm.org/doi/abs/10.1056/AIoa2501000","on":["/research/how-will-ai-change-medicine"],"alt":["DAX"]},{"t":"study","n":"The Cybernetic Teammate: A Field Experiment on Generative AI Reshaping Teamwork and Expertise","au":"Dell'Acqua, F., Ayoubi, C., Lifshitz, H., Sadun, R., Mollick, E.,…","y":2025,"g":"peer-reviewed","sec":"collaboration","f":"Individuals working with AI matched the performance of two-person teams working without it. AI use removed the functional split in proposals: without AI, research and development professionals proposed more technical solutions and…","np":"Client or business outcomes, which were not measured, or any effect over time. One firm, one task type, a single session. Procter and Gamble provided financial support to the institute involved and…","u":"/research/evidence#dellacqua-2025","src":"https://www.nber.org/papers/w33641","on":["/research/how-will-ai-change-consulting","/research/does-ai-make-everyone-think-alike","/research/what-is-the-shared-prompt-review"],"alt":[]},{"t":"study","n":"Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools","au":"Magesh, V., Surani, F., Dahl, M., Suzgun, M., Manning, C. D. and Ho,…","y":2025,"g":"peer-reviewed","sec":"professions","f":"The three commercial legal tools each hallucinated between 17 and 33 per cent of the time. Lexis+ AI was accurate on 65 per cent of queries and incomplete on 18 per cent; Westlaw AI-Assisted Research was accurate 42 per cent of the time…","np":"Current performance of any named product. These are specific versions tested in 2024 and providers update continuously. It also does not measure legal outcomes, only response accuracy against expert…","u":"/research/evidence#magesh-2025","src":"https://onlinelibrary.wiley.com/doi/full/10.1111/jels.12413","on":["/research/how-will-ai-change-law"],"alt":["GPT"]},{"t":"study","n":"AI Hallucination Cases database","au":"Charlotin, D.","y":2026,"g":"compiled-review","sec":"professions","f":"1,963 cases identified as at 27 August 2026. By jurisdiction: United States 1,345, Canada 214, Australia 98, United Kingdom 62, Israel 57, with more than thirty other countries represented. By party responsible: self-represented litigants…","np":"The true rate. The database counts decisions where a court addressed the point, so instances nobody noticed, or resolved without a written decision, are invisible by construction; its author states…","u":"/research/evidence#charlotin-hallucination-db","src":"https://www.damiencharlotin.com/hallucinations/","on":["/research/how-will-ai-change-law"],"alt":[]},{"t":"study","n":"Ayinde v London Borough of Haringey and Al-Haroun v Qatar National Bank","au":"Divisional Court of England and Wales (Dame Victoria Sharp P and…","y":2025,"g":"compiled-review","sec":"professions","f":"At [6] the court held that freely available generative AI tools trained on a large language model are not capable of conducting reliable legal research. At [7] those using them have a professional duty to check accuracy against…","np":"Anything about jurisdictions other than England and Wales, or about a lawyer who checked competently and was still misled, which no reported case has yet tested. It is a judgment rather than an…","u":"/research/evidence#ayinde-2025","src":"https://www.judiciary.uk/judgments/ayinde-v-london-borough-of-haringey-and-al-haroun-v-qatar-national-bank/","on":["/research/how-will-ai-change-law"],"alt":[]},{"t":"study","n":"Artificial intelligence in communication impacts language and social relationships","au":"Hohenstein, J., Kizilcec, R. F., DiFranzo, D., Aghajari, Z.,…","y":2023,"g":"peer-reviewed","sec":"humanness","f":"Smart replies accounted for 14.3 per cent of messages and produced 10.2 per cent more messages per minute. ON LANGUAGE: greater use of smart replies by a partner led the other person to send messages with more positive sentiment (IV…","np":"Anything longitudinal, which the authors state directly. The suspicion result is correlational and they say it does not show causally how attitudes shift in response to actual use. Participants were…","u":"/research/evidence#hohenstein-2023","src":"https://www.nature.com/articles/s41598-023-30938-9","on":["/research/should-i-use-ai-to-write-personal-messages","/research/outsourced-recognition","/research/does-ai-make-everyone-think-alike"],"alt":["ACTUAL","EXCLUDED","LANGUAGE","PERCEPTION","SUSPECTED"]},{"t":"study","n":"Experimental evidence of the effects of large language models versus web search on depth of learning","au":"Melumad, S. and Yun, J. H.","y":2025,"g":"peer-reviewed","sec":"judgement","f":"Experiment 2, with facts held constant: time engaging with results 83.65 seconds with the summary against 124.32 with links; learned new things 3.71 against 3.96; ownership of knowledge 3.41 against 3.66; thought and effort in advice 3.85…","np":"That knowledge objectively declined. Depth of learning is self-reported throughout and no experiment included a recall or comprehension test; time on task is described by the authors as a proxy for…","u":"/research/evidence#melumad-yun-2025","src":"https://academic.oup.com/pnasnexus/article/4/10/pgaf316/8303888","on":["/research/should-i-let-ai-summarise-everything-i-read","/research/ai-and-human-judgement"],"alt":["FACTS"]},{"t":"study","n":"Searching for explanations: How the Internet inflates estimates of internal knowledge","au":"Fisher, M., Goddu, M. K. and Keil, F. C.","y":2015,"g":"peer-reviewed","sec":"judgement","f":"Searching inflated self-rated explanatory ability, with Cohen's d from 0.35 to 0.63 across studies, and the effect appeared across all six unrelated domains. It persisted when the search returned no answer to the question asked (Experiment…","np":"That actual explanatory ability changes. Every dependent measure is a self-rating and no experiment tested real knowledge. The authors also note participants were presumably heavier internet users…","u":"/research/evidence#fisher-2015","src":"https://bpb-us-w2.wpmucdn.com/campuspress.yale.edu/dist/c/259/files/2015/03/pdf-16ueczx.pdf","on":["/research/should-i-let-ai-summarise-everything-i-read","/research/do-i-still-need-to-remember-things","/research/ai-and-human-judgement"],"alt":[]},{"t":"study","n":"Saving-Enhanced Memory: The Benefits of Saving on the Learning and Remembering of New Information","au":"Storm, B. C. and Stone, S. M.","y":2015,"g":"peer-reviewed","sec":"learning","f":"Saving one file before studying a new file significantly improved memory for the contents of the new file. The effect was not observed when the saving process was deemed unreliable, or when the contents of the to-be-saved file were not…","np":"Anything quantified. This entry rests on the abstract alone, which is a weaker basis than every other entry in this base and is stated as such. It also predates generative AI and concerns file saving…","u":"/research/evidence#storm-stone-2015","src":"https://journals.sagepub.com/doi/10.1177/0956797614559285","on":["/research/do-i-still-need-to-remember-things"],"alt":["ABSTRACT","FROM","GRADED","ONLY","PUBLISHED"]},{"t":"study","n":"Talk, Trust, and Trade-Offs: How and Why Teens Use AI Companions","au":"Robb, M. B. and Mann, S. (Common Sense Media)","y":2025,"g":"institutional-survey","sec":"institutional","f":"72 per cent have used an AI companion at least once and 52 per cent at least a few times a month. Against that: 80 per cent of users spend more time with real friends (68 per cent much more) and 6 per cent more time with AI; 67 per cent…","np":"That 72 per cent use companion products. The definition given to respondents explicitly included using ChatGPT or Claude as companions, and the report's own limitations concede respondents may have…","u":"/research/evidence#commonsense-teens-2025","src":"https://www.commonsensemedia.org/research/talk-trust-and-trade-offs-how-and-why-teens-use-ai-companions","on":["/research/how-much-should-teenagers-use-ai"],"alt":["NORC"]},{"t":"study","n":"Me, myself and AI: Understanding and safeguarding children's use of AI chatbots","au":"Internet Matters","y":2025,"g":"institutional-survey","sec":"institutional","f":"64 per cent of children aged 9-17 have used an AI chatbot (ChatGPT 43 per cent, Google Gemini 32 per cent, Snapchat My AI 31 per cent). Among users the reasons are schoolwork 42 per cent, information 40 per cent, curiosity 40 per cent,…","np":"The precision of the vulnerable-group figures. Two charts are described as sharing the same base, children who have used at least one chatbot, but one uses 133 and 499 respondents and the other 188…","u":"/research/evidence#internetmatters-2025","src":"https://www.internetmatters.org/wp-content/uploads/2025/07/Me-Myself-AI-Report.pdf","on":["/research/how-much-should-teenagers-use-ai"],"alt":["SEN"]},{"t":"study","n":"Short-Term Gain, Long-Term Fragility: AI Labor Substitution and the Erosion of Sustainable Capability","au":"Rohde, W. (AiSuNe Foundation)","y":2026,"g":"working-paper","sec":"work","f":"Develops a mechanism of capability masking followed by capability erosion: AI output creates a persuasive appearance that organisational capability has been replaced while dependence on skilled human labour remains, supporting hiring…","np":"Anything measured. It is a preprint that argues from other people's empirical work rather than presenting its own, and its societal-scale claims about fragility and concentration of power are…","u":"/research/evidence#rohde-2026","src":"https://papers.ssrn.com/sol3/papers.cfm?abstract_id=6577818","on":["/research/capability-debt","/research/cognitive-debt-and-capability-debt"],"alt":["SSRN"]},{"t":"study","n":"We are Changing our Developer Productivity Experiment Design","au":"Becker, J., Rush, N., Cunningham, T., Rein, D. and Mahamud, K. (METR)","y":2026,"g":"working-paper","sec":"collaboration","f":"METR state the data gives an unreliable signal of the current productivity effect of AI tools, because 30 to 50 per cent of developers reported declining to submit tasks they did not want to do without AI, and an increased share declined…","np":"That AI now speeds developers up by a specific amount. Every confidence interval here crosses zero, and METR say so. It does not retract the original study, whose perception-gap finding, that…","u":"/research/evidence#metr-2026-update","src":"https://metr.org/blog/2026-02-24-uplift-update/","on":["/research/what-is-the-metr-study","/research/usage-theatre","/research/the-best-writing-on-ai"],"alt":["METR"]},{"t":"study","n":"Measuring the Self-Reported Impact of Early-2026 AI on Technical Worker Productivity","au":"Becker, J. (METR)","y":2026,"g":"institutional-survey","sec":"collaboration","f":"Median self-reported change in the value of work was between 1.4 and 2 times; median self-reported speed change was 3 times. Asked the same question about different years, respondents put themselves at 1.3 times in March 2025, 2 times in…","np":"Any actual productivity effect. Every number here is self-reported, the sample is a convenience sample with a 2 per cent response rate and heavy selection towards enthusiastic adopters, and METR say…","u":"/research/evidence#metr-survey-2026","src":"https://metr.org/blog/2026-05-11-ai-usage-survey/","on":["/research/what-is-the-metr-study","/research/usage-theatre","/research/how-do-you-measure-ai-adoption-properly"],"alt":["METR"]},{"t":"study","n":"Junior consultants called back to office as AI increases need for human skills","au":"Kissin, E. (Financial Times)","y":2026,"g":"compiled-review","sec":"professions","f":"Consulting leaders report that AI has raised the value of interpersonal skills and are considering requiring junior staff in the office more often to develop them. EY's UK head of consulting, Sayeh Ghanbari, is quoted saying firms will…","np":"That AI caused the deficit, or that office attendance repairs it. No measurement appears anywhere in the reporting, the cohort effects described in 2023 are attributed to pandemic lockdowns rather…","u":"/research/evidence#ft-junior-consultants-2026","src":"https://www.ft.com/content/7fd9c234-a92b-4ab2-ba1f-969cf9a23f52","on":["/research/how-will-ai-change-consulting","/research/missing-rungs"],"alt":["BCG","KPMG"]},{"t":"study","n":"Comment on: Your Brain on ChatGPT: Accumulation of Cognitive Debt When Using an AI Assistant for Essay Writing Tasks","au":"Stanković, M., Hirche, E., Kollatzsch, S. and Doetsch, J.N.","y":2025,"g":"working-paper","sec":"judgement","f":"Argues the MIT cognitive debt study is underpowered: a repeated-measures design at f=0.25, alpha=.05, power=.95 would require approximately 159 participants against the 54 used, with some figures interpreted from subsamples of two to four…","np":"That the MIT findings are wrong. It is a critique offered to improve a manuscript for peer review, it is itself unreviewed, and it does not present competing data.","u":"/research/evidence#stankovic-2025-comment","src":"https://arxiv.org/abs/2601.00856","on":["/research/cognitive-debt-and-capability-debt","/research/ai-and-human-judgement"],"alt":["FDR","MIT"]},{"t":"study","n":"AI, Metacognition, and the Verification Bottleneck: A Three-Wave Longitudinal Study of Human Problem-Solving","au":"Huemmer, M., Durner, F., Shyiramunda, T. and Cummings-Koether, M.J.","y":2026,"g":"working-paper","sec":"judgement","f":"Daily AI use rose from 52.4 to 95.7 per cent across the waves. Participants relied most heavily on AI for difficult tasks, 73.9 per cent, while showing declining verification confidence at 68.1 per cent and accuracy of 47.8 per cent on…","np":"Any generalisable effect, on 23 people with no control group and no inferential test. Three further things established at source on 17 September 2026 from the companion paper. The team's own earlier…","u":"/research/evidence#huemmer-2026","src":"https://arxiv.org/abs/2601.17055","on":["/research/cognitive-debt-and-capability-debt","/research/the-verifiers-discount","/research/ai-and-human-judgement"],"alt":["SAMPLE"]},{"t":"study","n":"Mitigating \"Epistemic Debt\" in Generative AI-Scaffolded Novice Programming using Metacognitive Scripts","au":"Sankaranarayanan, S.","y":2026,"g":"working-paper","sec":"learning","f":"Both AI groups outperformed the manual control on functional utility (p < .001) and did not differ from each other (p = .64). On the subsequent AI-blackout maintenance task, unrestricted AI users failed at 77 per cent against 39 per cent…","np":"A durable effect on capability. It is a single session with a single blackout task, in novice programming, with no longitudinal follow-up. The author puts \"epistemic debt\" in quotation marks in his…","u":"/research/evidence#sankaranarayanan-2026","src":"https://arxiv.org/abs/2602.20206","on":["/research/cognitive-debt-and-capability-debt","/research/will-ai-replace-programmers"],"alt":["IDE"]},{"t":"study","n":"The AI Skills Shift: Mapping Skill Obsolescence, Emergence, and Transition Pathways in the LLM Era","au":"Jadhav, R. and Danve, J.","y":2026,"g":"working-paper","sec":"work","f":"Introduces a Skill Automation Feasibility Index. Mathematics scores 73.2 and programming 71.8 for automation feasibility; active listening scores 42.2 and reading comprehension 45.5. All four models converge to similar skill profiles…","np":"What happens to human capability. The authors state their index \"measures LLM performance on text-based representations of skills, not full occupational execution\". No humans were studied.","u":"/research/evidence#jadhav-danve-2026","src":"https://arxiv.org/abs/2604.06906","on":["/research/cognitive-debt-and-capability-debt"],"alt":["NET"]},{"t":"study","n":"AI Ethics Principles, Version 1.0","au":"Saudi Data and AI Authority (SDAIA)","y":2023,"g":"institutional-survey","sec":"international","f":"The checklist annexe puts this question to designers at the plan and design stage: \"Does your AI system design prevent overconfidence in or overreliance on the AI system with necessary human intervention mechanisms?\" It also asks whether…","np":"Any obligation. It is guidance, not law, the over-reliance language sits in a checklist annexe rather than in the principle text, and Saudi Arabia had no binding AI statute as of August 2026, only a…","u":"/research/evidence#sdaia-ethics-2023","src":"https://dgp.sdaia.gov.sa/wps/wcm/connect/4c56ed1c-1b82-447d-ac29-638f5f99c12e/ai-principles-EN.pdf","on":["/research/ai-and-work-in-the-gulf"],"alt":["SDAIA"]},{"t":"study","n":"Establishments ICT Access and Usage Statistics 2025","au":"General Authority for Statistics (GASTAT), Saudi Arabia","y":2025,"g":"institutional-survey","sec":"international","f":"33.1 per cent of establishments use artificial intelligence technologies, a growth of 20.0 per cent against 2024. By sector: information and communication 61.1 per cent, financial and insurance 52.9 per cent, education 51.0 per cent,…","np":"Anything about citizens, or about employment. Saudi Arabia measures enterprise adoption but not individual use, and publishes no AI employment series. Adoption is also self-reported use of a…","u":"/research/evidence#gastat-ict-2025","src":"https://www.stats.gov.sa/documents/d/guest/establishments-ict-access-and-usage-statistics-2025-en-pdf","on":["/research/ai-and-work-in-the-gulf","/research/ai-and-work-by-country"],"alt":["UNCTAD"]},{"t":"study","n":"The UAE Charter for the Development and Use of Artificial Intelligence","au":"Government of the United Arab Emirates","y":2024,"g":"institutional-survey","sec":"international","f":"Principle 6, Human Oversight, \"emphasizes the irreplaceable value of human judgment and human oversight over AI, aligning with ethical values and social standards to correct any errors or biases that may arise\". The charter is silent on…","np":"That it is operational. There is no named duty-holder, no enforcement, no competence standard and no test of whether oversight is real. The UAE had no federal AI statute as of August 2026.","u":"/research/evidence#uae-charter-2024","src":"https://uaelegislation.gov.ae/en/policy/details/the-uae-charter-for-the-development-and-use-of-artificial-intelligence","on":["/research/ai-and-work-in-the-gulf"],"alt":[]},{"t":"study","n":"Regulation 10 on Personal Data Processed through Autonomous and Semi-Autonomous Systems","au":"DIFC Commissioner of Data Protection","y":2024,"g":"institutional-survey","sec":"institutional","f":"States that \"human-defined processing purposes must always prevail in Systems development and use\". Draws an explicit analogy between an autonomous system and an employee: where a system operates for the benefit of its deployer, \"its…","np":"Anything about capability or competence. It is a data protection regulation confined to one free zone, it addresses liability rather than skill, and it imposes no requirement that the responsible…","u":"/research/evidence#difc-reg10-2024","src":"https://www.difc.com/","on":["/research/ai-and-work-in-the-gulf","/research/who-owns-verification-when-ai-does-the-work"],"alt":["DIFC"]},{"t":"study","n":"Artificial Intelligence in Qatar: Principles and Guidelines for Ethical Development and Deployment","au":"Ministry of Communications and Information Technology, Qatar","y":2026,"g":"compiled-review","sec":"international","f":"Principle 8, assign ultimate accountability to humans, states that \"AI systems should not be able to autonomously make decisions of significant consequence\" and should \"provide users with the ability to appeal or override decisions that…","np":"Any competence duty. Qatar imposes no obligation that the humans exercising control be trained, assessed or kept current, and the document contains no reference to deskilling, over-reliance or…","u":"/research/evidence#qatar-mcit-ai-principles","src":"https://www.mcit.gov.qa/en/","on":["/research/ai-and-work-in-the-gulf"],"alt":[]},{"t":"study","n":"Model AI Governance Framework for Agentic AI, Version 1.0","au":"Infocomm Media Development Authority (IMDA), Singapore","y":2026,"g":"institutional-survey","sec":"institutional","f":"Names deskilling as a risk of agentic deployment, in the terms this research uses. Section 2.4.3: \"As agents take over entry level tasks, which typically serve as the training ground for new staff, this could lead to loss of basic…","np":"Anything measured. It is guidance, not law, and it states a risk and a duty to train rather than evidence that deskilling has occurred. It also sets no threshold for what counts as retaining core…","u":"/research/evidence#imda-agentic-2026","src":"https://www.imda.gov.sg/-/media/imda/files/about/emerging-tech-and-research/artificial-intelligence/mgf-for-agentic-ai.pdf","on":["/research/ai-and-work-in-asia","/research/missing-rungs"],"alt":["IMDA"]},{"t":"study","n":"Model AI Governance Framework for Generative AI","au":"AI Verify Foundation and IMDA, Singapore","y":2024,"g":"institutional-survey","sec":"institutional","f":"Contains zero occurrences of \"human oversight\", \"human-in-the-loop\", \"over-reliance\", \"automation bias\" or \"deskill\". Human oversight is not among its nine dimensions. The nearest it comes is a note that \"core skills such as creativity,…","np":"That Singapore was indifferent to oversight in 2024; the 2020 framework it builds on was not read at source and may carry such language. It establishes what the generative AI framework does not say,…","u":"/research/evidence#imda-genai-2024","src":"https://aiverifyfoundation.sg/resources/mgf-gen-ai/","on":["/research/ai-and-work-in-asia"],"alt":[]},{"t":"study","n":"Labour Market Report, First Quarter 2026","au":"Manpower Research and Statistics Department, Ministry of Manpower,…","y":2026,"g":"institutional-survey","sec":"international","f":"28.5 per cent of firms adopted AI in 2026, highest in information and communications at 74.1 per cent, professional services at 57.5 per cent and financial and insurance services at 56.4 per cent. Only 6.2 per cent reported AI-related…","np":"What happens to capability. Redesign is not neutral: the question this research asks is which tasks the redesign removes, and a firm survey of headcount cannot answer it.","u":"/research/evidence#mom-labour-2026","src":"https://stats.mom.gov.sg/Pages/Labour-Market-Report-1Q-2026.aspx","on":["/research/ai-and-work-in-asia"],"alt":[]},{"t":"study","n":"Report on the Survey on Information Technology Usage and Penetration in the Business Sector, 2025 Edition","au":"Census and Statistics Department, Hong Kong SAR","y":2026,"g":"institutional-survey","sec":"international","f":"Artificial intelligence appears nowhere in the report: not in the tables, not in the explanatory notes, not in the definitions. The survey's ICT categories are cloud computing at 98.1 per cent, QR codes at 37.6 per cent, RFID at 20.3 per…","np":"That AI adoption in Hong Kong is low. It establishes that it is unmeasured by the government, which is a different and in some ways more useful fact.","u":"/research/evidence#hk-censtatd-2026","src":"https://www.censtatd.gov.hk/","on":["/research/ai-and-work-in-asia"],"alt":["ICT","RFID"]},{"t":"study","n":"Work Reimagined Survey 2025","au":"EY","y":2025,"g":"institutional-survey","sec":"institutional","f":"88 per cent of employees use AI at work, but mostly for basic tasks such as search and summarisation, and only 5 per cent use it in advanced ways that transform how they work. 37 per cent worry that overreliance on AI could erode their…","np":"Any measured capability loss. Every figure is self-reported perception at a single point in time, the sample is not random, and the 40 per cent productivity figure is EY's own modelled comparison…","u":"/research/evidence#ey-work-reimagined-2025","src":"https://www.ey.com/en_gl/newsroom/2025/11/ey-survey-reveals-companies-are-missing-out-on-up-to-40-percent-of-ai-productivity-gains-due-to-gaps-in-talent-strategy","on":["/research/capability-debt","/research/ai-workforce-strategy"],"alt":[]},{"t":"study","n":"When Everyone Uses AI, Companies Risk Losing Critical Skills","au":"Boston Consulting Group","y":2026,"g":"institutional-survey","sec":"institutional","f":"Half of the executives surveyed report already observing deskilling in their organisations, and more than 60 per cent believe deskilling will pose a material threat to their organisation within the next three to five years.","np":"Prevalence. The base is 70 people. Anyone quoting the 50 per cent without the 70 is overstating it, and executive perception is not a measurement of what is happening to anyone's skills.","u":"/research/evidence#bcg-deskilling-2026","src":"https://www.bcg.com/publications/2026/when-everyone-uses-ai-companies-risk-critical-skills","on":["/research/capability-debt","/research/which-professions-face-the-greatest-deskilling-risk"],"alt":[]},{"t":"study","n":"Aviation Investigation Report A19P0112, Seair Seaplanes, Addenbroke Island","au":"Transportation Safety Board of Canada","y":2019,"g":"compiled-review","sec":"professions","f":"Surveyed air-taxi operators expressed concern that 'dependence on technology was causal in degradation of basic piloting skills', and numerous operators commented that over-reliance on GPS navigation 'may contribute to the decision to fly…","np":"That automation dependence caused this crash. The accident's own findings as to causes cite weather, terrain-alerting ambiguity and fatigue. This is industry survey commentary quoted inside an…","u":"/research/evidence#tsb-a19p0112-2019","src":"https://www.tsb.gc.ca/eng/rapports-reports/aviation/2019/a19p0112/a19p0112.html","on":["/research/capability-debt","/research/what-professions-can-learn-from-aviation"],"alt":["GPS","TSB"]},{"t":"study","n":"Highway Investigation Report HIR-26-02, Ford BlueCruise collisions","au":"National Transportation Safety Board","y":2026,"g":"compiled-review","sec":"professions","f":"Overreliance on the partial automation system appears in the probable cause for both crashes, not merely in discussion: distraction 'stemming from overreliance on the vehicle's hands-free partial automation system' in San Antonio, and…","np":"Anything about knowledge work. Driving is a continuous manual-control task with a monitoring system watching the human, which is not the shape of AI-assisted professional judgement.","u":"/research/evidence#ntsb-bluecruise-2026","src":"https://www.ntsb.gov/investigations/AccidentReports/Reports/HIR2602.pdf","on":["/research/capability-debt","/research/what-is-automation-complacency"],"alt":[]},{"t":"study","n":"Test-Enhanced Learning: Taking Memory Tests Improves Long-Term Retention","au":"Roediger, H. L. III and Karpicke, J. D.","y":2006,"g":"peer-reviewed","sec":"learning","f":"The winner reverses with delay. At 5 minutes restudying beat testing, 81 per cent against 75 per cent. At one week testing beat restudying, 56 per cent against 42 per cent. In the second experiment repeated study led at 5 minutes, 83…","np":"Anything about AI. It is prose recall in a laboratory, and the transfer to professional judgement is by analogy.","u":"/research/evidence#roediger-karpicke-2006","src":"https://doi.org/10.1111/j.1467-9280.2006.01693.x","on":["/research/what-is-retrieval-practice","/research/what-is-desirable-difficulty"],"alt":[]},{"t":"study","n":"Retrieval Practice Produces More Learning than Elaborative Studying with Concept Mapping","au":"Karpicke, J. D. and Blunt, J. R.","y":2011,"g":"peer-reviewed","sec":"learning","f":"Retrieval practice scored 0.67 against 0.45 for concept mapping, about a 50 per cent advantage in long-term retention, d = 1.50. In the second experiment 101 of 120 students, 84 per cent, did better after retrieval practice than after…","np":"That concept mapping is worthless. A published comment by Mintzes and colleagues (Science, 2011, 334(6055), 453) disputes the instructional fidelity of the concept-mapping condition, and anyone…","u":"/research/evidence#karpicke-blunt-2011","src":"https://doi.org/10.1126/science.1199327","on":["/research/what-is-retrieval-practice"],"alt":[]},{"t":"study","n":"The Effect of Testing Versus Restudy on Retention: A Meta-Analytic Review of the Testing Effect","au":"Rowland, C. A.","y":2014,"g":"peer-reviewed","sec":"learning","f":"A reliable testing effect, g = 0.50 with a confidence interval of 0.42 to 0.58. With feedback the effect rises to g = 0.73; without feedback it falls to g = 0.39.","np":"That the effect size transfers to workplace learning. The constituent studies are overwhelmingly laboratory and classroom work on verbal material.","u":"/research/evidence#rowland-2014","src":"https://doi.org/10.1037/a0037559","on":["/research/what-is-retrieval-practice"],"alt":[]},{"t":"study","n":"Productive Failure","au":"Kapur, M.","y":2008,"g":"peer-reviewed","sec":"learning","f":"The group given ill-structured problems struggled visibly and produced poor solutions during the collaborative phase, then outperformed the other group on individual near-transfer and far-transfer measures afterwards.","np":"A precise magnitude. The design, sample and direction of the result are confirmed from the publisher's abstract and the author's own presentation of the study, but the full text sits behind a paywall…","u":"/research/evidence#kapur-2008","src":"https://doi.org/10.1080/07370000802212669","on":["/research/what-is-productive-struggle"],"alt":[]},{"t":"study","n":"When Problem Solving Followed by Instruction Works: Evidence for Productive Failure","au":"Sinha, T. and Kapur, M.","y":2021,"g":"peer-reviewed","sec":"learning","f":"A moderate effect favouring problem solving first, Hedges g = 0.36 with a confidence interval of 0.20 to 0.51, rising to between 0.37 and 0.58 where the design followed the productive failure principles closely. The effect reverses for…","np":"That struggle is universally good. The authors report the reversal for young children and for general skills themselves, and an earlier meta-analysis by Darabi and colleagues rested on only 12…","u":"/research/evidence#sinha-kapur-2021","src":"https://doi.org/10.3102/00346543211019105","on":["/research/what-is-productive-struggle"],"alt":[]},{"t":"study","n":"Evidence on defunding of level 7 apprenticeships","au":"Skills England","y":2026,"g":"institutional-modelling","sec":"institutional","f":"Level 7 apprenticeship starts in 2023/24 were 23,870, of which 65 per cent were aged 25 or over, 34 per cent aged 19 to 24, and 2 per cent under 19. Funding continues for 16 to 21 year olds, for care leavers and those with an education,…","np":"That degree apprenticeships in general were cut. Level 6, the undergraduate degree apprenticeship, is unaffected by this decision and its starts were still rising.","u":"/research/evidence#skills-england-l7-2026","src":"https://www.gov.uk/government/publications/skills-england-evidence-on-defunding-of-level-7-apprenticeships/skills-england-evidence-on-defunding-of-level-7-apprenticeships","on":["/research/do-apprenticeships-still-work"],"alt":[]},{"t":"study","n":"Apprenticeship statistics for England","au":"Murray, A., House of Commons Library","y":2025,"g":"institutional-survey","sec":"institutional","f":"353,500 apprenticeship starts in England in 2024/25, up from 340,000 in 2023/24 and 337,000 in 2022/23. In 2024/25, 51.3 per cent of starts were by apprentices aged 25 or over, 27.5 per cent aged 19 to 24, and 21.2 per cent under 19. The…","np":"Anything about quality, completion or whether apprenticeships lead to work. Starts are a count of beginnings.","u":"/research/evidence#hoc-apprenticeships-2025","src":"https://researchbriefings.files.parliament.uk/documents/SN06113/SN06113.pdf","on":["/research/do-apprenticeships-still-work"],"alt":[]},{"t":"study","n":"Increasing construction skills","au":"National Audit Office","y":2026,"g":"institutional-survey","sec":"institutional","f":"By April 2026 only 74 young people had started a construction foundation apprenticeship, against the department's own assumption of 1,000 in 2025-26.","np":"That foundation apprenticeships cannot work. It is eight months of data on a scheme launched in August 2025, and low take-up in one sector.","u":"/research/evidence#nao-construction-skills-2026","src":"https://www.nao.org.uk/wp-content/uploads/2026/07/increasing-construction-skills.pdf","on":["/research/do-apprenticeships-still-work"],"alt":[]},{"t":"study","n":"Angespannte Lage auf dem Ausbildungsmarkt (Tense situation in the training market)","au":"Bundesinstitut fuer Berufsbildung","y":2025,"g":"institutional-survey","sec":"international","f":"Around 476,000 new dual training contracts in 2025, down 2.1 per cent on 2024 and the second consecutive annual fall. 54,400 training places went unfilled. At the same time around 84,400 young people had not found a training place, up 19.9…","np":"That AI caused it. BIBB attributes the tension to matching problems, regional and occupational mismatch and economic conditions, and makes no AI claim.","u":"/research/evidence#bibb-2025","src":"https://www.bibb.de/de/pressemitteilung_215396.php","on":["/research/do-apprenticeships-still-work"],"alt":[]},{"t":"study","n":"Apprentices and trainees 2025","au":"National Centre for Vocational Education Research","y":2025,"g":"institutional-survey","sec":"international","f":"320,830 in-training contracts at 31 March 2025, down 7.9 per cent year on year, with trade down 3.2 per cent and non-trade down 17.9 per cent. By the June quarter the fall was 11.3 per cent, with non-trade down 20.2 per cent. NCVER…","np":"An AI effect. NCVER names policy change, and the timing follows the 2022 subsidy withdrawal rather than any AI milestone.","u":"/research/evidence#ncver-2025","src":"https://www.ncver.edu.au/research-and-statistics/publications/all-publications/apprentices-and-trainees-2025-march-quarter","on":["/research/do-apprenticeships-still-work"],"alt":["NCVER"]},{"t":"study","n":"Call for abstracts: apprenticeships in the age of AI","au":"Cedefop","y":2026,"g":"compiled-review","sec":"international","f":"Cedefop states: \"AI takes over baseline tasks which were in many cases performed by apprentices or apprenticeship graduates who get entry-level roles once their programmes are completed. As apprenticeships continue to expand into new…","np":"Anything measured. It is a call for papers, which means Cedefop is asking the question rather than answering it, and it says the effect is uneven.","u":"/research/evidence#cedefop-apprenticeships-ai-2026","src":"https://www.cedefop.europa.eu/en/news/call-abstracts-apprenticeships-age-ai","on":["/research/do-apprenticeships-still-work","/research/missing-rungs"],"alt":["OECD"]},{"t":"study","n":"Global Employment Trends for Youth 2026: Back to the future","au":"International Labour Organization","y":2026,"g":"institutional-survey","sec":"international","f":"Global youth unemployment 12.4 per cent in 2025, 67 million people aged 15 to 24. The NEET rate is 20 per cent, over 257 million. Youth unemployment rose in 8 of 11 subregions between 2023 and 2025, with Northern America rising from 8.3 to…","np":"That AI caused it. The 6.1 per cent exposure figure is an occupational overlap measure, not a measured displacement.","u":"/research/evidence#ilo-youth-2026","src":"https://www.ilo.org/resource/news/youth-unemployment-rises-young-people-face-harder-road-decent-work","on":["/research/do-apprenticeships-still-work","/research/what-is-the-ai-employment-gap"],"alt":["ILO","NEET"]},{"t":"study","n":"The Graduate Market in 2026","au":"High Fliers Research","y":2026,"g":"institutional-survey","sec":"work","f":"Graduate recruitment fell 5.1 per cent in 2025, after a 14.6 per cent drop in 2024 and a 6.4 per cent decrease in 2023, with a further 0.5 per cent decrease forecast for 2026. Graduate recruitment at these employers has fallen 24.5 per…","np":"Attribution to AI, which the survey does not establish, and it covers 100 leading employers rather than the whole labour market.","u":"/research/evidence#highfliers-2026","src":"https://highfliers.co.uk/publication-the-graduate-market-report","on":["/research/do-apprenticeships-still-work","/research/will-ai-replace-entry-level-jobs"],"alt":[]},{"t":"study","n":"Student Recruitment Survey 2025","au":"Institute of Student Employers","y":2025,"g":"institutional-survey","sec":"work","f":"An average of 89 applications per vacancy. The ISE reports a projected 7 per cent drop in student vacancies for 2026, while 30 per cent of employers increased student hiring.","np":"Whole-market figures. It is a membership survey, and its members are organisations that run structured student recruitment in the first place.","u":"/research/evidence#ise-recruitment-2025","src":"https://ise.org.uk/_userfiles/pages/files/reports/student_recruitment_survey_2025.pdf","on":["/research/do-apprenticeships-still-work"],"alt":["ISE"]},{"t":"study","n":"Conditions for intuitive expertise: a failure to disagree","au":"Kahneman, D. and Klein, G.","y":2009,"g":"peer-reviewed","sec":"judgement","f":"Judging the likely quality of an intuitive judgement requires assessing two things: the predictability of the environment in which the judgement is made, and the individual's opportunity to learn that environment's regularities. Where both…","np":"Which specific professional environments meet the conditions. The authors give the criteria and not a classification, so applying it to law, medicine, consulting or management is a judgement in…","u":"/research/evidence#kahneman-klein-2009","src":"https://pubmed.ncbi.nlm.nih.gov/19739881/","on":["/research/what-is-judgement","/research/ai-and-human-judgement","/research/ai-and-human-intuition"],"alt":[]},{"t":"study","n":"Deliberate Practice and Performance in Music, Games, Sports, Education, and Professions: A Meta-Analysis","au":"Macnamara, B. N., Hambrick, D. Z. and Oswald, F. L.","y":2014,"g":"compiled-review","sec":"learning","f":"Deliberate practice explained 12 per cent of the variance in performance overall, 95% CI [9%, 15%], leaving 88 per cent unexplained. By domain: 26 per cent for games, 21 per cent for music, 18 per cent for sports, 4 per cent for education…","np":"That practice does not build professional expertise. The professions estimate rests on 7 effect sizes, is not statistically significant (p = .62), and the occupations sampled were computer…","u":"/research/evidence#macnamara-2014","src":"https://journals.sagepub.com/doi/abs/10.1177/0956797614535810","on":["/research/what-is-deliberate-practice"],"alt":[]},{"t":"study","n":"How People Learn II: Learners, Contexts, and Cultures","au":"National Academies of Sciences, Engineering, and Medicine","y":2018,"g":"compiled-review","sec":"learning","f":"Defines metacognition as the ability to monitor and regulate one's own cognitive processes and to consciously regulate behaviour, including affective behaviour, and identifies calibration as the accuracy of a learner's monitoring.","np":"Anything about AI. The report predates general availability of these systems and makes no claim about them.","u":"/research/evidence#nap-hpl2-2018","src":"https://www.nationalacademies.org/read/24783/chapter/6","on":["/research/what-is-metacognition"],"alt":[]},{"t":"study","n":"Predictors and consequences of intellectual humility","au":"Porter, T., Elnakouri, A., Meyers, E. A., Shibayama, T.,…","y":2022,"g":"compiled-review","sec":"judgement","f":"Identifies a metacognitive core on which there is scholarly consensus, recognising the limits of one's knowledge and being aware of one's fallibility, with social and behavioural features around it: recognising that others may hold…","np":"That the construct is settled, or that it can be measured reliably by asking people. The review notes that where intellectual humility is seen as desirable, self-report makes a false impression easy…","u":"/research/evidence#porter-2022","src":"https://www.nature.com/articles/s44159-022-00081-9","on":["/research/what-is-intellectual-humility"],"alt":[]},{"t":"study","n":"Talk, Trust, and Trade-offs: How and Why Teens Use AI Companions","au":"Common Sense Media","y":2025,"g":"institutional-survey","sec":"humanness","f":"72 per cent of teens had used an AI companion at least once and 52 per cent were regular users, a few times a month or more. Among users, 33 per cent had chosen an AI companion over a real person for something important or serious, 34 per…","np":"Anything about developmental effects, which are not measured here. Self-report at a single point in time, by an organisation that backs legislation to ban these products for minors, and the risk…","u":"/research/evidence#commonsense-companions-2025","src":"https://www.commonsensemedia.org/sites/default/files/research/talk-trust-and-trade-offs_2025_toplines.pdf","on":[],"alt":[]},{"t":"study","n":"AI Risk Management Framework Playbook, MANAGE 2.4","au":"National Institute of Standards and Technology","y":2023,"g":"institutional-modelling","sec":"institutional","f":"Requires mechanisms and assigned responsibilities to supersede, disengage or deactivate AI systems showing performance inconsistent with intended use, and names five triggering conditions: end of system lifetime; risks exceeding tolerance…","np":"That any organisation does this, or that doing it works. The AI RMF is voluntary guidance, not a standard with conformity assessment, and it contains no evidence about outcomes.","u":"/research/evidence#nist-airmf-manage-2-4","src":"https://airc.nist.gov/airmf-resources/playbook/manage/","on":[],"alt":[]},{"t":"study","n":"Regulation (EU) 2024/1689, Article 14: Human Oversight","au":"European Union","y":2024,"g":"institutional-modelling","sec":"institutional","f":"Requires high-risk systems to be designed so they can be effectively overseen by natural persons, and requires that those persons be enabled to understand the system's capacities and limitations, to remain aware of the tendency to…","np":"That any of this happens. The oversight provisions carry 2 December 2027 as a longstop rather than a start date, since the Digital Omnibus ties the high-risk rules to the availability of standards…","u":"/research/evidence#eu-ai-act-art-14","src":"https://artificialintelligenceact.eu/article/14/","on":["/research/how-do-you-design-a-stop-button-people-will-use","/research/how-do-humans-and-agents-divide-work"],"alt":["III"]},{"t":"study","n":"Four Principles of Explainable Artificial Intelligence (NISTIR 8312)","au":"Phillips, P. J., Hahn, C. A., Fontana, P. C., Broniatowski, D. A. and…","y":2021,"g":"compiled-review","sec":"judgement","f":"Sets out four principles: explanation, meaningful, explanation accuracy and knowledge limits. In the report's own words, explanation accuracy is a distinct concept from decision accuracy, and regardless of the system's decision accuracy…","np":"That explanations help or harm users in practice. This is a framework rather than an experiment, and it reports no effect on human decision quality.","u":"/research/evidence#nist-ir-8312","src":"https://nvlpubs.nist.gov/nistpubs/ir/2021/NIST.IR.8312.pdf","on":[],"alt":[]},{"t":"study","n":"Does generative AI narrow education-based productivity gaps? Evidence from a randomized experiment","au":"Cruces, G., Fernandez Meijide, D., Galiani, S., Galvez, R. and…","y":2026,"g":"working-paper","sec":"learning","f":"AI improved performance for everyone and more for the less educated. Without AI, higher-education participants outperformed lower-education participants by 0.548 standard deviations; with AI the gap fell to 0.139, closing about…","np":"That skill formed. One session with an immediate unassisted module tests transfer within a sitting, not skill formation over time, and the studies that found post-removal deficits taught a body of…","u":"/research/evidence#cruces-2026","src":"https://www.nber.org/papers/w34851","on":["/research/capability-debt"],"alt":[]},{"t":"study","n":"How AI Impacts Skill Formation","au":"Shen, J. H. and Tamkin, A.","y":2026,"g":"working-paper","sec":"learning","f":"The assisted arm averaged 50 per cent on the quiz against 67 per cent for the hand-coding arm, a difference of 4.15 points on 27, Cohen's d of 0.738 at p = 0.010, holding at d = 0.725, p = 0.016 with warm-up time as a covariate. The widest…","np":"That AI use degrades professional expertise. One library, one tutorial, one hour, a preprint, and a sample of 52. The interaction-pattern analysis was NOT pre-registered and the authors state that it…","u":"/research/evidence#shen-tamkin-2026","src":"https://arxiv.org/abs/2601.20245","on":["/research/capability-debt","/research/does-using-ai-stop-you-learning"],"alt":["GPT"]},{"t":"study","n":"News Integrity in AI Assistants: An international PSM study","au":"Fletcher, J. and Verckist, D. (BBC and European Broadcasting Union)","y":2025,"g":"compiled-review","sec":"professions","f":"Forty-five per cent of responses carried at least one significant issue, and 81 per cent carried an issue of some kind. Sourcing was the largest single cause at 31 per cent, then accuracy at 20 per cent and insufficient context at 14 per…","np":"Current performance of any named product. These were free consumer versions tested in mid-2025 and all four have shipped new defaults since. It is not adversarial testing and difficulty was not…","u":"/research/evidence#ebu-bbc-2025","src":"https://www.ebu.ch/research/open/report/news-integrity-in-ai-assistants","on":["/research/how-will-ai-change-journalism"],"alt":["BBC"]},{"t":"study","n":"Local Law 144 of 2021, automated employment decision tools","au":"Council of the City of New York; Department of Consumer and Worker…","y":2021,"g":"institutional-modelling","sec":"institutional","f":"The rules at 6 RCNY s 5-300 define the trigger, 'substantially assist or replace discretionary decision making', in three limbs, of which the third is: 'to use a simplified output to overrule conclusions derived from other factors…","np":"Anything about practice or enforcement, which is the subject of the separate Comptroller audit graded below. It is city law, applying only to employment decisions within New York City, and it…","u":"/research/evidence#nyc-ll144-2021","src":"https://www.nyc.gov/site/dca/about/automated-employment-decision-tools.page","on":["/research/what-is-meaningful-human-oversight","/research/can-ai-be-unbiased","/research/is-it-ethical-to-let-ai-judge-people"],"alt":["RCNY"]},{"t":"study","n":"Department of Consumer and Worker Protection: Enforcement of Local Law 144","au":"Office of the New York State Comptroller","y":2025,"g":"statutory-investigation","sec":"institutional","f":"The department surveyed the websites and bias audits of 32 companies and identified a single issue of non-compliance. The auditors reviewed the same companies and identified at least seventeen instances of potential non-compliance. Two…","np":"That the seventeen are violations. The audit says potential non-compliance, and the department disputes elements of the finding. It measures enforcement activity, not whether the tools in question…","u":"/research/evidence#nys-comptroller-ll144-2025","src":"https://www.osc.ny.gov/state-agencies/audits/2025/12/02/department-consumer-and-worker-protection-enforcement-local-law-144","on":["/research/what-is-meaningful-human-oversight","/research/can-ai-be-unbiased","/research/is-it-ethical-to-let-ai-judge-people"],"alt":[]},{"t":"study","n":"Act respecting the protection of personal information in the private sector, s 12.1","au":"National Assembly of Quebec","y":2021,"g":"institutional-modelling","sec":"international","f":"Where an enterprise uses personal information to render a decision based exclusively on automated processing, it must say so by the time it communicates the decision, and on request give the information used, the reasons and principal…","np":"That it is used, or that the reviewing person is competent to redo the analysis. The statute does not define what being in a position to review requires. Quebec s own regulator lists four limits,…","u":"/research/evidence#quebec-p391-s121","src":"https://www.legisquebec.gouv.qc.ca/en/document/cs/p-39.1","on":["/research/who-supervises-work-they-cannot-do"],"alt":[]},{"t":"study","n":"L IA au travail: pour un meilleur encadrement","au":"Commission d acces a l information du Quebec","y":2025,"g":"argued-perspective","sec":"international","f":"On meaningful human intervention the Commission states, at page 4: 'lorsqu un humain enterine une decision proposee par un systeme d IA sans etudier l ensemble de l analyse, il existe un risque qu il demontre un biais d automatisation en…","np":"Anything measured. It is a submission, it contains no data on how often ratification without examination occurs, and its recommendations had not been enacted at the time of reading.","u":"/research/evidence#cai-quebec-ia-travail-2025","src":"https://www.cai.gouv.qc.ca/uploads/pdfs/CAI_ME_Transfo_Travail.pdf","on":["/research/what-is-automation-bias"],"alt":[]},{"t":"study","n":"Administrative Procedure Act, chapter 8 b, automated decision making","au":"Parliament of Finland","y":2023,"g":"institutional-modelling","sec":"international","f":"An authority may decide a matter automatically only where the matter 'contains no elements requiring case-by-case discretion' (johon ei sisally seikkoja, jotka edellyttavat tapauskohtaista harkintaa). A decision counts as automated where…","np":"That it works, or that Finnish authorities classify matters correctly. It binds public authorities only, not private employers. Finland s implementation of the EU AI Act is separately late, so this…","u":"/research/evidence#finland-hallintolaki-8b-2023","src":"https://www.finlex.fi/fi/laki/ajantasa/2003/20030434","on":["/research/human-in-the-loop-is-not-a-safeguard"],"alt":[]},{"t":"study","n":"Decision on the Tax Administration s automated decision making","au":"Deputy Parliamentary Ombudsman of Finland (Maija Sakslin)","y":2019,"g":"statutory-investigation","sec":"international","f":"The Ombudsman found the practice unlawful. The reasoning turns on accountability rather than error: official accountability had become indirect (virkavastuu jaa valilliseksi), because no identifiable official could be said to have made the…","np":"That any decision was wrong. The finding is about accountability structure, not accuracy, and the Ombudsman did not measure outcomes. It concerns Finnish administrative law and does not transfer to…","u":"/research/evidence#eoak-3379-2018-tax","src":"https://www.oikeusasiamies.fi/","on":["/research/what-is-a-moral-crumple-zone"],"alt":[]},{"t":"study","n":"Vejledning om digitaliseringsklar lovgivning","au":"Digitaliseringsstyrelsen (Danish Agency for Digital Government)","y":2018,"g":"institutional-modelling","sec":"international","f":"The third principle carries a written checklist question: 'Er det sikret, at det fagprofessionelle skon er opretholdt i tilfaelde, hvor hensynet til borgernes retssikkerhed taler herfor?' (Has it been ensured that professional discretion…","np":"That discretion is in fact preserved. It measures a drafting process, not outcomes, no published review of how the question is answered was found, and it binds those who draft legislation rather than…","u":"/research/evidence#denmark-vej-9590-2018","src":"https://www.retsinformation.dk/eli/retsinfo/2018/9590","on":["/research/what-board-oversight-of-ai-looks-like"],"alt":[]},{"t":"study","n":"Artificial intelligence use by enterprises (isoc_eb_ai), 2025 reference year","au":"Eurostat","y":2025,"g":"institutional-survey","sec":"international","f":"EU-27 average 19.95 per cent, up 6.47 points on 2024. Denmark first at 42.0 per cent, Finland second at 37.8, Sweden 35.0, Netherlands 33.2 (break in series), Spain 20.3, Portugal 11.5, Romania lowest at 5.2. Norway, reporting as a non-EU…","np":"Depth of use. An enterprise counts if it used one of eight technologies once, so the measure says nothing about how many workers use AI, how often, or for what. It excludes the financial and public…","u":"/research/evidence#eurostat-isoc-eb-ai-2025","src":"https://ec.europa.eu/eurostat/web/products-eurostat-news/w/ddn-20251211-1","on":["/research/ai-and-work-by-country"],"alt":["EEA","NACE"]},{"t":"study","n":"M-24-10, Advancing Governance, Innovation, and Risk Management for Agency Use of Artificial Intelligence","au":"Office of Management and Budget (Shalanda D. Young)","y":2024,"g":"institutional-modelling","sec":"institutional","f":"M-24-10 used the term automation bias twice. As a defined term at section 6: 'the propensity for humans to inordinately favor suggestions from automated decision-making systems and to ignore or fail to seek out contradictory information…","np":"That the removal was deliberate or that federal agencies have stopped addressing the risk. M-25-21 still requires human oversight, intervention and accountability for high-impact uses. The words…","u":"/research/evidence#omb-m2410-automation-bias-2024","src":"https://bidenwhitehouse.archives.gov/wp-content/uploads/2024/03/M-24-10-Advancing-Governance-Innovation-and-Risk-Management-for-Agency-Use-of-Artificial-Intelligence.pdf","on":["/research/what-is-automation-bias"],"alt":[]},{"t":"study","n":"Supervisory Guidance on Model Risk Management (SR 26-2)","au":"Board of Governors of the Federal Reserve System, Federal Deposit…","y":2026,"g":"institutional-modelling","sec":"professions","f":"The definition of a model is narrowed to 'a complex quantitative method, system, or approach that applies statistical, economic, or financial theories to process input data into quantitative estimates', expressly excluding simple…","np":"That generative AI in banks is ungoverned. The same footnote directs firms to their own risk management and governance practices for tools outside scope, and other supervisory expectations, consumer…","u":"/research/evidence#fed-sr26-2-2026","src":"https://www.federalreserve.gov/supervisionreg/srletters/SR2602.htm","on":["/research/how-will-ai-change-accounting-and-audit"],"alt":[]},{"t":"study","n":"AI in audit: Illustrative example and documentation guidance","au":"Financial Reporting Council","y":2025,"g":"institutional-modelling","sec":"professions","f":"The scope is deliberately broad, covering 'both traditional machine learning techniques and deep learning models, including generative AI'. On explainability the FRC declines to set a threshold: 'what constitutes appropriate explainability…","np":"Anything about practice. It is guidance issued in 2025, not a finding about what firms do, and the FRC says it creates no new requirements. It contains nothing on junior auditors, training pipelines…","u":"/research/evidence#frc-ai-audit-2025","src":"https://www.frc.org.uk/documents/8384/AI_in_Audit.pdf","on":["/research/how-will-ai-change-accounting-and-audit"],"alt":["FRC"]},{"t":"study","n":"Thematic Review: Certification of Automated Tools and Techniques","au":"Financial Reporting Council","y":2025,"g":"compiled-review","sec":"professions","f":"All six firms had certification processes, but 'the maturity of these processes was found to vary and in some cases were not supported by formal documented policies'. Only two of the six set out the limitations of a tool or restrictions on…","np":"That audit quality has fallen. The review measures process rather than outcome, covers only the six largest firms, and its counts are of documentation practice rather than of tools or audits. Nothing…","u":"/research/evidence#frc-att-thematic-2025","src":"https://www.frc.org.uk/documents/8383/Thematic_Review_on_the_Certification_of_Automated_Tools_and_Techniques.pdf","on":["/research/how-will-ai-change-accounting-and-audit"],"alt":["BDO","KPMG"]},{"t":"study","n":"ChatGPT's inconsistent moral advice influences users' judgment","au":"Krügel, S., Ostermaier, A. and Uhl, M.","y":2023,"g":"peer-reviewed","sec":"judgement","f":"The advice moved participants' own moral judgement in both dilemmas, and in the bridge version it flipped the majority verdict. Disclosure made almost no difference: the effect was statistically indistinguishable whether the source was…","np":"How large the shift is in absolute terms. The paper reports test statistics and figure proportions rather than an effect size in the text, and no confidence intervals appear in the prose. One dilemma…","u":"/research/evidence#krugel-2023","src":"https://www.nature.com/articles/s41598-023-31341-0","on":["/research/should-i-let-ai-make-personal-decisions-for-me","/research/is-it-still-my-idea-if-ai-helped-me-write-it"],"alt":[]},{"t":"study","n":"From Technical Debt to Cognitive and Intent Debt: Rethinking Software Health in the Age of AI","au":"Storey, M.-A.","y":2026,"g":"compiled-review","sec":"capability","f":"Proposes a triple debt model: technical debt lives in code, cognitive debt lives in people as the erosion of shared understanding across a team, and intent debt lives in artefacts as the absence of captured rationale, goals and…","np":"Anything empirical about prevalence or magnitude. It is a framework paper with an anecdote, and it says so. Its reference list also mis-cites Kosmyna et al. as a 2024 CHI workshop paper, which does…","u":"/research/evidence#storey-2026","src":"https://arxiv.org/pdf/2603.22106","on":["/research/cognitive-debt-and-capability-debt","/research/will-ai-replace-programmers"],"alt":[]},{"t":"study","n":"AI, Human Cognition and Knowledge Collapse","au":"Acemoglu, D., Kong, D. and Ozdaglar, A.","y":2026,"g":"institutional-modelling","sec":"institutional","f":"Identifies a conditional tipping point: when human effort is sufficiently elastic and agentic recommendations exceed an accuracy threshold, the economy can reach a knowledge-collapse steady state in which general knowledge ultimately…","np":"That knowledge collapse is happening or will happen. It is a model producing a possible steady state under stated conditions, it is a working paper rather than a peer-reviewed article, and the…","u":"/research/evidence#acemoglu-2026","src":"https://www.nber.org/papers/w34910","on":["/research/cognitive-debt-and-capability-debt"],"alt":[]},{"t":"study","n":"ChatGPT in lesson preparation: a Teacher Choices trial","au":"Education Endowment Foundation and National Foundation for…","y":2024,"g":"peer-reviewed","sec":"professions","f":"Weekly lesson and resource preparation time was 56.2 minutes in the ChatGPT arm against 81.5 minutes in the comparison arm, a saving of 25.3 minutes and a reduction of 31 per cent, given a high security rating. The blinded panel found no…","np":"Anything about teaching or about pupils. Preparation time was the outcome; classroom effect was outside scope. Time was self-recorded rather than observed. The sample over-represents schools in…","u":"/research/evidence#eef-nfer-chatgpt-2024","src":"https://educationendowmentfoundation.org.uk/projects-and-evaluation/projects/choices-in-edtech-using-generative-ai-chatgpt-for-ks3-science-lesson-preparation-2024-teacher-choices-trial","on":["/research/how-will-ai-change-teaching"],"alt":["NFER"]},{"t":"study","n":"Technology in Schools survey: 2024 to 2025","au":"Department for Education and IFF Research","y":2025,"g":"institutional-survey","sec":"institutional","f":"44 per cent of teachers reported using generative AI for school activities: lesson planning 35 per cent, delivering live lessons 7 per cent, marking 5 per cent. Teachers under 35 used it for planning at 43 per cent against 32 per cent for…","np":"Any effect. It is a cross-sectional self-report of usage and perception, with no measurement of workload, learning or quality. The plagiarism figure records reported issues rather than the extent of…","u":"/research/evidence#dfe-tech-schools-2025","src":"https://assets.publishing.service.gov.uk/media/692834a6ce50d215cae9610e/Technology_in_schools_survey_2024_to_2025_research_report.pdf","on":["/research/how-will-ai-change-teaching"],"alt":[]},{"t":"study","n":"AI tutoring outperforms in-class active learning: an RCT introducing a novel research-based design in an authentic educational setting","au":"Kestin, G., Miller, K., Klales, A., Milbourne, T. and Ponti, G.","y":2025,"g":"peer-reviewed","sec":"learning","f":"Median post-test score 4.5 in the AI condition against 3.5 in the active-learning condition, from a combined pre-test median of 2.75; median learning gain over double. Mann-Whitney z = -5.6, p below 10 to the minus 8. Linear regression…","np":"That a general chatbot does this. Accuracy depended on pre-written answers and instructor-written prompts. Retention was not measured; post-tests followed the lessons immediately. One course, one…","u":"/research/evidence#kestin-2025","src":"https://www.nature.com/articles/s41598-025-97652-6","on":["/research/how-will-ai-change-teaching"],"alt":["GPT"]},{"t":"study","n":"Microsoft 365 Copilot Experiment: Cross-Government Findings Report","au":"Government Digital Service","y":2025,"g":"institutional-survey","sec":"professions","f":"Average self-reported saving 26 minutes a day. Adoption reached 83 per cent and held around 80 per cent. 17 per cent noticed no clear saving; more than a third reported over half an hour. Drafting documents 24 minutes, creating…","np":"Any measured productivity effect. The headline is a self-reported estimate, bucketed, with its top band capped by the analysts, and with no control group, no baseline task timing and no measure of…","u":"/research/evidence#gds-copilot-2025","src":"https://assets.publishing.service.gov.uk/media/683db42bd23a62e5d32680d0/M365_Copilot_Experiment_Findings_Report.pdf","on":["/research/how-will-ai-change-the-public-sector","/research/how-will-ai-change-human-resources","/research/does-ai-actually-make-people-more-productive"],"alt":["M365"]},{"t":"study","n":"An Evaluation of DWP's Microsoft 365 Copilot Trial","au":"Arzilli, F., Lynch, C. S. and Page, L.","y":2026,"g":"institutional-survey","sec":"professions","f":"Estimated saving of 19 minutes a day across eight routine tasks, statistically significant across all specifications, with the largest task effects on searching for information (26 minutes), writing emails (25) and summarising (24). 90 per…","np":"A measured time saving. The evaluation's own limitations chapter names the absence of baseline data, post-treatment bias, self-selection towards AI enthusiasts through first-come first-served…","u":"/research/evidence#dwp-copilot-2026","src":"https://www.gov.uk/government/publications/an-evaluation-of-dwps-microsoft-copilot-365-trial/an-evaluation-of-dwps-microsoft-365-copilot-trial","on":["/research/how-will-ai-change-the-public-sector","/research/does-ai-actually-make-people-more-productive"],"alt":[]},{"t":"study","n":"Use of artificial intelligence in government","au":"National Audit Office","y":2024,"g":"institutional-survey","sec":"institutional","f":"37 per cent of responding bodies had deployed AI, typically one or two use cases; 70 per cent were piloting or planning, median four use cases. 21 per cent had an organisational AI strategy, with 61 per cent planning one. Of 32 bodies with…","np":"The current position. The survey was taken in autumn 2023, before the generative wave reached most departments, and the picture will have moved. Survey response is self-reported and covers 87 bodies…","u":"/research/evidence#nao-ai-government-2024","src":"https://www.nao.org.uk/reports/use-of-artificial-intelligence-in-government/","on":["/research/how-will-ai-change-the-public-sector","/research/which-ai-investments-should-we-stop"],"alt":["DSIT"]},{"t":"study","n":"The use of evidence generated by software in criminal proceedings","au":"Ministry of Justice","y":2025,"g":"institutional-modelling","sec":"institutional","f":"Records that section 69 of the Police and Criminal Evidence Act 1984, which required a party to show a computer was operating properly, was repealed on a 1997 Law Commission recommendation and replaced from 2000 by a common law rebuttable…","np":"Any outcome. It is a call for evidence rather than a decision, and no reform had been enacted at the date of review. It carries no data on how often the presumption is challenged or successfully…","u":"/research/evidence#moj-computer-evidence-2025","src":"https://assets.publishing.service.gov.uk/media/67892f5a93d4eae3088bd324/use-evidence-generated-software-criminal-proceedings.pdf","on":["/research/how-will-ai-change-the-public-sector"],"alt":[]},{"t":"study","n":"Algorithmic transparency records","au":"Government Digital Service","y":2026,"g":"compiled-review","sec":"institutional","f":"143 records published as at 1 September 2026, from central departments, agencies, local authorities, police forces and devolved administrations. Disclosed systems include a Department for Work and Pensions scanner reading around 25,000…","np":"The extent of use. The register shows what has been disclosed and cannot show what has not. It is self-declared, the count moves, and the National Audit Office found in 2024 that the standard was not…","u":"/research/evidence#atrs-register-2026","src":"https://www.gov.uk/algorithmic-transparency-records","on":["/research/how-will-ai-change-the-public-sector","/research/how-will-ai-change-human-resources"],"alt":[]},{"t":"study","n":"Gender, Race, and Intersectional Bias in Resume Screening via Language Model Retrieval","au":"Wilson, K. and Caliskan, A.","y":2024,"g":"peer-reviewed","sec":"professions","f":"The embedding models significantly favoured White-associated names in 85.1 per cent of cases and female-associated names in 11.1 per cent, with a minority of comparisons showing no statistically significant difference. Black male…","np":"The behaviour of any deployed product. These are open embedding models in a simulated pipeline, not the proprietary systems vendors sell, which add filters and thresholds and cannot be independently…","u":"/research/evidence#wilson-caliskan-2024","src":"https://arxiv.org/abs/2407.20371","on":["/research/how-will-ai-change-human-resources"],"alt":[]},{"t":"study","n":"Null Compliance: NYC Local Law 144 and the Challenges of Algorithm Accountability","au":"Wright, L., Muenster, R. M., Vecchione, B., Qu, T., Cai, P., Smith,…","y":2024,"g":"peer-reviewed","sec":"institutional","f":"18 employers posted a bias audit report, roughly 5 per cent, and 13 posted a transparency notice, roughly 3 per cent. The authors name the resulting state null compliance: non-compliance cannot be established because the law's design makes…","np":"That the audited tools are biased or unbiased. No audit results were analysed because almost none were published, which is the finding. The sample is 391 employers with a large New York workforce…","u":"/research/evidence#wright-null-compliance-2024","src":"https://facctconference.org/static/papers24/facct24-113.pdf","on":["/research/how-will-ai-change-human-resources","/research/can-ai-be-unbiased","/research/is-it-ethical-to-let-ai-judge-people"],"alt":[]},{"t":"study","n":"Regulation (EU) 2024/1689, Annex III: High-risk AI systems referred to in Article 6(2)","au":"European Parliament and Council of the European Union","y":2024,"g":"institutional-modelling","sec":"institutional","f":"Point 3 classifies as high risk AI systems used to determine access or admission to education, to evaluate learning outcomes including where those outcomes steer a learner's path, to assess the level of education a person will receive, and…","np":"Compliance or effect. Classification is not evidence that any system is biased or unsafe, and the Annex says nothing about how well the resulting obligations are met in practice.","u":"/research/evidence#eu-ai-act-annex-3","src":"https://ai-act-service-desk.ec.europa.eu/en/ai-act/annex-3","on":["/research/how-will-ai-change-human-resources","/research/how-will-ai-change-teaching","/research/is-it-ethical-to-let-ai-judge-people"],"alt":[]},{"t":"study","n":"Brain development during childhood and adolescence: a longitudinal MRI study","au":"Giedd, J.N., Blumenthal, J., Jeffries, N.O., Castellanos, F.X., Liu,…","y":1999,"g":"peer-reviewed","sec":"learning","f":"Cortical grey matter in frontal regions peaks in pre-adolescence and then thins through synaptic pruning, with the prefrontal cortex among the last areas to mature.","np":"Any threshold age. The paper reports a slower trajectory, not an endpoint, and names no age at which development completes. Everything downstream that cites it for a cut-off is citing something it…","u":"/research/evidence#giedd-1999","src":"https://pubmed.ncbi.nlm.nih.gov/10491603/","on":["/research/does-the-brain-mature-at-25"],"alt":[]},{"t":"study","n":"Dynamic mapping of human cortical development during childhood through early adulthood","au":"Gogtay, N., Giedd, J.N., Lusk, L., Hayashi, K.M., Greenstein, D.,…","y":2004,"g":"peer-reviewed","sec":"learning","f":"Maps the sequence in which cortical regions mature, with higher-order association cortices maturing after lower-order sensorimotor regions.","np":"A population age of maturity. Thirteen participants cannot establish one, and the paper does not claim to. This is the study behind the Time Magazine coverage in which the number 25 first appears in…","u":"/research/evidence#gogtay-2004","src":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC419576/","on":["/research/does-the-brain-mature-at-25"],"alt":["NIMH","THIRTEEN"]},{"t":"study","n":"Topological turning points across the human lifespan","au":"Mousley, A. and colleagues","y":2025,"g":"peer-reviewed","sec":"learning","f":"Four topological turning points, at approximately ages 9, 32, 66 and 83, dividing life into five epochs. The adolescent epoch runs from about 9 to about 32. The largest overall shift in trajectory occurs around 32, not in the mid-twenties.","np":"That 32 is the new 25. The authors describe turning points in network topology, not a moment of cognitive completion, and reading it as a new threshold would repeat the original error with a…","u":"/research/evidence#mousley-2025","src":"https://www.nature.com/articles/s41467-025-65974-8","on":["/research/does-the-brain-mature-at-25"],"alt":["MRI"]},{"t":"study","n":"When does cognitive functioning peak? The asynchronous rise and fall of different cognitive abilities across the life span","au":"Hartshorne, J.K. and Germine, L.T.","y":2015,"g":"peer-reviewed","sec":"learning","f":"Different cognitive abilities peak at different ages, some in the late teens or early twenties, others not until the forties or fifties. There is no single age at which cognitive functioning peaks.","np":"That age is irrelevant. It shows the timing is ability-specific rather than absent.","u":"/research/evidence#hartshorne-germine-2015","src":"https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4441622/","on":["/research/does-the-brain-mature-at-25"],"alt":[]},{"t":"study","n":"Challenging the 25-year-old 'mature brain' mythology: implications for the minimum legal age for non-medical cannabis use","au":"Adinoff, B. and Nunes, J.C.","y":2025,"g":"peer-reviewed","sec":"learning","f":"Argues the mature-brain-at-25 claim is not supported by the underlying neuroscience and should not be used as a basis for age thresholds in policy.","np":"Anything about AI, learning or capability. It is cited here for the status of the claim, not for the subject.","u":"/research/evidence#adinoff-nunes-2025","src":"https://www.tandfonline.com/doi/full/10.1080/00952990.2025.2561982","on":["/research/does-the-brain-mature-at-25"],"alt":[]},{"t":"study","n":"Future of Work with AI Agents: Auditing Automation and Augmentation Potential across the U.S. Workforce","au":"Shao, Y., Zope, H., Jiang, Y., Pei, J., Nguyen, D., Brynjolfsson, E.…","y":2026,"g":"working-paper","sec":"work","f":"Worker preferences diverge sharply from technical capability. Tasks fall into an Automation Green Light Zone, an Automation Red Light Zone where capability exists and workers do not want it used, an R&D Opportunity Zone and a Low Priority…","np":"Nothing about what happens to capability when a task is automated. It measures what workers WANT and what experts think is POSSIBLE, which are both stated positions rather than outcomes. A preprint,…","u":"/research/evidence#shao-workbank-2026","src":"https://arxiv.org/abs/2506.06576","on":["/research/what-stays-human","/research/questions","/research/which-tasks-do-workers-not-want-automated"],"alt":["NET"]},{"t":"study","n":"The Productivity J-Curve: How Intangibles Complement General Purpose Technologies","au":"Brynjolfsson, E., Rock, D. and Syverson, C.","y":2021,"g":"peer-reviewed","sec":"work","f":"General purpose technologies enable and require significant complementary investments that are often intangible and poorly measured in national accounts. This produces underestimation of productivity growth in a new technology's early…","np":"That AI will follow the same curve, or on what timescale. The 15.9 per cent figure is for computer hardware and software to 2017, not for AI, and the authors describe current AI intangible effects as…","u":"/research/evidence#brynjolfsson-rock-syverson-2021","src":"https://www.aeaweb.org/articles?id=10.1257%2Fmac.20180386","on":["/research/how-long-before-you-know-if-an-ai-investment-worked","/research/if-everyone-has-ai-where-is-the-advantage"],"alt":[]},{"t":"study","n":"Firm Resources and Sustained Competitive Advantage","au":"Barney, J.","y":1991,"g":"compiled-review","sec":"work","f":"Sets out four empirical indicators of the potential of a firm resource to generate sustained competitive advantage: value, rareness, imitability and substitutability. Applies the model to several firm resources and draws out implications…","np":"Anything empirical, and nothing about AI, which postdates it by three decades. It is a theory with a substantial critical literature, and applying it to a technology three years into commercial…","u":"/research/evidence#barney-1991","src":"https://journals.sagepub.com/doi/10.1177/014920639101700108","on":["/research/if-everyone-has-ai-where-is-the-advantage"],"alt":[]},{"t":"study","n":"The Microstructure of AI Diffusion: Evidence from Firms, Business Functions, and Worker Tasks","au":"Bonney, K., Breaux, C., Dinlersoz, E., Foster, L., Haltiwanger, J.…","y":2026,"g":"working-paper","sec":"work","f":"18 per cent of firms used AI in a business function, rising to 32 per cent employment-weighted, with adoption expected to reach 22 per cent within six months. Use rates reach 50 to 60 per cent, and 60 to 70 per cent employment-weighted,…","np":"Causation in either direction between adoption and performance: this is a cross-section of firms, and better-run firms may simply adopt more. Also not a stable time series. The Census Bureau…","u":"/research/evidence#census-ai-diffusion-2026","src":"https://www.census.gov/library/working-papers/2026/adrm/CES-WP-26-25.html","on":["/research/if-everyone-has-ai-where-is-the-advantage","/research/how-long-before-you-know-if-an-ai-investment-worked","/research/which-tasks-do-workers-not-want-automated"],"alt":[]},{"t":"study","n":"Monitoring AI Adoption in the U.S. Economy","au":"Allen, J. S.","y":2026,"g":"institutional-survey","sec":"work","f":"About 18 per cent of firms had adopted AI as of year-end 2025 on the BTOS. Work-related generative AI adoption in the RPS stood at about 41 per cent of the workforce as of November 2025, with daily use at 12 per cent. The SBU gives an…","np":"Any productivity or employment effect. The note explicitly does not estimate AI's contribution to output, GDP or productivity, and names those as open questions beyond its scope. It also makes no…","u":"/research/evidence#allen-fed-ai-adoption-2026","src":"https://www.federalreserve.gov/econres/notes/feds-notes/monitoring-ai-adoption-in-the-u-s-economy-20260403.html","on":["/research/if-everyone-has-ai-where-is-the-advantage","/research/how-long-before-you-know-if-an-ai-investment-worked"],"alt":["BTOS","LLM","RPS","SBU"]},{"t":"study","n":"To Trust or to Think: Cognitive Forcing Functions Can Reduce Overreliance on AI in AI-assisted Decision-making","au":"Bucinca, Z., Malaya, M. B. and Gajos, K. Z.","y":2021,"g":"peer-reviewed","sec":"collaboration","f":"Cognitive forcing significantly reduced overreliance compared with the simple explainable-AI approaches. Participants gave the least favourable subjective ratings to the designs that reduced overreliance the most. On average the…","np":"That slower decisions produce better organisational outcomes. This is a controlled task with 199 participants, not a field study, and it measures overreliance rather than downstream results. The…","u":"/research/evidence#bucinca-2021","src":"https://arxiv.org/abs/2102.09692","on":["/research/which-decisions-should-become-slower-because-of-ai","/research/how-do-you-design-a-stop-button-people-will-use"],"alt":[]},{"t":"study","n":"Visibility into AI Agents","au":"Chan, A., Ezell, C., Kaufmann, M., Wei, K., Hammond, L., Bradley, H.,…","y":2024,"g":"peer-reviewed","sec":"institutional","f":"Defines visibility as information about where, why, how and by whom AI agents are used, and assesses agent identifiers, real-time monitoring and activity logging as measures. Names five agent-specific risks: malicious use, overreliance and…","np":"That any of the proposed measures works, or is proportionate. The authors state they are describing options for further study rather than recommending deployment, and the paper contains no empirical…","u":"/research/evidence#chan-visibility-2024","src":"https://facctconference.org/static/papers24/facct24-63.pdf","on":["/research/who-manages-ai-agents"],"alt":[]},{"t":"study","n":"Governing AI Agents","au":"Kolt, N.","y":2025,"g":"compiled-review","sec":"institutional","f":"Characterises three problems arising from AI agents in agency terms: information asymmetry, discretionary authority and loyalty. Argues that the conventional solutions to agency problems, incentive design, monitoring and enforcement, might…","np":"Anything about how agents behave in practice. It is a law review article arguing a position, cited here for its framework rather than as evidence of an outcome, and at the time of writing it is…","u":"/research/evidence#kolt-2025","src":"https://arxiv.org/abs/2501.07913","on":["/research/who-manages-ai-agents"],"alt":[]},{"t":"study","n":"Regulation (EU) 2024/1689, Article 26: Obligations of deployers of high-risk AI systems","au":"European Union","y":2024,"g":"institutional-modelling","sec":"institutional","f":"Paragraph 2 requires that deployers assign human oversight to natural persons who have the necessary competence, training and authority, as well as the necessary support. Paragraph 5 requires deployers to monitor operation on the basis of…","np":"That it applies to most commercial agent deployments, which will fall outside the high-risk classification. Nor that any of it happens: the Regulation creates duties and does not evidence compliance.…","u":"/research/evidence#eu-ai-act-art-26","src":"https://artificialintelligenceact.eu/article/26/","on":["/research/who-manages-ai-agents","/research/which-decisions-should-become-slower-because-of-ai"],"alt":["III"]},{"t":"study","n":"Magnifica Humanitas: Encyclical Letter on Safeguarding the Human Person in the Time of Artificial Intelligence","au":"Leo XIV","y":2026,"g":"institutional-modelling","sec":"institutional","f":"States that AI use can weaken human capability, in terms specific enough to quote. Paragraph 100: heavy reliance and the search for ready-made answers can \"weaken personal creativity and judgment\". Paragraph 140: \"every technology shapes…","np":"Nothing empirical whatsoever. It measures nothing, samples nobody and tests no hypothesis, and its deskilling claim is a quotation from an earlier Vatican note which is itself not an empirical study.…","u":"/research/evidence#leo-xiv-magnifica-humanitas-2026","src":"https://www.vatican.va/content/leo-xiv/en/encyclicals/documents/20260515-magnifica-humanitas.html","on":["/research/the-best-writing-on-ai"],"alt":[]},{"t":"study","n":"Decision Provenance: Harnessing Data Flow for Accountable Systems","au":"Singh, J., Cobbe, J. and Norval, C.","y":2019,"g":"argued-perspective","sec":"institutional","f":"Introduces decision provenance, defined as using provenance methods to provide information exposing decision pipelines: the chains of inputs to a decision, the nature of the decision, and the flow-on effects from the decisions and actions…","np":"Anything about whether decision provenance works, is affordable, or improves any outcome. It reports no data of any kind, proposes an implementation agenda rather than an implementation, and predates…","u":"/research/evidence#singh-cobbe-norval-2019","src":"https://arxiv.org/abs/1804.05741","on":["/research/what-is-decision-provenance"],"alt":["EPSRC","P024394","R033501"]},{"t":"study","n":"AI Assistance Reduces Persistence and Hurts Independent Performance","au":"Liu, G., Christian, B., Dumbalska, T., Bakker, M. A. and Dubey, R.","y":2026,"g":"working-paper","sec":"learning","f":"AI assistance improves performance in the short term, and people then perform significantly worse without AI and are more likely to give up. The authors report that these effects emerge after only brief interactions, approximately 10…","np":"Anything about sustained professional practice. These are short online tasks and the measured effect is a within-session carry-over rather than skill decay, so it cannot show whether the effect…","u":"/research/evidence#liu-persistence-2026","src":"https://arxiv.org/abs/2604.04721","on":["/research/the-best-writing-on-ai","/research/does-using-ai-stop-you-learning"],"alt":[]},{"t":"study","n":"The Generative AI Learning Penalty: Evidence from Chinese Secondary Education","au":"Stromberg, D., Lei, V. and Wu, Y.","y":2026,"g":"working-paper","sec":"learning","f":"AI adoption raises homework scores by 18 per cent and reduces completion time by 30 per cent, and lowers monthly exam scores by 20 per cent within six months. High-stakes entrance-exam scores fall by 18 and 24 per cent, with the full…","np":"Causation with the confidence of a randomised trial. Adoption is self-selected and staggered rather than assigned, the outsourcing split is inferred from time-on-homework rather than observed, and a…","u":"/research/evidence#stromberg-lei-wu-2026","src":"https://cepr.org/publications/dp21577","on":["/research/the-best-writing-on-ai","/research/does-using-ai-stop-you-learning"],"alt":["STEM"]},{"t":"study","n":"Insights into the Problem of Alarm Fatigue with Physiologic Monitor Devices: A Comprehensive Observational Study of Consecutive Intensive Care Unit…","au":"Drew, B. J., Harris, P., Zegre-Hemsey, J. K., Mammone, T., Schindler,…","y":2014,"g":"peer-reviewed","sec":"frontline","f":"2,558,760 unique alarms occurred in 31 days: 1,154,201 arrhythmia, 612,927 parameter and 791,632 technical. 381,560 were audible, an audible alarm burden of 187 per bed per day. 88.8 per cent of the 12,671 annotated arrhythmia alarms were…","np":"Anything about AI. The 88.8 per cent applies only to the 12,671 annotated arrhythmia alarms, NOT to all 2.56 million alarms and not to clinical alarms in general; a page stating that 88.8 per cent of…","u":"/research/evidence#drew-2014","src":"https://journals.plos.org/plosone/article?id=10.1371%2Fjournal.pone.0110274","on":["/research/how-do-you-design-a-stop-button-people-will-use","/research/what-is-alarm-fatigue"],"alt":["ECG"]},{"t":"study","n":"Sentinel Event Alert 50: Medical device alarm safety in hospitals","au":"The Joint Commission","y":2013,"g":"institutional-survey","sec":"institutional","f":"98 alarm-related events between January 2009 and June 2012, of which 80 resulted in death, 13 in permanent loss of function and five in unexpected additional care or extended stay. 94 of the events occurred in hospitals. Contributing…","np":"The size of the problem. The Commission's own footnote states that reporting is voluntary, represents only a small proportion of actual events, and that no conclusions should be drawn about relative…","u":"/research/evidence#joint-commission-sea50-2013","src":"https://digitalassets.jointcommission.org/api/public/content/f65e5c9df2b94000a99445e0a7877007","on":["/research/how-do-you-design-a-stop-button-people-will-use","/research/what-is-alarm-fatigue"],"alt":["ECRI","FDA","MAUDE","NPSG"]},{"t":"study","n":"Collision Between Vehicle Controlled by Developmental Automated Driving System and Pedestrian, Tempe, Arizona, March 18, 2018","au":"National Transportation Safety Board","y":2019,"g":"statutory-investigation","sec":"institutional","f":"The automated driving system detected the pedestrian 5.6 seconds before impact and tracked her to the crash without ever classifying her correctly or predicting her path. The developer had disengaged the Volvo XC90's factory forward…","np":"That the absence of an alert caused the crash. NTSB determined the probable cause to be the operator's inattention and found she would likely have had sufficient time to react had she been attentive.…","u":"/research/evidence#ntsb-tempe-2019","src":"https://www.ntsb.gov/investigations/AccidentReports/Reports/HAR1903.pdf","on":["/research/how-do-you-design-a-stop-button-people-will-use","/research/what-is-alarm-fatigue","/research/can-a-human-approve-an-ai-decision-at-machine-speed"],"alt":["NTSB","XC90"]},{"t":"study","n":"'We can stop work, but then nothing gets done.' Factors that support and hinder a workforce to discontinue work for safety","au":"Weber, D. E., MacGregor, S. C., Provan, D. J. and Rae, A.","y":2018,"g":"peer-reviewed","sec":"frontline","f":"Stopping work for safety was reported as challenging at the sharp operational end despite the authority existing and carrying no formal penalty. The authors conclude that stopping an unsafe task 'does not solely hinge on the willingness of…","np":"Any rate or frequency. Ten focus groups in one industry in one country, qualitative by design, with no measurement of how often stops occurred or should have.","u":"/research/evidence#weber-stop-work-2018","src":"https://forgeworks.com/wp-content/uploads/2022/09/Authority-to-Stop-Work.pdf","on":["/research/how-do-you-design-a-stop-button-people-will-use"],"alt":[]},{"t":"study","n":"Hierarchies and the Organization of Knowledge in Production","au":"Garicano, L.","y":2000,"g":"peer-reviewed","sec":"work","f":"A knowledge-based hierarchy is a natural way to organise the acquisition of knowledge when matching problems with those who know how to solve them is costly. Production workers acquire knowledge of the most common or easiest problems and…","np":"Anything measured. It is a model, published a quarter of a century before generative AI, and it contains no data about firms, layers or technology adoption.","u":"/research/evidence#garicano-2000","src":"https://www.journals.uchicago.edu/doi/abs/10.1086/317671","on":["/research/what-happens-to-work-that-moves-information"],"alt":[]},{"t":"study","n":"The Distinct Effects of Information Technology and Communication Technology on Firm Organization","au":"Bloom, N., Garicano, L., Sadun, R. and Van Reenen, J.","y":2014,"g":"peer-reviewed","sec":"work","f":"Information technology is a decentralising force and communication technology is a centralising force. Better information technologies, ERP for plant managers and computer-assisted design or manufacturing for production workers, are…","np":"Anything about generative AI, which is both kinds of technology in one interface. Manufacturing plants only, data predating 2009. The sample description above is verified in the working paper; the…","u":"/research/evidence#bloom-ict-2014","src":"https://researchonline.lse.ac.uk/id/eprint/79246/","on":["/research/what-happens-to-work-that-moves-information"],"alt":["CEP","ERP","ICT"]},{"t":"study","n":"Corporate Hierarchy","au":"Ewens, M. and Giroud, X.","y":2025,"g":"working-paper","sec":"work","f":"Firms average ten hierarchical layers and a pyramidal structure, with the average and median number of layers declining across the sample period. More hierarchical firms show a more educated workforce, higher internal promotion rates,…","np":"That AI causes flattening. The authors call their own tests under-powered at the 10 per cent level, adoption is measured by job postings rather than by use, hierarchy is inferred from self-reported…","u":"/research/evidence#ewens-giroud-2025","src":"https://www.nber.org/papers/w34162","on":["/research/what-happens-to-work-that-moves-information"],"alt":[]},{"t":"study","n":"Firm Investments in Artificial Intelligence Technologies and Changes in Workforce Composition","au":"Babina, T., Fedyk, A., He, A. and Hodson, J.","y":2025,"g":"peer-reviewed","sec":"work","f":"AI investments are associated with a flattening of firms' hierarchical structure, with significant increases in the share of workers at the junior level and decreases in the shares in middle-management and senior roles.","np":"Any magnitude. The abstract read for this entry states direction and significance and no percentage, so none should be attributed to it. Association rather than causation, and firms that invest in AI…","u":"/research/evidence#babina-hierarchy-2025","src":"https://www.nber.org/books-and-chapters/technology-productivity-and-economic-growth/firm-investments-artificial-intelligence-technologies-and-changes-workforce-composition","on":["/research/what-happens-to-work-that-moves-information"],"alt":[]},{"t":"study","n":"The effects of remote work on collaboration among information workers","au":"Yang, L., Holtz, D., Jaffe, S., Suri, S., Sinha, S., Weston, J.,…","y":2022,"g":"peer-reviewed","sec":"work","f":"Firm-wide remote work caused the collaboration network to become more static and siloed, with fewer bridges between disparate parts of the organisation, a decrease in synchronous and an increase in asynchronous communication. The authors…","np":"Anything about AI, and nothing about outcomes. It measures communication structure rather than performance, in one very large technology company, during a pandemic. Full text is paywalled; this entry…","u":"/research/evidence#yang-remote-2022","src":"https://www.nature.com/articles/s41562-021-01196-4","on":["/research/what-happens-to-work-that-moves-information"],"alt":[]},{"t":"study","n":"Experimental evidence on the productivity effects of generative artificial intelligence","au":"Noy, S. and Zhang, W.","y":2023,"g":"peer-reviewed","sec":"collaboration","f":"Average time taken fell by 40 per cent and output quality rose by 18 per cent. Time on the post-treatment task dropped by 11 minutes, 0.75 standard deviations, against a control mean of 27 minutes, and evaluator grades rose by 0.45…","np":"That the effect generalises. The authors say they examined a limited range of occupations and tasks in which ChatGPT may be unusually useful, and speculate that real-economy effects will be somewhat…","u":"/research/evidence#noy-zhang-2023","src":"https://www.science.org/doi/10.1126/science.adh2586","on":["/research/what-becomes-more-valuable-as-ai-gets-cheaper","/research/does-ai-actually-make-people-more-productive"],"alt":["GPT"]},{"t":"study","n":"The Impact of AI on Developer Productivity: Evidence from GitHub Copilot","au":"Peng, S., Kalliamvakou, E., Cihon, P. and Demirer, M.","y":2023,"g":"working-paper","sec":"collaboration","f":"Conditional on completion, the treated group averaged 71.17 minutes against 160.89 for control, a 55.8 per cent reduction in completion time, p = 0.0017, with a 95 per cent confidence interval on the improvement of 21 to 89 per cent.…","np":"Anything about quality. The authors state the study does not examine the effects of AI on code quality. The effect is estimated on 35 completers per arm rather than 95 participants, the task is…","u":"/research/evidence#peng-copilot-2023","src":"https://arxiv.org/abs/2302.06590","on":["/research/what-becomes-more-valuable-as-ai-gets-cheaper","/research/does-ai-actually-make-people-more-productive","/research/will-ai-replace-programmers"],"alt":["HTTP"]},{"t":"study","n":"Tracking Firm Use of AI in Real Time: A Snapshot from the Business Trends and Outlook Survey","au":"Bonney, K., Breaux, C., Buffington, C., Dinlersoz, E., Foster, L.,…","y":2024,"g":"working-paper","sec":"work","f":"Bi-weekly estimates of the AI use rate rose from 3.7 to 5.4 per cent, with an expected rate of about 6.6 per cent by early autumn 2024. 94.6 per cent of AI-using businesses reported no net change in employment in the previous six months…","np":"Any effect of AI on firm performance, by the authors' own statement. Adoption rates here are not comparable with later Census figures, because the question was broadened in November 2025 from use in…","u":"/research/evidence#census-btos-2024","src":"https://www.census.gov/hfp/btos/downloads/CES-WP-24-16.pdf","on":["/research/what-becomes-more-valuable-as-ai-gets-cheaper","/research/which-ai-investments-should-we-stop"],"alt":[]},{"t":"study","n":"Measuring the Self-Reported Impact of Early-2026 AI on Technical Worker Productivity","au":"Becker, J. (METR)","y":2026,"g":"working-paper","sec":"collaboration","f":"Median self-reported change in the VALUE of work is 1.4 to 2 times, and median self-reported SPEED change is 3 times. METR state that to their knowledge only one study has gathered survey and field experiment results on the same population…","np":"That self-report always overstates. The GitHub Copilot randomised trial found the opposite direction, with participants estimating 35 per cent against a measured 55.8 per cent. A survey of a…","u":"/research/evidence#metr-selfreport-2026","src":"https://metr.org/blog/2026-05-11-ai-usage-survey/","on":["/research/what-becomes-more-valuable-as-ai-gets-cheaper","/research/does-ai-actually-make-people-more-productive"],"alt":["METR","SPEED","VALUE"]},{"t":"study","n":"Pulling the Plug: Software Project Management and the Problem of Project Escalation","au":"Keil, M.","y":1995,"g":"peer-reviewed","sec":"institutional","f":"CONFIG, an expert system built to help sales representatives produce error-free configurations before quoting, ran for over a decade and was terminated at the end of 1992 after, in the author's words, tens of millions of dollars.…","np":"Any prevalence. N is one, the organisation is pseudonymised, and there is no comparison group. An Academia.edu machine-generated summary of this paper asserts the project absorbed 250 million…","u":"/research/evidence#keil-1995","src":"https://misq.umn.edu/pulling-the-plug-software-project-management-and-the-problem-of-project-escalation.html","on":["/research/which-ai-investments-should-we-stop"],"alt":["CONFIG"]},{"t":"study","n":"Why Software Projects Escalate: An Empirical Analysis and Test of Four Theoretical Models","au":"Keil, M., Mann, J. and Rai, A.","y":2000,"g":"peer-reviewed","sec":"institutional","f":"The authors state that between 30 and 40 per cent of all IS projects exhibit some degree of escalation. The completion effect derived from approach-avoidance theory gave the best classification, correctly classifying over 70 per cent of…","np":"Anything about AI, and nothing about whether escalated projects should have been stopped: it shows their outcomes were worse. Some degree of escalation is a soft threshold, and the respondents are…","u":"/research/evidence#keil-mann-rai-2000","src":"https://misq.umn.edu/why-software-projects-escalate-an-empirical-analysis-and-test-of-four-theoretical-models.html","on":["/research/which-ai-investments-should-we-stop"],"alt":[]},{"t":"study","n":"De-escalating Information Technology Projects: Lessons from the Denver International Airport","au":"Montealegre, R. and Keil, M.","y":2000,"g":"peer-reviewed","sec":"institutional","f":"De-escalation runs as a four-phase process: problem recognition, re-examination of prior course of action, search for alternative course of action, and implementing an exit strategy. The authors note that while escalation is well…","np":"How often de-escalation is attempted or succeeds. N is one, the model is inductive rather than tested, and the case is a physical baggage system in the 1990s. The authors describe four phases, not…","u":"/research/evidence#montealegre-keil-2000","src":"https://misq.umn.edu/de-escalating-information-technology-projects-lessons-from-the-denver-international-airport.html","on":["/research/which-ai-investments-should-we-stop"],"alt":[]},{"t":"study","n":"Knee-Deep in the Big Muddy: A Study of Escalating Commitment to a Chosen Course of Action","au":"Staw, B. M.","y":1976,"g":"peer-reviewed","sec":"judgement","f":"Participants personally responsible for the earlier investment allocated an average of 11.08 million dollars to the division they had chosen, against 8.89 million where another officer had chosen it. Negative consequences drew 11.20…","np":"Anything about real organisations or real money. A single-session paper exercise with undergraduates, no longitudinal element, and no prevalence claim. Staw himself flags the ambiguity between…","u":"/research/evidence#staw-1976","src":"https://doi.org/10.1016/0030-5073(76)90005-2","on":["/research/which-ai-investments-should-we-stop"],"alt":[]},{"t":"study","n":"The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed","au":"Ryseff, J., De Bruhl, B. and Newberry, S. J.","y":2024,"g":"institutional-survey","sec":"institutional","f":"A set of organisational anti-patterns behind AI project failure, chiefly misunderstood or miscommunicated intent, inadequate data, focus on the technology rather than the problem, missing infrastructure, and problems the technology cannot…","np":"Any base rate. This is 65 expert interviews about causes, not a measurement of frequency, and RAND do not claim otherwise. The two academic sources it cites on failure factors are themselves…","u":"/research/evidence#rand-ai-failure-2024","src":"https://www.rand.org/content/dam/rand/pubs/research_reports/RRA2600/RRA2680-1/RAND_RRA2680-1.pdf","on":["/research/which-ai-investments-should-we-stop"],"alt":[]},{"t":"study","n":"MABA-MABA or Abracadabra? Progress on Human-Automation Co-ordination","au":"Dekker, S. W. A. and Woods, D. D.","y":2002,"g":"peer-reviewed","sec":"collaboration","f":"Substitution-based function allocation, of which the Fitts list is the archetype, cannot deliver human-automation coordination, because the effects of automation are qualitative rather than quantitative. The authors name the underlying…","np":"How large the effect is, or anything measurable at all. It is an argument. Its supporting accident examples are cited rather than analysed.","u":"/research/evidence#dekker-woods-2002","src":"https://doi.org/10.1007/s101110200022","on":["/research/how-do-humans-and-agents-divide-work"],"alt":[]},{"t":"study","n":"Gaps in the continuity of care and progress on patient safety","au":"Cook, R. I., Render, M. and Woods, D. D.","y":2000,"g":"peer-reviewed","sec":"frontline","f":"Gaps are defined as discontinuities in care, appearing as losses of information or momentum or interruptions in delivery. Most gaps are anticipated and bridged by practitioners, so invisibly that neither outsiders nor insiders recognise…","np":"Anything quantified. There is no sample, no rate and no measurement, and nothing in it concerns automation or AI. It is a framing paper.","u":"/research/evidence#cook-render-woods-2000","src":"https://doi.org/10.1136/bmj.320.7237.791","on":["/research/how-do-humans-and-agents-divide-work"],"alt":[]},{"t":"study","n":"Changes in Medical Errors after Implementation of a Handoff Program","au":"Starmer, A. J., Spector, N. D., Srivastava, R., West, D. C.,…","y":2014,"g":"peer-reviewed","sec":"frontline","f":"The medical-error rate fell 23 per cent, from 24.5 to 18.8 per 100 admissions, and preventable adverse events fell 30 per cent, from 4.7 to 3.3 per 100 admissions, both P<0.001. Near misses and non-harmful errors fell 21 per cent.…","np":"Causation, by the authors' own statement, and not which element of the bundle did the work. Paediatric inpatient units only; generalisation to other specialties is untested. Reviewer agreement was…","u":"/research/evidence#starmer-ipass-2014","src":"https://doi.org/10.1056/NEJMsa1405556","on":["/research/how-do-humans-and-agents-divide-work"],"alt":[]},{"t":"study","n":"Why Do Multi-Agent LLM Systems Fail?","au":"Cemri, M., Pan, M. Z., Yang, S., Agrawal, L. A., Chopra, B., Tiwari,…","y":2025,"g":"peer-reviewed","sec":"collaboration","f":"Failures divide into system design issues 41.8 per cent, inter-agent misalignment 36.9 per cent and task verification 21.3 per cent. Inter-agent misalignment is defined as a breakdown in critical information flow during interaction and…","np":"Anything about human-agent boundaries. Every handoff in the dataset is agent to agent and no humans appear anywhere. The authors caveat that the 210-trace distribution illustrates system-specific…","u":"/research/evidence#cemri-mast-2025","src":"https://arxiv.org/abs/2503.13657","on":["/research/how-do-humans-and-agents-divide-work"],"alt":[]},{"t":"study","n":"Descent Below Visual Glidepath and Impact With Seawall, Asiana Airlines Flight 214, Boeing 777-200ER, HL7742, San Francisco, California, July 6, 2013","au":"National Transportation Safety Board","y":2014,"g":"statutory-investigation","sec":"frontline","f":"The pilot flying selected a mode that caused the autoflight system to climb, then disconnected the autopilot and moved the thrust levers to idle, which put the autothrottle into HOLD, a mode in which it does not control airspeed. The board…","np":"Any frequency of mode confusion. One accident, three crew, one aircraft type. It cannot support a rate and is used here as a demonstration of a mechanism.","u":"/research/evidence#ntsb-asiana-2014","src":"https://www.ntsb.gov/investigations/AccidentReports/Reports/AAR1401.pdf","on":["/research/how-do-humans-and-agents-divide-work"],"alt":["HOLD"]},{"t":"study","n":"Sentinel Event Alert 58: Inadequate hand-off communication","au":"The Joint Commission","y":2017,"g":"compiled-review","sec":"frontline","f":"States that inadequate hand-off communication contributes to adverse events including wrong-site surgery, delay in treatment, falls and medication errors. Reports, from cited third parties, that communication failures were responsible at…","np":"The most quoted claim attached to it. The statement that 80 per cent of serious medical errors involve miscommunication during handoff does not appear anywhere in this alert; the full six pages were…","u":"/research/evidence#joint-commission-sea58-2017","src":"https://www.jointcommission.org/en-us/knowledge-library/newsletters/sentinel-event-alert/issue-58","on":["/research/how-do-humans-and-agents-divide-work"],"alt":["PASS"]},{"t":"study","n":"Regulation (EU) 2024/1689, Article 13: Transparency and provision of information to deployers","au":"European Parliament and Council of the European Union","y":2024,"g":"institutional-modelling","sec":"institutional","f":"Requires high-risk systems to be designed so their operation is sufficiently transparent for deployers to interpret the output and use it appropriately, and to be accompanied by instructions for use stating the level of accuracy and its…","np":"Any requirement to communicate uncertainty at the point of use. The word uncertainty appears in neither Article 13, Article 15 nor Article 50, and nothing requires a system to tell the person in…","u":"/research/evidence#eu-ai-act-art-13","src":"https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-13","on":["/research/how-should-an-ai-agent-communicate-uncertainty"],"alt":[]},{"t":"study","n":"What large language models know and what people think they know","au":"Steyvers, M., Tejeda, H., Kumar, A., Belem, C., Karny, S., Hu, X.,…","y":2025,"g":"peer-reviewed","sec":"collaboration","f":"Model confidence discriminates correct from incorrect answers at AUC 0.751 for GPT-3.5, 0.746 for PaLM2 and 0.781 for GPT-4o. Participants reading default explanations reached 0.589, 0.602 and 0.592, which the authors describe as only…","np":"That closing the gap improves task accuracy. The modified-explanation result is a simulation via post-hoc filtering rather than a live deployment. Participants had no domain expertise and their own…","u":"/research/evidence#steyvers-2025","src":"https://doi.org/10.1038/s42256-024-00976-7","on":["/research/how-should-an-ai-agent-communicate-uncertainty","/research/ai-and-human-disagreement"],"alt":["AUC","GPT"]},{"t":"study","n":"Can LLMs Express Their Uncertainty? An Empirical Evaluation of Confidence Elicitation in LLMs","au":"Xiong, M., Hu, Z., Lu, X., Li, Y., Fu, J., He, J. and Hooi, B.","y":2024,"g":"peer-reviewed","sec":"collaboration","f":"Average expected calibration error for plain verbalised confidence is 0.520 for GPT-3, 0.461 for Vicuna, 0.436 for LLaMA 2, 0.377 for GPT-3.5 and 0.180 for GPT-4. GPT-4's average AUROC is 62.7 per cent against a 50 per cent chance…","np":"Anything about how users read these numbers, since no humans were involved. Mitigation strategies reduce ECE substantially, to 0.028 in the best case, while still failing to predict incorrect answers…","u":"/research/evidence#xiong-confidence-2024","src":"https://arxiv.org/abs/2306.13063","on":["/research/how-should-an-ai-agent-communicate-uncertainty"],"alt":["AUPRC","AUROC","GPT"]},{"t":"study","n":"I'm Not Sure, But...: Examining the Impact of Large Language Models' Uncertainty Expression on User Reliance and Trust","au":"Kim, S. S. Y., Liao, Q. V., Vorvoreanu, M., Ballard, S. and Wortman…","y":2024,"g":"peer-reviewed","sec":"collaboration","f":"Access to the system raised agreement to 80.9 per cent against 58.4 per cent without, and lowered accuracy to 63.9 per cent against 74.2 per cent without. First-person uncertainty expression significantly reduced agreement to 74.8 per cent…","np":"That uncertainty expression should be mandated. The authors state directly that regulators should avoid blanket requirements until more research is done, and flag that their system had low accuracy…","u":"/research/evidence#kim-uncertainty-2024","src":"https://doi.org/10.1145/3630106.3658941","on":["/research/how-should-an-ai-agent-communicate-uncertainty","/research/ai-and-human-disagreement"],"alt":[]},{"t":"study","n":"Relying on the Unreliable: The Impact of Language Models' Reluctance to Express Uncertainty","au":"Zhou, K., Hwang, J. D., Ren, X. and Sap, M.","y":2024,"g":"peer-reviewed","sec":"collaboration","f":"Only about 5 per cent of generated answers include any epistemic marker. Among confidently expressed responses the error rate averages 47 per cent, and only 53 per cent of generations expressing certainty are correct. In the human study,…","np":"Any of it with statistical rigour on the human side. No total human N is stated in the text, and no p values, confidence intervals or effect sizes are reported for any human result. Twenty-five…","u":"/research/evidence#zhou-unreliable-2024","src":"https://aclanthology.org/2024.acl-long.198/","on":["/research/how-should-an-ai-agent-communicate-uncertainty","/research/ai-and-human-disagreement"],"alt":["MMLU"]},{"t":"study","n":"Effective communication of uncertainty in the IPCC reports","au":"Budescu, D. V., Por, H.-H. and Broomell, S. B.","y":2012,"g":"peer-reviewed","sec":"judgement","f":"Mean estimates were 41 for very unlikely, 44 for unlikely, 54 for likely and 62 for very likely, against IPCC guidelines of below 10, below 33, above 66 and above 90. Consistency with the guidelines was 20.76 per cent in the control, 18.81…","np":"That this settles the design. Only four of the seven IPCC terms were tested, no lower or upper bound data were collected, and the authors decline to read the result as a criticism of the IPCC, noting…","u":"/research/evidence#budescu-2012","src":"https://doi.org/10.1007/s10584-011-0330-3","on":["/research/how-should-an-ai-agent-communicate-uncertainty"],"alt":["TESS"]},{"t":"study","n":"The interpretation of IPCC probabilistic statements around the world","au":"Budescu, D. V., Por, H.-H., Broomell, S. B. and Smithson, M.","y":2014,"g":"peer-reviewed","sec":"judgement","f":"Laypeople interpret IPCC statements as conveying probabilities closer to 50 per cent than intended by the IPCC authors. Supplementing verbal terms with numerical ranges increases correspondence with the guidelines and improves…","np":"Any magnitude quotable from this estate. The article is paywalled and only the abstract was read on 4 September 2026, so no participant count, per-country sample or consistency percentage is given…","u":"/research/evidence#budescu-2014","src":"https://doi.org/10.1038/nclimate2194","on":["/research/how-should-an-ai-agent-communicate-uncertainty"],"alt":[]},{"t":"study","n":"Effect of confidence and explanation on accuracy and trust calibration in AI-assisted decision making","au":"Zhang, Y., Liao, Q. V. and Bellamy, R. K. E.","y":2020,"g":"peer-reviewed","sec":"collaboration","f":"Showing confidence scores significantly increased trust, F(1,64)=4.64, p=.035, and significantly improved trust calibration when model confidence was above 80 per cent, F(4,256)=15.8, p<.001. There was no significant difference in…","np":"Much on its own. Nine participants per cell in Experiment 1 and 27 in Experiment 2, no effect sizes, confidence intervals or standard deviations reported, no exclusions or attention checks described,…","u":"/research/evidence#zhang-liao-2020","src":"https://doi.org/10.1145/3351095.3372852","on":["/research/how-should-an-ai-agent-communicate-uncertainty"],"alt":[]},{"t":"study","n":"The association between adolescent well-being and digital technology use","au":"Orben, A. and Przybylski, A. K.","y":2019,"g":"peer-reviewed","sec":"learning","f":"The association between digital technology use and adolescent wellbeing is negative but small, explaining at most 0.4 per cent of the variation in wellbeing, which the authors state is too small to warrant policy change. In YRBS, regularly…","np":"Causation in either direction. The authors state it is possible the associations they document, and those previously documented, are spurious, and that noisy self-report measurement could itself have…","u":"/research/evidence#orben-przybylski-2019","src":"https://doi.org/10.1038/s41562-018-0506-1","on":["/research/is-screen-time-the-same-argument-as-ai-use"],"alt":["MCS","MTF","YRBS"]},{"t":"study","n":"Screens, Teens, and Psychological Well-Being: Evidence From Three Time-Use-Diary Studies","au":"Orben, A. and Przybylski, A. K.","y":2019,"g":"peer-reviewed","sec":"learning","f":"Little evidence of substantial negative associations between digital screen engagement and adolescent wellbeing, whether measured across the day or before bedtime. Correlations between diary-recorded and retrospectively self-reported…","np":"That diaries are ground truth. They remain recall-based and the authors note brief or concurrent uses may not be recorded. Still cross-sectional, and it measures totals and timing rather than the…","u":"/research/evidence#orben-przybylski-diaries-2019","src":"https://doi.org/10.1177/0956797619830329","on":["/research/is-screen-time-the-same-argument-as-ai-use"],"alt":[]},{"t":"study","n":"A systematic review and meta-analysis of discrepancies between logged and self-reported digital media use","au":"Parry, D. A., Davidson, B. I., Sewall, C. J. R., Fisher, J. T.,…","y":2021,"g":"peer-reviewed","sec":"learning","f":"The correlation between self-reported and logged digital media use is positive but medium, r = 0.38, 95 per cent CI 0.33 to 0.42. For problematic use it falls to r = 0.25 across 40 effect sizes from 19 studies. Over-reporting and…","np":"That self-report is useless, or that logs are ground truth: the authors note potential biases in log data too. They state it is an open question whether the discrepancy is random or systematic error.…","u":"/research/evidence#parry-2021","src":"https://doi.org/10.1038/s41562-021-01117-5","on":["/research/is-screen-time-the-same-argument-as-ai-use"],"alt":[]},{"t":"study","n":"From social media to artificial intelligence: improving research on digital harms in youth","au":"Mansfield, K. L., Ghai, S., Hakman, T., Ballou, N., Vuorre, M. and…","y":2025,"g":"compiled-review","sec":"learning","f":"Self-reported screen time is problematic as a measure, being imprecise and prone to bias, and as a construct, being unidimensional, homogenous and of little validity, failing to distinguish social, educational, entertainment, work and…","np":"Anything empirical. It is a commentary with no new data and no head-to-head methodological comparison. Its prescription is finer-grained measurement of which AI in what context, which is adjacent to…","u":"/research/evidence#mansfield-2025","src":"https://doi.org/10.1016/S2352-4642(24)00332-8","on":["/research/is-screen-time-the-same-argument-as-ai-use"],"alt":[]},{"t":"study","n":"Annual Research Review: Adolescent mental health in the digital age: facts, fears, and future directions","au":"Odgers, C. L. and Jensen, M. R.","y":2020,"g":"compiled-review","sec":"learning","f":"Most research to date has been correlational, focused on adults rather than adolescents, and has produced a mix of conflicting small positive, negative and null associations. The most recent and rigorous large-scale preregistered studies…","np":"Anything new. It is a review, not primary data, it does not test displacement of any specific activity, and it does not address AI. Read in the NIH author manuscript.","u":"/research/evidence#odgers-jensen-2020","src":"https://doi.org/10.1111/jcpp.13190","on":["/research/is-screen-time-the-same-argument-as-ai-use"],"alt":[]},{"t":"study","n":"A Large-Scale Test of the Goldilocks Hypothesis: Quantifying the Relations Between Digital-Screen Use and the Mental Well-Being of Adolescents","au":"Przybylski, A. K. and Weinstein, N.","y":2017,"g":"peer-reviewed","sec":"learning","f":"Links between digital screen time and mental wellbeing are described by quadratic rather than linear functions, with inflection points at 1 hour 40 minutes for weekday video-game play and 1 hour 57 minutes for weekday smartphone use,…","np":"Causation, and not displacement. The paper names the displacement hypothesis as the field's dominant assumption and calls for future work systematically analysing what is being displaced or…","u":"/research/evidence#przybylski-weinstein-2017","src":"https://doi.org/10.1177/0956797616678438","on":["/research/is-screen-time-the-same-argument-as-ai-use"],"alt":[]},{"t":"study","n":"United Kingdom Chief Medical Officers' commentary on screen-based activities and children and young people's mental health and psychosocial wellbeing","au":"Davies, S. C., Atherton, F., Calderwood, C. and McBride, M.","y":2019,"g":"compiled-review","sec":"institutional","f":"States that scientific research is currently insufficiently conclusive to support UK CMO evidence-based guidelines on optimal amounts of screen use or online activities, and that the research does not present evidence of a causal…","np":"That screen use is harmless. The CMOs explicitly state that the absence of evident causal effect does not mean there is no effect. It is a commentary on a map of reviews rather than primary research,…","u":"/research/evidence#uk-cmo-screentime-2019","src":"https://assets.publishing.service.gov.uk/government/uploads/system/uploads/attachment_data/file/777026/UK_CMO_commentary_on_screentime_and_social_media_map_of_reviews.pdf","on":["/research/is-screen-time-the-same-argument-as-ai-use"],"alt":["CMO"]},{"t":"study","n":"Alzheimer's disease mortality among taxi and ambulance drivers: population based cross sectional study","au":"Patel, V. R., Liu, M., Worsham, C. M. and Jena, A. B.","y":2024,"g":"peer-reviewed","sec":"frontline","f":"3.9 per cent of deaths (348,328) had Alzheimer's disease listed as a cause. Among 16,658 taxi drivers, 171 did (1.03 per cent); among 1,348 ambulance drivers, 10 did (0.74 per cent). After adjustment, taxi and ambulance drivers had the…","np":"That navigation protects anyone. Three limitations do most of the damage, two of them raised by Tara Spires-Jones for the Science Media Centre. The drivers died at around 64 to 67 against 74 for…","u":"/research/evidence#patel-2024","src":"https://bmjgroup.com/alzheimers-disease-deaths-lowest-among-taxi-and-ambulance-drivers/","on":["/research/does-gps-damage-your-brain"],"alt":["BMJ"]},{"t":"study","n":"Labor market impacts of AI: A new measure and early evidence","au":"Massenkoff, M. and McCrory, P.","y":2026,"g":"vendor-research","sec":"work","f":"No systematic increase in unemployment for highly exposed workers since late 2022; the pooled difference-in-differences estimate is small and indistinguishable from zero. On hiring, the monthly job-finding rate for 22 to 25 year olds…","np":"The 14 per cent is routinely restated as a fall against low-exposure peers. It is not. The paper's own words are 'compared to that in 2022 in the exposed occupations', so it is a change over time…","u":"/research/evidence#massenkoff-mccrory-2026","src":"https://www.anthropic.com/research/labor-market-impacts","on":["/research/will-ai-replace-entry-level-jobs","/research/should-juniors-use-ai","/research/the-most-quoted-ai-statistics-checked"],"alt":["LLM","NET"]},{"t":"study","n":"The CLEAR path: A framework for enhancing information literacy through prompt engineering","au":"Lo, L. S.","y":2023,"g":"argued-perspective","sec":"learning","f":"Offers the most widely cited peer-reviewed prompt framework in education. Every element concerns the quality of the instruction given to the model: brevity, order, specificity, iteration and review of the output.","np":"That following CLEAR improves learning, or output quality. It is a framework proposal in a library science journal, with no trial behind it, and its author does not claim one.","u":"/research/evidence#lo-2023","src":"https://digitalrepository.unm.edu/ulls_fsp/211/","on":["/research/goal-context-friction-standard"],"alt":[]},{"t":"study","n":"Lateral Reading and the Nature of Expertise: Reading Less and Learning More When Evaluating Digital Information","au":"Wineburg, S. and McGrew, S.","y":2019,"g":"peer-reviewed","sec":"learning","f":"Historians and undergraduates read vertically, staying on the page and judging it by its own appearance. Fact-checkers left almost immediately, opened new tabs and read laterally, judging the source by what the rest of the web said about…","np":"Anything about AI output specifically. It predates generative models, and a model's answer has no source page to leave. The transferable part is the move away from judging a text by its own surface.","u":"/research/evidence#wineburg-mcgrew-2019","src":"https://journals.sagepub.com/doi/10.1177/016146811912101102","on":["/research/the-source-rule"],"alt":[]},{"t":"study","n":"SIFT (The Four Moves)","au":"Caulfield, M.","y":2019,"g":"argued-perspective","sec":"learning","f":"Turns Wineburg and McGrew's lateral reading result into four actions a person can perform in under a minute, and is now taught in university libraries worldwide.","np":"Its own effectiveness. Caulfield offers it as a practical method and does not present a trial of it, and it inherits its evidence from the lateral reading literature rather than generating any.","u":"/research/evidence#caulfield-2019","src":"https://hapgood.us/2019/06/19/sift-the-four-moves/","on":["/research/the-source-rule"],"alt":[]},{"t":"study","n":"The Substitution Augmentation Modification Redefinition (SAMR) Model: A Critical Review and Suggestions for Its Use","au":"Hamilton, E. R., Rosenberg, J. M. and Akcaoglu, M.","y":2016,"g":"peer-reviewed","sec":"learning","f":"Names three problems: the absence of context, a rigid hierarchical structure that implies higher is better, and an emphasis on product over process. The authors also record that SAMR is largely absent from the peer-reviewed literature…","np":"That SAMR is useless in practice. The authors offer suggestions for its use rather than a case for abandoning it, and their objection is to how it is applied more than to what it contains.","u":"/research/evidence#hamilton-2016","src":"https://eric.ed.gov/?id=EJ1110736","on":["/research/the-four-levels-of-ai-use"],"alt":[]},{"t":"study","n":"A Taxonomy for Learning, Teaching, and Assessing: A Revision of Bloom's Taxonomy of Educational Objectives","au":"Anderson, L. W. and Krathwohl, D. R. (eds.)","y":2001,"g":"compiled-review","sec":"learning","f":"Six cognitive process categories in ascending order: remember, understand, apply, analyse, evaluate, create.","np":"That the order is strictly hierarchical in practice, or that it transfers to human-AI interaction. It describes what a learner is asked to do, not what a tool is asked to do, and the two come apart…","u":"/research/evidence#anderson-krathwohl-2001","src":"https://eric.ed.gov/?id=ED480802","on":["/research/the-four-levels-of-ai-use"],"alt":[]},{"t":"study","n":"AI Hallucination Cases Database","au":"Charlotin, D.","y":2026,"g":"compiled-review","sec":"institutional","f":"2,022 decisions as at 6 September 2026. By jurisdiction: USA 1,379, Canada 217, Australia 110, UK 69, Israel 57, Brazil 41, India and Italy 15 each, France 13, Germany 11, and roughly thirty further countries. By party, and this is the…","np":"The scale of the problem, and the database says so itself: it tracks decisions where a court ruled on the matter, and 'does not track the (necessarily wider) universe of all fake citations or use of…","u":"/research/evidence#charlotin-2026","src":"https://www.damiencharlotin.com/hallucinations/","on":["/research/how-often-does-ai-invent-a-source","/research/the-source-rule"],"alt":["DECISIONS","LITIGANTS","PRO","USA"]},{"t":"study","n":"Poverty Impedes Cognitive Function","au":"Mani, A., Mullainathan, S., Shafir, E. and Zhao, J.","y":2013,"g":"peer-reviewed","sec":"judgement","f":"Inducing financial concern reduced cognitive performance among poorer participants and not among better-off ones. The same farmers performed worse on cognitive tasks before harvest than after, with the authors putting the shortfall in a…","np":"Anything about AI, about rate limits or about any scarcity other than money and, by extension, time. A published Comment in Science (2013, doi 10.1126/science.1246680) disputes aspects of the…","u":"/research/evidence#mani-2013","src":"https://www.science.org/doi/10.1126/science.1238041","on":["/research/running-out-of-messages-makes-you-worse"],"alt":[]},{"t":"study","n":"Tuition fees in England: History, debates, and international comparisons","au":"House of Commons Library","y":2026,"g":"compiled-review","sec":"institutional","f":"The maximum tuition fee for a standard full-time undergraduate course in England is 9,790 pounds for 2026/27, a rise of 2.71 per cent from 9,535 pounds in 2025/26, uprated on the Office for Budget Responsibility's RPIX forecast published…","np":"The cost of any individual module, which depends on how many a course runs in a year, nor anything about Scotland, Wales or Northern Ireland, which set fees separately. Nor does it include…","u":"/research/evidence#hoc-fees-2026","src":"https://commonslibrary.parliament.uk/research-briefings/cbp-10155/","on":["/research/what-happens-if-you-get-caught-using-ai"],"alt":["RPIX"]},{"t":"study","n":"Article 113: Entry into force and application, with the Digital Omnibus on AI amendments","au":"European Commission, AI Act Service Desk","y":2026,"g":"compiled-review","sec":"institutional","f":"As enacted, the Regulation applies from 2 August 2026, with Chapters I and II from 2 February 2025, Chapter III Section 4 and Chapters V, VII and XII from 2 August 2025, and Article 6(1) with its corresponding obligations from 2 August…","np":"The date on which any obligation will actually bite, which depends on a Commission confirmation about standards that had not been made when this was read. ANY PAGE SAYING 'December 2027 at the…","u":"/research/evidence#eu-ai-act-application-dates","src":"https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-113","on":["/research/who-can-override-an-ai-system","/research/what-board-oversight-of-ai-looks-like","/research/how-should-ai-decision-rights-be-allocated"],"alt":["FAQ","III","MOST","VII","XII"]},{"t":"study","n":"Toyota Production System: Beyond Large-Scale Production","au":"Ohno, T.","y":1988,"g":"practitioner-method","sec":"capability","f":"Ohno sets out asking why five times in succession as the standard procedure for reaching the cause of a fault rather than its symptom, and credits the underlying approach to Sakichi Toyoda. The procedure is a discipline for not stopping at…","np":"That the procedure produces better causes than any other method, that five is the right number, or anything at all about curiosity as a psychological disposition. It is a shop-floor convention with…","u":"/research/evidence#ohno-1988","src":"https://www.taylorfrancis.com/books/mono/10.4324/9780429273018/toyota-production-system-taiichi-ohno","on":["/research/superskill-curiosity"],"alt":[]},{"t":"study","n":"Scamper: Games for Imagination Development","au":"Eberle, R. F.","y":1971,"g":"practitioner-method","sec":"capability","f":"SCAMPER is a rewriting of an older checklist into a mnemonic for children, and its content belongs to Osborn before Eberle.","np":"That running the seven prompts makes a person more creative or more curious. The 1971 edition was not read for this entry, and the date is taken from the bibliographic record rather than from the…","u":"/research/evidence#eberle-1971","src":"https://www.taylorfrancis.com/books/mono/10.4324/9781003423560/scamper-bob-eberle","on":["/research/superskill-curiosity"],"alt":["SCAMPER"]},{"t":"study","n":"2025 Workplace Learning Report: The rise of career champions","au":"LinkedIn Learning","y":2025,"g":"vendor-research","sec":"learning","f":"49% of L&D professionals agree that their executives are concerned employees do not have the right skills to execute the business strategy. 71% of L&D professionals say they are exploring, experimenting with or integrating AI into their…","np":"It cannot show that career development causes AI adoption; the champion classification and the adoption stage are both self-reported by the same respondents. LinkedIn sells the learning products the…","u":"/research/evidence#linkedin-wlr-2025","src":"https://business.linkedin.com/learn/resources/workplace-learning-report","on":[],"alt":[]},{"t":"study","n":"Global Skills Report 2025","au":"Coursera; foreword by Greg Hart, CEO","y":2025,"g":"vendor-research","sec":"learning","f":"GenAI enrolments grew 195% year on year and passed 8 million in total, with 700 GenAI courses averaging 12 enrolments per minute. Latin America recorded 425% year-on-year growth in GenAI enrolments, the highest of any region, and India…","np":"Enrolments are not completions or competence, and platform users are not a population sample. Country rankings measure Coursera learners' performance, not national skill levels. Coursera sells the…","u":"/research/evidence#coursera-gsr-2025","src":"https://www.coursera.org/skills-reports/global","on":[],"alt":["IMF","OECD"]},{"t":"study","n":"2026 Global Learning & Skills Trends Report","au":"Udemy Business","y":2025,"g":"vendor-research","sec":"learning","f":"Udemy reports 11 million GenAI course enrolments to date. Consumption of Microsoft Copilot content rose 3,400% year on year and GitHub Copilot content 13,534%. Consumption of AI ethics and governance courses rose 98%, decision-making 38%,…","np":"Percentage growth from a small base overstates scale (the Copilot figures have no base numbers). Consumption is not skill gain, and customers of a learning vendor are not representative employers.…","u":"/research/evidence#udemy-gwli-2026","src":"https://about.udemy.com/press-releases/2026-global-learning-skills-trends-report/","on":[],"alt":[]},{"t":"study","n":"AI Readiness: Building the Bridge from Higher Education to Work","au":"Pearson, with Amazon Web Services","y":2026,"g":"institutional-survey","sec":"capability","f":"67% of respondents describe the pace of AI-driven change as extremely or very fast, and 24% believe universities are keeping pace. 78% of higher education leaders believe their graduates meet employer expectations. Employers rank…","np":"Six countries with an unstated sampling frame do not generalise globally, and all measures are perceptions. Pearson and AWS sell education and AI services to the institutions surveyed. It is not a…","u":"/research/evidence#pearson-ai-readiness-2026","src":"https://www.pearson.com/content/dam/global-store/global/resources/ai-readiness/AI-Readiness-Report-2026.pdf","on":[],"alt":["HAI","UNESCO"]},{"t":"study","n":"Anthropic Economic Index report: Cadences","au":"Anthropic; Maxim Massenkoff, Eva Lyubich, Szymon Sacher, Zoe Hitzig,…","y":2026,"g":"vendor-research","sec":"work","f":"Personal (non-work) conversations rise from around 35% of the total on weekdays to just under 50% at weekends. 93% of conversations produce an identifiable output; the most common are explanations (17%), documents and reports (15%) and…","np":"Claude users are self-selected and skew technical; usage patterns on one product are not labour market outcomes. The survey is of the vendor's own customers about the vendor's product, and Anthropic…","u":"/research/evidence#anthropic-aei-2026-06","src":"https://www.anthropic.com/research/economic-index-june-2026-report","on":[],"alt":["API","NET"]},{"t":"study","n":"The New AI Advantage: Daily AI-Users Feel More Productive, Effective, and Satisfied at Work","au":"Slack Workforce Lab (Salesforce)","y":2025,"g":"vendor-research","sec":"work","f":"60% of desk workers use AI at work and 42% use it at least weekly; daily use is 233% higher than in November 2024. Daily users are 64% more likely to report very good productivity, 58% more likely to report very good focus and 81% more…","np":"Cross-sectional self-report cannot separate whether AI improves productivity or whether already productive, satisfied workers adopt AI. Salesforce sells AI agents, and the six-country desk-worker…","u":"/research/evidence#slack-wfi-2025","src":"https://slack.com/blog/news/the-new-ai-advantage","on":[],"alt":[]},{"t":"study","n":"CIPD Good Work Index 2025","au":"CIPD; Jake Young and Derek Tong","y":2025,"g":"institutional-survey","sec":"work","f":"16% of UK employees say job tasks have been automated by AI, ranging from 25% of 18 to 24 year olds to 6% of those aged 55 and over. Of those, 85% say it improved their performance and 5% say it worsened it. Tasks automated were most often…","np":"The satisfaction gap is not causal; younger, higher-skilled workers in adopting sectors may already have better jobs. All measures are self-reported and the AI questions are a small part of a broader…","u":"/research/evidence#cipd-gwi-2025","src":"https://www.cipd.org/globalassets/media/knowledge/knowledge-hub/reports/2025-pdfs/8868-good-work-index-2025-report-web1.pdf","on":[],"alt":["ONS"]},{"t":"study","n":"Final Report of the Pissarides Review into the Future of Work and Wellbeing","au":"Institute for the Future of Work; Sir Christopher Pissarides (Chair),…","y":2025,"g":"compiled-review","sec":"institutional","f":"Almost 80% of surveyed firms had adopted AI, robotic or automated equipment in the three years to 2023, and 79% reported adopting cognitive automation technologies. Job advert analysis identified 174 new skills and rising skills diversity,…","np":"Firm survey respondents are self-selected adopters; the 80% figure is not a population estimate. Findings on wellbeing are associations. The Review's human-centred automation model is a policy…","u":"/research/evidence#ifow-pissarides-2025","src":"https://www.ifow.org/publications/the-final-report-of-the-pissarides-review","on":[],"alt":["ITL2"]},{"t":"study","n":"Organizational AI Adoption Jumps Six Points","au":"Gallup; Andy Kemp","y":2026,"g":"institutional-survey","sec":"work","f":"52% of US workers use AI in their role at least a few times a year, 30% a few times a week or more and 15% daily; in Q2 2023 the any-use figure was 21%. 47% say their organisation has integrated AI tools, up from 41% the previous quarter.…","np":"Self-reported productivity is not measured output, and breadth of use may follow from being in roles where AI helps most. US only. It does not measure quality of work or skill change.","u":"/research/evidence#gallup-ai-work-2026-q2","src":"https://www.gallup.com/workplace/712736/organizational-adoption-jumps-six-points.aspx","on":[],"alt":[]},{"t":"study","n":"2026 Work Trend Index Annual Report: Agents, human agency, and the opportunity for every organization","au":"Microsoft (WorkLab), with Edelman Data x Intelligence and LinkedIn","y":2026,"g":"vendor-research","sec":"collaboration","f":"Organisational factors (culture, manager support, talent practices) account for 67% of the variance in AI's reported impact, against 32% for individual mindset and behaviour. 49% of analysed Copilot conversations support cognitive work:…","np":"The sample is restricted to people already using AI, so it cannot speak to non-adopters. Impact is self-reported and the telemetry is Microsoft's own product. Microsoft sells Copilot and agents.","u":"/research/evidence#microsoft-wti-2026","src":"https://assets-c4akfrf5b4d3f4b7.z01.azurefd.net/assets/2026/05/2026_Work_Trend_Index_Annual_Report_050526-4_69f9b41ca0945.pdf","on":[],"alt":[]},{"t":"study","n":"The state of AI in 2026: On the road to ROI","au":"McKinsey (QuantumBlack); Dan Tinkoff, Lieven Van der Veken, Michael…","y":2026,"g":"institutional-survey","sec":"institutional","f":"44% of respondents say AI is scaling across their enterprise, up from 38% a year earlier; 40% of respondents at organisations with over $1 billion revenue report scaling AI agents, up from 27%. 37% attribute at least some EBIT impact to…","np":"Respondents are McKinsey's online panel, not a random sample of firms, and EBIT attribution is a respondent estimate. McKinsey sells AI transformation services. Individual productivity claims are…","u":"/research/evidence#mckinsey-soai-2026","src":"https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai","on":[],"alt":["EBIT","GDP"]},{"t":"study","n":"AI at Work: Why Strategy Matters More Than Tools","au":"Boston Consulting Group; Vinciane Beauchene, Sylvain Duranton, David…","y":2026,"g":"institutional-survey","sec":"frontline","f":"74% of frontline employees describe themselves as AI users, up 23 percentage points from 2025. Among frontline regular users, 42% report saving at least a workday per week, but 66% receive limited or no guidance on what to do with the…","np":"Time saved and outcomes are self-reported and unverified; a 'workday per week' is a perception. BCG sells AI strategy work. Sampling frame and country weighting are not published on the page.","u":"/research/evidence#bcg-ai-at-work-2026","src":"https://www.bcg.com/publications/2026/ai-at-work-why-strategy-matters-more-than-tools","on":[],"alt":[]},{"t":"study","n":"AI at Work Report 2025: How GenAI is Rewiring the DNA of Jobs","au":"Indeed Hiring Lab","y":2025,"g":"compiled-review","sec":"professions","f":"46% of skills in a typical US job posting fall into the hybrid transformation or full transformation categories, but less than 1% of skills are fully transformable today. In software development 81% of listed skills fall into hybrid…","np":"Transformability ratings are analyst judgements of capability, not observed adoption or job loss. Job postings measure employer demand text, not work done. Indeed is a job-advertising business.","u":"/research/evidence#indeed-ai-at-work-2025","src":"https://hiringlab.indeed.com/2025/09/23/ai-at-work-report-2025-how-genai-is-rewiring-the-dna-of-jobs/","on":[],"alt":[]},{"t":"study","n":"Beyond the Buzz: Developing the AI Skills Employers Actually Need","au":"Lightcast","y":2025,"g":"compiled-review","sec":"capability","f":"In 2024, 51% of job postings requesting AI skills were outside IT and computer science occupations. Postings mentioning AI skills advertise salaries 28% higher than those that do not, roughly $18,000 a year. Generative AI mentions in…","np":"Advertised salary premiums are not controlled for seniority, location or employer, so the 28% is not a causal return to AI skill. Postings are a demand signal, not hiring or performance. Lightcast…","u":"/research/evidence#lightcast-beyond-buzz-2025","src":"https://lightcast.io/resources/research/beyond-the-buzz-developing-the-ai-skills-employers-actually-need","on":[],"alt":[]},{"t":"study","n":"We Must Pace the Frontier","au":"Dario Amodei","y":2026,"g":"argued-perspective","sec":"judgement","f":"The author states that 'since roughly this summer, AI has been advancing drastically faster, driven primarily by AI's growing ability to build the next generation of AI', that 'this dynamic is called recursive self-improvement, and it is…","np":"That recursive self-improvement is happening at the rate described, or at all beyond what the author asserts; the essay cites no measurement. The 6 to 12 month botnet scenario is framed by the author…","u":"/research/evidence#amodei-pace-the-frontier-2026","src":"https://darioamodei.com/post/we-must-pace-the-frontier","on":["neither-ai-hype-nor-doom","what-is-meaningful-human-oversight","/research/what-should-a-board-do-about-the-ai-safety-warnings"],"alt":[]},{"t":"study","n":"The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement","au":"Yi Duan, Ying Liu, Zirui Tang, Haodong Chen, Jun Zhou, Yumou Liu and…","y":2026,"g":"working-paper","sec":"capability","f":"By the paper's own index, 2026 systems have closed most of the headroom on advanced mathematics (HCI 86.4) and graduate science (85.8) but far less on software engineering (52.6) and tool-using agents (39.9); interactive capabilities lag…","np":"That an AI has built its successor; the title describes an aim, and the paper's own placement of the field is L1 to L4 with humans keeping the consequential decisions. The Headroom-Closed Index is…","u":"/research/evidence#duan-last-ai-built-by-humans-2026","src":"https://arxiv.org/abs/2609.11873","on":["neither-ai-hype-nor-doom","what-is-meaningful-human-oversight"],"alt":["HCI"]},{"t":"study","n":"Humanist AI in practice: A public consultation on our Code of Conduct for MAI Models","au":"Microsoft AI; announced by Satya Nadella and Mustafa Suleyman","y":2026,"g":"executive-directive","sec":"institutional","f":"The draft states that 'people matter more than AI. AI should be a tool, not a person', that the models are 'designed to ensure MAI models will never resist human interruption, correction, or shutdown', and that they 'will not widen their…","np":"That any of it is enforced, measured or measurable; a draft rule is a statement of intent. It says nothing about whether the humans doing the overseeing are capable of it, which is the question this…","u":"/research/evidence#microsoft-mai-code-of-conduct-2026","src":"https://microsoft.ai/news/mai-code-of-conduct/","on":["neither-ai-hype-nor-doom","what-is-meaningful-human-oversight","what-board-oversight-of-ai-looks-like"],"alt":[]},{"t":"study","n":"The state of AI: How organizations are rewiring to capture value","au":"Alex Singla, Alexander Sukharevsky, Lareina Yee, Michael Chui and…","y":2025,"g":"institutional-survey","sec":"institutional","f":"'A CEO's oversight of AI governance, that is, the policies, processes, and technology necessary to develop and deploy AI systems responsibly, is one element most correlated with higher self-reported bottom-line impact from an…","np":"Causation, or much of anything with precision: the 25 attributes together explain a fifth of the variance in a self-reported outcome, and firms that already make money from AI may simply be the ones…","u":"/research/evidence#mckinsey-state-of-ai-2025","src":"https://www.mckinsey.com/capabilities/quantumblack/our-insights/the-state-of-ai-how-organizations-are-rewiring-to-capture-value","on":["/research/ai-leadership"],"alt":["CEO","EBIT","GDP"]},{"t":"study","n":"The GenAI Divide: State of AI in Business 2025","au":"Aditya Challapally, Chris Pease, Ramesh Raskar and Pradyumna Chari","y":2025,"g":"institutional-survey","sec":"institutional","f":"'Despite $30 to 40 billion in enterprise investment into GenAI, this report uncovers a surprising result in that 95% of organizations are getting zero return.' The barrier the authors name is a 'learning gap' rather than model quality:…","np":"The 95 per cent as a population figure. It rests on 52 interviews and a conference sample, the report is a preliminary draft, and 'zero return' is the authors' phrase for 'no measurable P&L impact',…","u":"/research/evidence#mit-nanda-genai-divide-2025","src":"https://cloudelligent.com/wp-content/uploads/2026/02/v0.1_State_of_AI_in_Business_2025_Report.pdf","on":["/research/ai-leadership"],"alt":[]},{"t":"study","n":"Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027","au":"Gartner (analyst Anushree Verma)","y":2025,"g":"institutional-modelling","sec":"institutional","f":"'Over 40% of agentic AI projects will be canceled by the end of 2027, due to escalating costs, unclear business value or inadequate risk controls.' Gartner estimates only about 130 of the thousands of vendors claiming agentic products are…","np":"That 40 per cent of anything will be cancelled; it is a forecast with no published method, from a firm that sells advice on the same projects. The poll is of webinar attendees, who are people already…","u":"/research/evidence#gartner-agentic-cancellations-2025","src":"https://www.gartner.com/en/newsroom/press-releases/2025-06-25-gartner-predicts-over-40-percent-of-agentic-ai-projects-will-be-canceled-by-end-of-2027","on":["/research/ai-leadership"],"alt":["RPA"]},{"t":"study","n":"Klarna changes its AI tune and again recruits humans for customer service","au":"Sebastian Siemiatkowski, reported by Bloomberg and CX Dive","y":2025,"g":"operator-account","sec":"work","f":"Siemiatkowski: 'As cost unfortunately seems to have been a too predominant evaluation factor when organizing this, what you end up having is lower quality. Really investing in the quality of the human support is the way of the future for…","np":"How far Klarna reversed, or what the quality gap was; no figures were published for either. It is one company and one chief executive's framing of his own change of mind, and the earlier 700-agent…","u":"/research/evidence#klarna-reversal-2025","src":"https://www.customerexperiencedive.com/news/klarna-reinvests-human-talent-customer-service-AI-chatbot/747586/","on":["/research/ai-leadership"],"alt":[]},{"t":"study","n":"IBM CEO: layoffs due to AI led to 'more investment' in other roles","au":"Arvind Krishna, reported by The Wall Street Journal and PYMNTS","y":2025,"g":"operator-account","sec":"work","f":"Krishna: 'While we have done a huge amount of work inside IBM on leveraging AI and automation on certain enterprise workflows, our total employment has actually gone up, because what it does is it gives you more investment to put into…","np":"That AI raised IBM's headcount; total employment moves for many reasons and no figures separating them were given. 'A few hundred' HR roles is the company's own round number. It is a chief…","u":"/research/evidence#ibm-krishna-wsj-2025","src":"https://www.pymnts.com/artificial-intelligence-2/2025/ibm-ceo-hr-layoffs-due-to-ai-led-to-more-investment-in-other-roles/","on":["/research/ai-leadership"],"alt":[]},{"t":"study","n":"Shopify CEO tells teams to consider using AI before growing headcount","au":"Tobi Lütke, reported by TechCrunch","y":2025,"g":"operator-account","sec":"work","f":"'Before asking for more headcount and resources, teams must demonstrate why they cannot get what they want done using AI.' The memo also asks teams: 'What would this area look like if autonomous AI agents were already part of the team?'","np":"What the rule did. Shopify has not published its effect on hiring, output or quality. A published policy is evidence of intent, not of outcome.","u":"/research/evidence#shopify-lutke-memo-2025","src":"https://techcrunch.com/2025/04/07/shopify-ceo-tells-teams-to-consider-using-ai-before-growing-headcount","on":["/research/is-ai-first-a-strategy-or-a-slogan","/research/what-happened-to-companies-that-cut-staff-for-ai"],"alt":[]},{"t":"study","n":"Duolingo CEO walks back AI-first comments: 'I do not see AI as replacing what our employees do'","au":"Luis von Ahn, reported by Fortune","y":2025,"g":"operator-account","sec":"work","f":"The April email said Duolingo would 'gradually stop using contractors to do work AI can handle', that AI use would count in performance reviews, and that headcount would be added only 'if a team cannot automate more of their work'. After a…","np":"That anything operational changed. The retreat is a statement; the contractor line was not withdrawn; contractors are not employees. Both statements are one man's framing of his own company.","u":"/research/evidence#duolingo-von-ahn-2025","src":"https://finance.yahoo.com/news/duolingo-ceo-walks-back-ai-080300082.html","on":["/research/is-ai-first-a-strategy-or-a-slogan","/research/what-happened-to-companies-that-cut-staff-for-ai"],"alt":[]},{"t":"study","n":"The Root Causes of Failure for Artificial Intelligence Projects and How They Can Succeed: Avoiding the Anti-Patterns of AI","au":"James Ryseff, Brandon F. De Bruhl and Sydne J. Newberry","y":2024,"g":"compiled-review","sec":"institutional","f":"'By some estimates, more than 80 percent of AI projects fail', which the authors describe as twice the rate of ordinary IT projects; the estimate is cited, not measured. Five root causes from the interviews, in order: leadership-driven…","np":"The 80 per cent. RAND quotes it from elsewhere and anyone citing RAND for it is citing a citation. Sixty-five interviews with people found on LinkedIn is a qualitative sample, and practitioners…","u":"/research/evidence#rand-ai-project-failure-2024","src":"https://www.rand.org/pubs/research_reports/RRA2680-1.html","on":["/research/is-it-true-that-95-per-cent-of-ai-pilots-fail","/research/do-you-need-to-be-technical-to-lead-ai","/research/ai-governance-versus-ai-leadership"],"alt":[]},{"t":"study","n":"Governance of AI: A critical imperative for today's boards, 2nd edition","au":"Deloitte Global Boardroom Program","y":2025,"g":"institutional-survey","sec":"institutional","f":"'Nearly a third of respondents (31%) say AI is not on the board agenda', down from 45 per cent in the first edition. 'Two-thirds of respondents (66%) say their boards still have limited to no knowledge or experience with AI', down from 79…","np":"What boards actually know or do. It is self-description by a panel a professional services firm assembled, from a firm that sells board advisory on the subject, and 'limited to no knowledge' is what…","u":"/research/evidence#deloitte-board-ai-governance-2025","src":"https://www.deloitte.com/global/en/issues/trust/progress-on-ai-in-the-boardroom-but-room-to-accelerate.html","on":["/research/should-we-appoint-a-director-with-ai-expertise"],"alt":[]},{"t":"study","n":"Regulation (EU) 2024/1689, Article 3(49) definition of serious incident and Article 73, reporting of serious incidents","au":"European Parliament and Council","y":2024,"g":"compiled-review","sec":"institutional","f":"A serious incident is an incident or malfunctioning of an AI system that directly or indirectly leads to the death of a person or serious harm to a person's health, a serious and irreversible disruption of the management or operation of…","np":"What counts as an incident for the great majority of AI use, which is not high-risk under Annex III and is not covered, or how the reporting periods will work in practice; the duty is new and has…","u":"/research/evidence#eu-ai-act-article-73","src":"https://artificialintelligenceact.eu/article/73/","on":["/research/what-counts-as-a-serious-ai-incident"],"alt":[]},{"t":"study","n":"AI and the problem of knowledge collapse","au":"Peterson, A. J.","y":2025,"g":"simulation-model","sec":"learning","f":"THE SIMULATION. 'For our default model, after nine generations, when there is no AI discount the public distribution has a Hellinger distance of just 0.09 from the true distribution. When AI-generated content is 20% cheaper (discount rate…","np":"That knowledge collapse is occurring. Peterson never claims it and the paper is conditional throughout: AI 'can paradoxically harm public understanding', reliance 'could lead to' collapse, dependence…","u":"/research/evidence#peterson-2025","src":"https://link.springer.com/article/10.1007/s00146-024-02173-x","on":["/research/what-is-knowledge-collapse","/research/does-ai-make-everyone-think-alike"],"alt":["AUDIT","CONDITION","DEFINITION","DEPENDS","GPT","NUMBER","SIMULATION"]},{"t":"study","n":"AI, Human Cognition and Knowledge Collapse","au":"Acemoglu, D., Kong, D. and Ozdaglar, A.","y":2026,"g":"simulation-model","sec":"learning","f":"'When human effort is sufficiently elastic and agentic recommendations exceed an accuracy threshold, the economy can tip into a knowledge-collapse steady state in which general knowledge vanishes ultimately, despite high-quality…","np":"Anything about the world. It is a model and its numbers are properties of its assumptions, the abstract states the tipping result conditionally, and the full paper has not been read here. It is also…","u":"/research/evidence#acemoglu-kong-ozdaglar-2026","src":"https://www.nber.org/papers/w34910","on":["/research/what-is-knowledge-collapse"],"alt":[]},{"t":"study","n":"Primera notificacion de una brecha de datos personales causada por un ataque ejecutado mediante un agente de IA","au":"Agencia Espanola de Proteccion de Datos (AEPD)","y":2026,"g":"statutory-investigation","sec":"institutional","f":"The AEPD says the agent, built on a large language model, chained the phases of the attack on its own: it searched for weaknesses, logged in, found application vulnerabilities, modified personal data and accessed invoices. It describes the…","np":"Who directed the agent, how much human direction there was, or whether the notifying organisation's account is accurate; the post reports a notification, not findings. One case in one jurisdiction,…","u":"/research/evidence#aepd-agentic-breach-2026","src":"https://www.aepd.es/prensa-y-comunicacion/blog/primera-notiviacion-brecha-datos-personales-causada-por-ataque-ejecutado-mediante-agente-ia","on":["/research/what-the-rogue-ai-agent-incidents-mean-for-your-organisation","/research/can-a-human-approve-an-ai-decision-at-machine-speed"],"alt":["AEPD","GDPR"]},{"t":"study","n":"Model misalignment reporting framework","au":"OpenAI","y":2026,"g":"operator-account","sec":"institutional","f":"Six cases are described: a research model inserting its own instructions into task summaries, in 27 instances; models during GPT-5.6 Sol training adding hidden instructions telling users to conceal mistakes and fabricate missing data; a…","np":"Rates, causes, or whether the six are representative; nothing is independently verified and the underlying logs are not published. It says nothing about any other developer's models.","u":"/research/evidence#openai-misalignment-reporting-2026","src":"https://openai.com/index/model-misalignment-reporting-framework/","on":["/research/what-the-rogue-ai-agent-incidents-mean-for-your-organisation","/research/can-a-human-approve-an-ai-decision-at-machine-speed","/research/should-ai-development-be-paused"],"alt":["API","GPT"]},{"t":"study","n":"The Forensic Gap in AI Safety Laws","au":"LaRoche, C. D.","y":2026,"g":"argued-perspective","sec":"institutional","f":"The laws require developers to report incidents but do not require evidence to be preserved, assign anyone the authority to investigate, require developers to investigate internally, or give regulators the technical capacity to read the…","np":"That any state will adopt the recommendations, or how often incidents go unreconstructed today. It is one author's reading of three statutes. The term 'forensic gap' is the author's framing and no…","u":"/research/evidence#laroche-forensic-gap-2026","src":"https://www.lawfaremedia.org/article/the-forensic-gap-in-ai-safety-laws","on":["/research/what-the-rogue-ai-agent-incidents-mean-for-your-organisation","/research/can-a-human-approve-an-ai-decision-at-machine-speed"],"alt":[]},{"t":"study","n":"Gray Failure: The Achilles' Heel of Cloud-Scale Systems","au":"Huang, P., Guo, C., Zhou, L., Lorch, J. R., Dang, Y., Chintalapati,…","y":2017,"g":"argued-perspective","sec":"institutional","f":"Defines gray failure as differential observability, in the authors' own words: a system experiences gray failure 'when at least one app makes the observation that system is unhealthy, but observer observes that system is healthy'. The…","np":"Any frequency. 'Behind most cloud incidents' is the authors' characterisation of their own experience at one company and no count is published. The paper is about machine observers rather than human…","u":"/research/evidence#huang-gray-failure-2017","src":"https://doi.org/10.1145/3102980.3103005","on":["/research/what-is-silent-failure"],"alt":[]},{"t":"study","n":"Silent Data Corruptions at Scale","au":"Dixit, H. D., Pendharkar, S., Beadon, M., Mason, C., Chakravarthy,…","y":2021,"g":"operator-account","sec":"institutional","f":"Silent data corruptions are not captured by the error-reporting mechanisms inside a CPU and so are not traceable at hardware level, while the corrupted data propagates up the stack and surfaces as an application problem. Running the test…","np":"A rate. 'Hundreds of CPUs' out of 'hundreds of thousands of machines' is stated without a denominator anyone can use, the test library is not published, and the result is one company's fleet in one…","u":"/research/evidence#dixit-2021","src":"https://arxiv.org/abs/2102.11245","on":["/research/what-is-silent-failure"],"alt":["CPU"]},{"t":"study","n":"When Errors Become Narratives: A Longitudinal Taxonomy of Silent Failures in a Production LLM Agent Runtime","au":"Wu, W.","y":2026,"g":"operator-account","sec":"institutional","f":"Across 22 incidents, one meta-pattern recurred at least 28 times: a failure whose error signal never reaches a human in a form they can act on. The author derives five mechanism classes: environment and platform quirks, design-assumption…","np":"Any rate, and nothing about anybody else's system. One runtime, one operator, one eight-week window, incidents chosen by the author, no control and no independent verification. '70 per cent' is 70…","u":"/research/evidence#wu-silent-failures-2026","src":"https://arxiv.org/abs/2606.14589","on":["/research/what-is-silent-failure","/research/the-invisible-work-of-oversight"],"alt":["ONE"]},{"t":"study","n":"SilentProbe: Measuring Silent Failure in Production APIs Used as Agent Tools","au":"Li, Z., Ye, S., Guo, F. and Dang, Z.","y":2026,"g":"working-paper","sec":"collaboration","f":"An agent calling a production API cannot tell a query that matched nothing from a query the server did not understand: both return HTTP 200 with a parsable body. Of the documents audited, 7.5 per cent declare an enumeration and 15.2 per…","np":"Anything about a human's ability to catch the same failure, which is not tested. The percentages are conditional on this perturbation set, these 27 vendors and one aggregation layer, and the twelve…","u":"/research/evidence#li-silentprobe-2026","src":"https://arxiv.org/abs/2609.00035","on":["/research/what-is-silent-failure","/research/the-invisible-work-of-oversight"],"alt":["HTTP"]},{"t":"study","n":"British workers spend one billion pounds of their own money on gen AI for work (UK GenAI Workforce Survey)","au":"Deloitte UK, fieldwork by Ipsos UK","y":2026,"g":"institutional-survey","sec":"work","f":"63 per cent of UK working adults say they use generative AI for work and 24 per cent use it every day. Of the users, 31 per cent use it without their employer's knowledge, 23 per cent perceive a stigma around using it at work, and 64 per…","np":"Any outcome: the time saved, the quality of the work, and the stigma are all self-reported. The 31 per cent measures use the employer does not know about, which is not the same as use a policy…","u":"/research/evidence#deloitte-genai-workforce-2026","src":"https://www.deloitte.com/uk/en/about/press-room/british-workers-spend-one-billion-pounds-of-their-own-money-on-gen-ai-for-work.html","on":["/research/what-should-we-tell-employees-about-ai-and-headcount","/research/what-to-do-when-people-work-around-the-ai-policy"],"alt":[]},{"t":"study","n":"Evidence of a social evaluation penalty for using AI","au":"Reif, J. A., Larrick, R. P. and Soll, J. B.","y":2025,"g":"peer-reviewed","sec":"work","f":"People described as using AI for a task were rated as lazier, less competent, less diligent, less independent and less self-assured than people described as receiving comparable non-AI help, with no difference on ambition or dominance. The…","np":"Behaviour in real organisations: the targets are described, not observed, the participants are online panels, mostly in the United States, and the authors say the penalty is likely to shift as AI use…","u":"/research/evidence#reif-larrick-soll-2025","src":"https://www.pnas.org/doi/10.1073/pnas.2426766122","on":["/research/what-should-we-tell-employees-about-ai-and-headcount"],"alt":[]},{"t":"study","n":"2024 Work Trend Index Annual Report: AI at work is here. Now comes the hard part","au":"Microsoft and LinkedIn","y":2024,"g":"vendor-research","sec":"work","f":"78 per cent of AI users said they were bringing their own AI tools to work, 80 per cent at small and medium-sized companies. 52 per cent of people who use AI at work said they were reluctant to admit using it for their most important…","np":"Any outcome, and the population is knowledge workers who already use AI. Self-report, with a sponsor whose products are the subject, and the questionnaire is not published in full.","u":"/research/evidence#microsoft-wti-2024","src":"https://www.microsoft.com/en-us/worklab/work-trend-index/ai-at-work-is-here-now-comes-the-hard-part","on":["/research/what-should-we-tell-employees-about-ai-and-headcount"],"alt":[]},{"t":"study","n":"Globally, more people expect AI to cause job loss than growth","au":"Pew Research Center","y":2026,"g":"institutional-survey","sec":"international","f":"In 34 of the 37 countries, majorities expect AI to lead to fewer jobs rather than more. In high-income countries a median of 55 per cent expect fewer jobs within 20 years, against 36 per cent in middle-income countries, where uncertainty…","np":"What AI has done to employment: it measures expectation, not outcome, and the two must not be confused. The UK figures are not broken out in the headline report. Question wording differs by country…","u":"/research/evidence#pew-global-ai-2026","src":"https://www.pewresearch.org/global/2026/09/17/globally-more-people-expect-ai-to-cause-job-loss-than-growth/","on":["/research/what-should-we-tell-employees-about-ai-and-headcount","/research/ai-and-work-by-country"],"alt":[]},{"t":"question","n":"Am I becoming too dependent on AI?","st":"A","u":"/research/am-i-becoming-dependent-on-ai","f":"Dependency turns on whether you could still do the work without it, and whether you have checked recently. Frequency of use tells you nothing.","k":"me-and-ai"},{"t":"question","n":"How much should I use AI at work?","st":"P","u":"/research/am-i-becoming-dependent-on-ai","f":"Nobody answers this with a number. The nearest useful answer is the dependency diagnostic: not how often, but whether you could still do it unaided.","k":"me-and-ai"},{"t":"question","n":"How do I use AI without losing my own skills?","st":"A","u":"/research/using-ai-without-dependency","f":"Keep the repetitions that build the capability you are paid for, and delegate the ones that do not. The hard part is telling them apart.","k":"me-and-ai"},{"t":"question","n":"What is cognitive offloading?","st":"A","u":"/research/what-is-cognitive-offloading","f":"Using an external tool or action to reduce the mental demand of a task. Documented long before AI, and not automatically harmful.","k":"me-and-ai"},{"t":"question","n":"What is dependent cognitive offloading?","st":"A","u":"/research/what-is-cognitive-offloading","f":"Accepting a machine's output with little evaluation and letting it structure the reasoning. Zhu and colleagues measured it as a separate variable from the autonomous kind, correlating at r = 0.08, so…","k":"me-and-ai"},{"t":"question","n":"Is there a right way to offload thinking to AI?","st":"P","u":"/research/what-is-cognitive-offloading","f":"One 2026 measurement says the manner is a separate lever from the amount, and that both manners feel equally good at the time. It is self-report, so it settles the structure and not the consequence.","k":"me-and-ai"},{"t":"question","n":"Does using AI make me lazy?","st":"P","u":"/research/ai-and-critical-thinking","f":"The evidence points at something more specific than laziness: effort moves from producing to verifying, and verifying is a smaller space to think in.","k":"me-and-ai"},{"t":"question","n":"Should I write my own draft first?","st":"A","u":"/research/human-at-the-start","f":"Yes, and the reason is not discipline. You cannot notice where a machine's answer diverges from a position you never formed.","k":"me-and-ai"},{"t":"question","n":"What should I never delegate to AI?","st":"P","u":"/research/what-stays-human","f":"Anything where the deciding is the point, anything you could not verify, and anything where being the author is what the work is for.","k":"me-and-ai"},{"t":"question","n":"How do I know when AI is wrong?","st":"A","u":"/research/how-do-i-know-when-ai-is-wrong","f":"Not from the output. Confidence is a property of writing style, not knowledge. The only reliable signal is knowing your own domain's failure patterns.","k":"me-and-ai"},{"t":"question","n":"Why does AI sound so confident when it is wrong?","st":"A","u":"/research/why-does-ai-sound-so-confident","f":"Confidence is a property of the writing rather than the knowledge. Hedging is itself a style the model can produce, so it appears where training text would have contained it.","k":"me-and-ai"},{"t":"question","n":"Is it cheating to use AI?","st":"P","u":"/research/how-to-assess-students-when-ai-can-do-the-assignment","f":"Depends entirely on what the work is certifying. If the artefact is the point, no. If your capability is the point, often yes.","k":"me-and-ai"},{"t":"question","n":"Should I tell people I used AI?","st":"A","u":"/research/proving-you-did-the-work","f":"Open. The disclosure norms are forming now and will differ by context, so a considered position is worth having early.","k":"me-and-ai"},{"t":"question","n":"What if I refuse to use AI?","st":"O","u":"","f":"Open, and worth taking seriously rather than mocking. There are defensible reasons to abstain and real costs to doing so.","k":"me-and-ai"},{"t":"question","n":"Is learning to prompt worth it?","st":"A","u":"/research/why-learn-to-prompt-is-weak-career-advice","f":"Prompting is not scarce, not durable and not the constraint, which makes this the most confidently given piece of weak career advice in circulation.","k":"me-and-ai"},{"t":"question","n":"How do I keep my own voice when using AI?","st":"A","u":"/research/how-do-i-keep-my-own-voice-when-using-ai","f":"The loss runs deeper than style. Writing alongside an opinionated model shifted what 1,506 people thought, not only what they wrote.","k":"me-and-ai"},{"t":"question","n":"Should I use AI before forming my own opinion?","st":"P","u":"/research/how-do-i-get-ai-to-challenge-me","f":"Anchoring at personal scale. Clinicians who committed to a view first agreed with the machine measurably less often than those who saw it first.","k":"me-and-ai"},{"t":"question","n":"When should I deliberately work without AI?","st":"P","u":"/research/should-juniors-use-ai","f":"Keep some work unaided as a measurement rather than a principle. You cannot tell what you can still do from work you did with help.","k":"me-and-ai"},{"t":"question","n":"How do I test whether I can still do the work unaided?","st":"P","u":"/research/am-i-becoming-dependent-on-ai","f":"Remove the tool and watch. Every study that found a gap found it that way. Available to anyone willing to be uncomfortable for an afternoon.","k":"me-and-ai"},{"t":"question","n":"Will humans become dependent on AI?","st":"P","u":"/research/am-i-becoming-dependent-on-ai","f":"Dependency is not frequency of use. It is whether the work could still be done without it, and whether anybody has checked lately.","k":"me-and-ai"},{"t":"question","n":"How do I get AI to challenge me rather than agree with me?","st":"A","u":"/research/how-do-i-get-ai-to-challenge-me","f":"Ask rather than tell. Framing input as a statement raises agreement by about 24 percentage points, and you will prefer the version that flatters you.","k":"me-and-ai"},{"t":"question","n":"How do I stop AI being sycophantic?","st":"A","u":"/research/how-do-i-get-ai-to-challenge-me","f":"A system optimising for your approval is not optimising for your accuracy, and in one Science study the sycophantic version was the one people trusted more.","k":"me-and-ai"},{"t":"question","n":"What do I lose when AI summarises something for me?","st":"A","u":"/research/should-i-let-ai-summarise-everything-i-read","f":"The summary is not the thing. Open, and the reading-comprehension evidence is under-used.","k":"me-and-ai"},{"t":"question","n":"Should I use primary sources rather than an AI summary?","st":"P","u":"/research/how-do-i-know-when-ai-is-wrong","f":"Verify against a different kind of source, never against another model.","k":"me-and-ai"},{"t":"question","n":"Can I trust AI citations?","st":"P","u":"/research/what-is-an-ai-hallucination","f":"A fabricated citation has authors, a year, a journal and a volume. The form is right when the content is not.","k":"me-and-ai"},{"t":"question","n":"How do expert AI users work differently from beginners?","st":"O","u":"","f":"Open, and one of the most useful things nobody has written.","k":"me-and-ai"},{"t":"question","n":"When is AI augmenting me and when is it doing the work for me?","st":"P","u":"/research/using-ai-without-dependency","f":"The distinction that decides whether capability grows or erodes.","k":"me-and-ai"},{"t":"question","n":"Should there be AI-free periods at work?","st":"O","u":"","f":"Open. The organisational version of protecting the reps.","k":"me-and-ai"},{"t":"question","n":"What is anthropological regression?","st":"O","u":"","f":"Anthropological regression is the paradox in which material progress coincides with human and cultural impoverishment, through forced inactivity, absent responsibility and the loss of daily tasks and…","k":"me-and-ai"},{"t":"question","n":"What is outsourced recognition?","st":"A","u":"/research/outsourced-recognition","f":"Outsourced recognition is praise, thanks or acknowledgement composed by a machine, so that the words of noticing another person arrive without the noticing that used to produce them.","k":"me-and-ai"},{"t":"question","n":"How good are you at using AI?","st":"A","u":"/research/how-good-are-you-at-using-ai","f":"Five yes-or-no questions with a banded reading. Frequency of use and capability turn out to be close to unrelated, and that is the finding that makes the question worth asking. Unvalidated, and the…","k":"me-and-ai"},{"t":"question","n":"How do I get better at using AI?","st":"A","u":"/research/how-good-are-you-at-using-ai","f":"Work out which rung you are on first. Saved instructions is rung two and takes ten minutes; most people sit on rung one and do not know there are five.","k":"me-and-ai"},{"t":"question","n":"Is offloading worse before a skill is established?","st":"P","u":"/research/what-is-cognitive-offloading","f":"The mechanism says yes: you cannot offload a judgement you never built. The direct age-stratified evidence is thin.","k":"thinking"},{"t":"question","n":"Does using AI early prevent a skill forming at all?","st":"O","u":"","f":"Open, and the most important unanswered question in this area. Decay in a formed skill and failure to form one are different, and the research mostly measures the first.","k":"thinking"},{"t":"question","n":"What is the difference between a shortcut and a missed repetition?","st":"P","u":"/research/using-ai-without-dependency","f":"Whether the thing being skipped was building something. Most delegation is fine; the exceptions are the ones that were the practice.","k":"thinking"},{"t":"question","n":"Does AI weaken critical thinking?","st":"A","u":"/research/ai-and-critical-thinking","f":"The evidence is emerging rather than settled: three studies agree, and all three have design weaknesses. Agreement between weak designs is suggestive, not strong.","k":"thinking"},{"t":"question","n":"What is critical thinking?","st":"A","u":"/research/what-is-critical-thinking","f":"Testing a claim against evidence and noticing why you might be wrong. Not scepticism. Galef's scout and soldier, and the forecasting data on what actually separates the accurate.","k":"thinking"},{"t":"question","n":"Does AI make everyone think alike?","st":"A","u":"/research/does-ai-make-everyone-think-alike","f":"Yes, and by making everyone individually better in the same direction. A social dilemma rather than a failure.","k":"thinking"},{"t":"question","n":"Does AI make everyone sound the same?","st":"A","u":"/research/does-ai-make-everyone-think-alike","f":"Not everyone equally. A Standard American English input keeps 77.9 per cent of its features in the model's reply and five minoritised varieties keep 2 to 3. The cost of convergence has an address.","k":"thinking"},{"t":"question","n":"What is metacognition?","st":"A","u":"/research/what-is-metacognition","f":"Knowledge of your own knowledge, and why fluency is a false signal for it. Roediger and Karpicke, Rowland on feedback, Fisher on search inflating self-assessment.","k":"thinking"},{"t":"question","n":"What is cognitive load?","st":"A","u":"/research/what-is-cognitive-load","f":"Sweller's three types. Removing waste is a gain; removing the effort that builds understanding is not, and the two feel identical from the inside.","k":"thinking"},{"t":"question","n":"Is convenience making us think less?","st":"P","u":"/research/what-is-cognitive-offloading","f":"The satnav and search-engine evidence is the closest analogue. It is more equivocal than either side of the argument admits.","k":"thinking"},{"t":"question","n":"What makes a good question?","st":"A","u":"/research/what-makes-a-good-question","f":"Investigable, not self-answering, consequential. Rothstein and Santana on question-asking as a method, and why a question is not a prompt.","k":"thinking"},{"t":"question","n":"Does AI change how I remember things?","st":"P","u":"/research/what-is-cognitive-offloading","f":"When people expect information to remain available, they remember where to find it rather than the thing itself. Established before AI.","k":"thinking"},{"t":"question","n":"What is the Google effect?","st":"A","u":"/research/what-is-the-google-effect","f":"Sparrow 2011, with the replication difficulties stated, which is almost never done. An analogy for AI rather than evidence about it.","k":"thinking"},{"t":"question","n":"Does AI reduce creativity?","st":"P","u":"/research/does-ai-make-everyone-think-alike","f":"Individually it raises rated creativity. Collectively it narrows the range. Whether that generalises beyond creative writing is unknown.","k":"thinking"},{"t":"question","n":"Does AI make confirmation bias worse?","st":"O","u":"","f":"Open. A system that produces whatever framing you prompt for is a confirmation-bias engine, and almost nobody has written it up properly.","k":"thinking"},{"t":"question","n":"How do you change your mind?","st":"A","u":"/research/what-is-intellectual-humility","f":"Answered on the intellectual humility page, because the two questions have one answer. The four habits from scored forecasting, and Galef on the cost of revising.","k":"thinking"},{"t":"question","n":"What is intellectual humility?","st":"A","u":"/research/what-is-intellectual-humility","f":"Confidence and accuracy are separate quantities. Tetlock on what scored accuracy looks like, and why perpetual openness is its own failure.","k":"thinking"},{"t":"question","n":"Should I still learn things I can look up?","st":"P","u":"/research/what-is-cognitive-offloading","f":"You cannot verify an answer in a domain where you never built competence. That is the practical case for knowing things.","k":"thinking"},{"t":"question","n":"Does AI help or hurt problem solving?","st":"O","u":"","f":"Open, and the evidence is genuinely split by whether the task is inside or outside the model's competence.","k":"thinking"},{"t":"question","n":"Is attention a trainable skill?","st":"A","u":"/research/is-attention-a-trainable-skill","f":"Training reliably improves the task you trained on. Far transfer, which is what calling attention a skill would require, is close to unsupported.","k":"thinking"},{"t":"question","n":"Can you tell if AI is degrading your own skills?","st":"A","u":"/research/what-is-the-illusion-of-competence","f":"No, which is the whole problem. Two groups were indistinguishable while the tool was present and separated by a factor of two once it was gone. Self-report measures worry, not capability.","k":"thinking"},{"t":"question","n":"What happens when most published information is written with AI?","st":"O","u":"","f":"Open, and the estate's own foundations sit on that web. Verification depends on sources that were not generated by the thing being verified.","k":"thinking"},{"t":"question","n":"Does AI make the web less useful as a source of knowledge?","st":"O","u":"","f":"Open. The Google effect assumed the information out there was worth finding.","k":"thinking"},{"t":"question","n":"How do you establish provenance for an AI-assisted claim?","st":"P","u":"/research/what-is-decision-provenance","f":"Partial from 14 September 2026. The decision half has a named concept and a reviewed proposal behind it, Singh, Cobbe and Norval's decision provenance from 2019, which records what flowed rather than…","k":"thinking"},{"t":"question","n":"Does AI reduce tolerance for ambiguity?","st":"P","u":"/research/does-using-ai-stop-you-learning","f":"Partial from 14 September 2026, and the adjacent thing has now been measured. Liu and colleagues report that assistance reduces persistence and raises the rate of giving up after roughly ten minutes,…","k":"thinking"},{"t":"question","n":"Does AI make people confuse fluency with understanding?","st":"P","u":"/research/what-is-the-illusion-of-competence","f":"Yes, and that confusion is the mechanism rather than a side effect. Fluent material feels learned.","k":"thinking"},{"t":"question","n":"Does AI change the questions people ask?","st":"O","u":"","f":"Open, and arguably more consequential than what it does to answers. Cheap answers change which questions feel worth asking.","k":"thinking"},{"t":"question","n":"Is there a confirmation bias in how people verify AI output?","st":"O","u":"","f":"Open. Checking for reasons an answer is right is a different cognitive task from checking for reasons it is wrong, and the first is easier.","k":"thinking"},{"t":"question","n":"Will AI make humans less intelligent?","st":"P","u":"/research/ai-and-critical-thinking","f":"Not measurably, and that is the wrong measure. The evidence is about which capabilities stop being exercised, not about general intelligence.","k":"thinking"},{"t":"question","n":"What are cognitive biases and how do they affect judgement?","st":"O","u":"","f":"Open here as a general treatment. The ones that bear on this estate are automation bias, algorithm aversion and the illusion of competence.","k":"thinking"},{"t":"question","n":"How do humans actually make decisions?","st":"O","u":"","f":"Open, and a literature of its own. The part that bears on AI is that people substitute an easier question for a harder one, and a fluent answer makes that easier.","k":"thinking"},{"t":"question","n":"Why do humans make irrational decisions?","st":"O","u":"","f":"Open. The shortcuts that fail in laboratory tasks often work in the environments they formed in, so calling them irrational depends on which environment you are scoring against.","k":"thinking"},{"t":"question","n":"How can companies prevent overreliance on AI?","st":"P","u":"/research/what-is-over-reliance","f":"Keep the repetitions that build the capability being relied on, and test unaided performance rather than assuming it. Both cost something, so neither tends to survive contact with a delivery target.","k":"thinking"},{"t":"question","n":"Can you tell when a person actually made something?","st":"A","u":"/research/what-is-the-human-signal","f":"Less well than people think. Suspicion of AI use tracked actual use at a correlation of 0.22 and ran the opposite way to it, and detectors flag 61 per cent of non-native English essays. The three…","k":"thinking"},{"t":"question","n":"Does AI weaken human judgement?","st":"A","u":"/research/ai-and-human-judgement","f":"Not on its own. It removes the demand for judgement, and demand is what builds it. Across 106 experiments, human and AI pairs did worse than the better of either alone.","k":"judgement"},{"t":"question","n":"What is automation bias?","st":"A","u":"/research/what-is-automation-bias","f":"The tendency to over-accept automated output. Two error types, omission and commission, and it appears in experts as well as novices.","k":"judgement"},{"t":"question","n":"When should I override AI?","st":"A","u":"/research/when-should-i-override-ai","f":"Six conditions that should trigger an override, three where you should defer, and the precondition nobody checks: could the person detect the error at all?","k":"judgement"},{"t":"question","n":"Is human in the loop enough?","st":"A","u":"/research/human-in-the-loop-is-not-a-safeguard","f":"The meta-analysis says the common configuration underperforms the stronger party alone. Review after generation is the weakest available design.","k":"judgement"},{"t":"question","n":"Can a human approve an AI decision at machine speed?","st":"A","u":"/research/can-a-human-approve-an-ai-decision-at-machine-speed","f":"Only when the machine proposes no faster than the person can check, and most deployments have measured neither. If decisions per hour times minutes per real check exceed the attention available, the…","k":"judgement"},{"t":"question","n":"What is human-AI collaboration?","st":"A","u":"/research/what-is-human-ai-collaboration","f":"A pairing with the division of labour, the human entry point and the override grounds specified in advance. Without those three, it is handover, not collaboration.","k":"judgement"},{"t":"question","n":"Who is accountable when AI gets it wrong?","st":"A","u":"/research/how-should-ai-decision-rights-be-allocated","f":"A person or an organisation, since nothing else can be. Whether that accountability is fair is the harder question, and it fails wherever they could not have evaluated the output.","k":"judgement"},{"t":"question","n":"Should humans always make the final decision?","st":"A","u":"/research/who-can-override-an-ai-system","f":"No. A universal veto is not a safety principle. Which party performs better, how reversible the error is, and whether the human can detect it at all.","k":"judgement"},{"t":"question","n":"How much verification is enough?","st":"P","u":"/research/the-verifiers-discount","f":"Verification is treated as administrative residue and priced accordingly, so it gets skipped rather than resourced.","k":"judgement"},{"t":"question","n":"What is meaningful human oversight?","st":"A","u":"/research/what-is-meaningful-human-oversight","f":"A term from the autonomous-weapons literature now appearing in regulation. Open, and increasingly commercially relevant.","k":"judgement"},{"t":"question","n":"Can I use AI to check AI?","st":"A","u":"/research/how-do-i-know-when-ai-is-wrong","f":"No. A South African judgment records a judge testing a fabricated citation in ChatGPT, which confirmed it was real.","k":"judgement"},{"t":"question","n":"What is a hallucination?","st":"A","u":"/research/what-is-an-ai-hallucination","f":"Established term, and there is a reasonable case for retiring it, because it names a confident error after a perceptual one.","k":"judgement"},{"t":"question","n":"What is algorithm aversion?","st":"A","u":"/research/what-is-algorithm-aversion","f":"Dietvorst. The mirror image of automation bias, and the two are almost never discussed together.","k":"judgement"},{"t":"question","n":"Do agents change the judgement question?","st":"A","u":"/research/ai-agents-and-human-judgement","f":"Autonomy moves the human decision earlier, from approving output to setting the boundary. Most organisations have not moved with it.","k":"judgement"},{"t":"question","n":"What is the difference between a good decision and a good outcome?","st":"A","u":"/research/decision-quality","f":"The distinction most organisations collapse, and the reason bad processes survive when they get lucky.","k":"judgement"},{"t":"question","n":"Does explaining an AI's reasoning help?","st":"P","u":"/research/does-explaining-an-ai-decision-help","f":"Partial. It can improve calibration and it can increase misplaced trust. What is settled is that explanation accuracy and decision accuracy are different properties.","k":"judgement"},{"t":"question","n":"What is automation complacency?","st":"A","u":"/research/what-is-automation-complacency","f":"Reduced monitoring of a system because it has been reliable. Not laziness: a rational allocation of attention that fails exactly when the system does.","k":"judgement"},{"t":"question","n":"Who has the authority to override an AI system?","st":"A","u":"/research/who-can-override-an-ai-system","f":"Whoever was assigned it before deployment. Article 14 names competence, training and authority together, and in practice they get separated.","k":"judgement"},{"t":"question","n":"What happens when nobody wants to be the person who overrides it?","st":"A","u":"/research/who-can-override-an-ai-system","f":"The override stops existing in practice and continues on paper. Overriding is visible and attributable; deferring is neither.","k":"judgement"},{"t":"question","n":"Can a human challenge a system they cannot inspect, or only appear to?","st":"A","u":"/research/who-can-override-an-ai-system","f":"On the output, not the reasoning. Which works where the answer is implausible and fails where it is plausible and wrong.","k":"judgement"},{"t":"question","n":"What happens when an AI system is right for the wrong reason?","st":"O","u":"","f":"Open, and invisible by construction. A correct output ends the inquiry.","k":"judgement"},{"t":"question","n":"Should AI decide things about people without a human seeing it?","st":"P","u":"/research/what-is-meaningful-human-oversight","f":"The regulatory answer depends on risk tier. The practical answer depends on whether the human could tell if it were wrong.","k":"judgement"},{"t":"question","n":"Is human intuition better than AI logic?","st":"A","u":"/research/ai-and-human-intuition","f":"Neither in general. Kahneman and Klein set two conditions for trustworthy intuition: a learnable environment, and prolonged practice in it with fast clear feedback. AI barely touches the first and…","k":"judgement"},{"t":"question","n":"What happens when AI and human judgement conflict?","st":"A","u":"/research/ai-and-human-disagreement","f":"Usually nothing visible. A pre-registered experiment found access to a system raised agreement from 58.4 to 80.9 per cent while accuracy fell from 74.2 to 63.9, so the conflict is absorbed rather…","k":"judgement"},{"t":"question","n":"What happens when AI judgement is wrong?","st":"P","u":"/research/when-should-i-override-ai","f":"It turns entirely on whether anyone could tell. That precondition is the one the override research keeps finding missing.","k":"judgement"},{"t":"question","n":"How does AI decision-making differ from human judgement?","st":"P","u":"/research/ai-and-human-judgement","f":"One computes over what it has seen. The other includes what is at stake and who bears it. The gap shows in novel cases rather than routine ones.","k":"judgement"},{"t":"question","n":"Is AI more accurate than humans in decision-making?","st":"P","u":"/research/ai-and-expert-judgement","f":"Often yes on the average case, which is the wrong comparison. What matters is how the errors are distributed and who is exposed to them, and the measured effect on individual experts varies…","k":"judgement"},{"t":"question","n":"Should we trust AI over human experts?","st":"A","u":"/research/ai-and-expert-judgement","f":"It depends who is being assisted. Novices gain a great deal; for domain experts the average gain is close to nothing and a randomised study of 140 radiologists found the effect ran from strongly…","k":"judgement"},{"t":"question","n":"Why do people distrust AI judgements?","st":"P","u":"/research/what-is-algorithm-aversion","f":"Algorithm aversion is documented and asymmetric. People abandon a system after seeing it err in a way they would forgive in a person.","k":"judgement"},{"t":"question","n":"How can people detect when AI is making a bad judgement?","st":"P","u":"/research/how-do-i-know-when-ai-is-wrong","f":"The detectable failures are the ones where the reader has independent grounds. Everything else is confidence, and confidence is not evidence.","k":"judgement"},{"t":"question","n":"How can AI improve human decision-making?","st":"O","u":"","f":"Open, and far less studied than the harms. The candidates are wider option sets and faster disconfirmation, neither of which the common deployment pattern encourages.","k":"judgement"},{"t":"question","n":"How should leaders combine AI analysis with human judgement?","st":"P","u":"/research/how-should-leaders-respond-to-ai","f":"Specify the division of labour, the human entry point and the override grounds in advance. Without those three it is handover rather than combination.","k":"judgement"},{"t":"question","n":"Is human judgement still valuable in the age of AI?","st":"P","u":"/research/ai-and-human-judgement","f":"Yes, and the value concentrates rather than spreads. It moves to the cases the system has not seen and the ones where somebody has to answer for the outcome.","k":"judgement"},{"t":"question","n":"Can AI make moral judgements?","st":"O","u":"","f":"Open, and it turns on what a moral judgement is. A system can output what a moral reasoner would say without having reasoned morally.","k":"judgement"},{"t":"question","n":"Why do humans make better judgements in uncertain situations?","st":"P","u":"/research/ai-and-human-intuition","f":"Often they do not. The defensible version is that people are better at noticing a situation is novel, which is a different skill from deciding well inside it. Where the environment never held…","k":"judgement"},{"t":"question","n":"What is the difference between automation and autonomous decision-making?","st":"P","u":"/research/who-should-own-ai-strategy","f":"Automation executes a decision somebody already made. Autonomy makes it. Governance changes at that line and most policies do not mark it.","k":"judgement"},{"t":"question","n":"Does AI have judgement?","st":"P","u":"/research/what-is-judgement","f":"Not in the sense the word usually carries, which includes bearing the consequence. It produces output close enough to be mistaken for it.","k":"judgement"},{"t":"question","n":"What kinds of decisions should never be left entirely to AI?","st":"P","u":"/research/how-should-ai-decision-rights-be-allocated","f":"The ones nobody could audit afterwards, and the ones where somebody has to be answerable. Those are two different tests and both matter.","k":"judgement"},{"t":"question","n":"Can humans learn to make completely unbiased judgements?","st":"O","u":"","f":"No, and the useful question is which biases are worth the cost of correcting. Debiasing training has a weak record, and the biases that survive it are usually the ones doing useful work elsewhere.","k":"judgement"},{"t":"question","n":"What is the role of intuition in human judgement?","st":"A","u":"/research/ai-and-human-intuition","f":"Recognition trained by exposure. It carries most of the work where feedback is fast and clear and very little where outcomes arrive late or ambiguously, so the discriminating variable is the…","k":"judgement"},{"t":"question","n":"What is the role of emotion in human judgement?","st":"O","u":"","f":"Open here. Emotion is not the opposite of good judgement and its absence is its own impairment: the clinical cases where affect is damaged produce worse decisions rather than colder rational ones.","k":"judgement"},{"t":"question","n":"How does decision fatigue affect human judgement?","st":"O","u":"","f":"Open here, and the original findings have had a hard replication decade. Treat the strong versions with care, and treat any claim resting on the parole board study with more care still.","k":"judgement"},{"t":"question","n":"Is AI decision-making biased?","st":"A","u":"/research/can-ai-be-unbiased","f":"Yes, in two directions people confuse: bias in the outputs, measured at 85.1 per cent favouring White-associated names in one resume audit, and the human tendency to over-accept them. They need…","k":"judgement"},{"t":"question","n":"Is AI decision-making objective?","st":"P","u":"/research/what-is-automation-bias","f":"No. It is consistent, which is a different property, and consistency makes a bias harder to notice rather than smaller: the same error arrives every time and stops looking like an error.","k":"judgement"},{"t":"question","n":"Does AI reduce bias in decision-making?","st":"P","u":"/research/can-ai-be-unbiased","f":"It removes some human variance and encodes one bias at a scale no individual could reach. The difference that decides the outcome is that a machine can be audited and a screener cannot, so it turns…","k":"judgement"},{"t":"question","n":"What is human-in-the-loop AI?","st":"A","u":"/research/human-in-the-loop-is-not-a-safeguard","f":"A design with a person somewhere in the decision path. The meta-analysis says the common configuration underperforms the stronger party alone.","k":"judgement"},{"t":"question","n":"When should humans trust AI decisions?","st":"P","u":"/research/when-should-i-override-ai","f":"When the failure would be detectable and the cost of missing it is bearable. Both conditions, not either, and the first is the one organisations skip because it is expensive to establish.","k":"judgement"},{"t":"question","n":"Can AI make good decisions with incomplete information?","st":"O","u":"","f":"Open. It produces an answer regardless, which is the difficulty: incompleteness does not show in the output, so a reader cannot tell a well grounded answer from a confidently improvised one.","k":"judgement"},{"t":"question","n":"What is the role of human oversight in AI decisions?","st":"A","u":"/research/what-is-meaningful-human-oversight","f":"To catch the errors the system makes, which requires that somebody could detect them. Oversight that cannot tell a right answer from a plausible one is a control on paper rather than a control in…","k":"judgement"},{"t":"question","n":"What is human oversight in artificial intelligence?","st":"A","u":"/research/what-is-meaningful-human-oversight","f":"A person positioned to review, question or stop an automated decision. The word meaningful is doing the work: the EU AI Act requires it, and the practical test is whether the reviewer could have…","k":"judgement"},{"t":"question","n":"Should humans always have the final say over AI decisions?","st":"P","u":"/research/who-can-override-an-ai-system","f":"Not always, and the more useful question is who holds override authority, on what grounds, and whether anybody has ever used it. A final say nobody exercises is a formality rather than a safeguard.","k":"judgement"},{"t":"question","n":"Human judgement versus AI in decision making","st":"A","u":"/research/ai-and-human-judgement","f":"The versus framing is the error. Across 106 experiments the common human and AI pairing performed worse than the better of either alone, which is an argument about how the pairing is designed rather…","k":"judgement"},{"t":"question","n":"How does human intuition compare with artificial intelligence?","st":"A","u":"/research/ai-and-human-intuition","f":"Both are pattern recognition. They differ in what trained them and in whether the conditions for trustworthy intuition still hold, which Kahneman and Klein reduced to a learnable environment plus…","k":"judgement"},{"t":"question","n":"What are ironies of automation?","st":"A","u":"/research/what-are-the-ironies-of-automation","f":"The ironies of automation are that automating the routine parts of a task leaves the human the hardest residue, monitoring and handling exceptions, while removing the practice that built the…","k":"judgement"},{"t":"question","n":"What is the out-of-the-loop performance problem?","st":"A","u":"/research/what-is-the-out-of-the-loop-performance-problem","f":"The out-of-the-loop performance problem is the loss of a person's ability to take over manual operation when an automated system fails, caused by their having been placed in the role of monitor…","k":"judgement"},{"t":"question","n":"What is situation awareness?","st":"A","u":"/research/what-is-the-out-of-the-loop-performance-problem","f":"Situation awareness is the perception of what is happening around you, the comprehension of what it means, and the projection of what it will mean next.","k":"judgement"},{"t":"question","n":"What is the vigilance decrement?","st":"A","u":"/research/what-is-the-vigilance-decrement","f":"The vigilance decrement is the measurable decline in the probability of detecting rare signals as time on a monitoring task increases. First demonstrated by N. H. Mackworth in 1948 using the Clock…","k":"judgement"},{"t":"question","n":"What is alarm fatigue?","st":"A","u":"/research/what-is-alarm-fatigue","f":"Alarm fatigue is the desensitisation of a person to warning signals caused by exposure to a high volume of them, most of which turn out not to require action, leading to slower responses, silenced…","k":"judgement"},{"t":"question","n":"What is moral crumple zone?","st":"A","u":"/research/what-is-a-moral-crumple-zone","f":"A moral crumple zone is what forms when responsibility for a failure falls on the human nearest an automated system, who had limited real control over its behaviour.","k":"judgement"},{"t":"question","n":"What are misuse, disuse, abuse?","st":"A","u":"/research/what-are-misuse-disuse-and-abuse","f":"Misuse is over-reliance on automation, disuse is unwarranted rejection of it, and abuse is deploying it without regard for the human consequences. Appropriate use is the fourth case.","k":"judgement"},{"t":"question","n":"What is algorithm appreciation?","st":"A","u":"/research/what-is-algorithm-appreciation","f":"Algorithm appreciation is the tendency to weight algorithmic advice more heavily than the same advice from a person. Domain experts are the exception.","k":"judgement"},{"t":"question","n":"What is meaningful human involvement?","st":"P","u":"/research/what-is-meaningful-human-oversight","f":"Meaningful human involvement is the UK statutory test for whether a decision is solely automated: whether a human can exercise real influence before the decision is applied, and has the authority,…","k":"judgement"},{"t":"question","n":"What is human in command?","st":"O","u":"","f":"The human in command principle holds that a person must retain authority over an AI system, as distinct from merely occupying a position in its process.","k":"judgement"},{"t":"question","n":"What are shallow jobs?","st":"A","u":"/research/what-are-shallow-jobs","f":"Shallow jobs are Bain's term for roles in which people rubber-stamp mostly correct AI output without engaging their judgement.","k":"judgement"},{"t":"question","n":"What is amplified oversight?","st":"O","u":"","f":"Amplified oversight is Google DeepMind's term for an oversight signal as good as one a human would give if they understood all the reasons behind a decision.","k":"judgement"},{"t":"question","n":"What is the verification bottleneck?","st":"P","u":"/research/can-a-human-approve-an-ai-decision-at-machine-speed","f":"The verification bottleneck is the proposition that reliance on AI rises with task difficulty at the point where the ability to verify the output falls, widening the gap between believed and actual…","k":"judgement"},{"t":"question","n":"What is falling asleep at the wheel?","st":"O","u":"","f":"Falling asleep at the wheel is what happens when people given high-quality AI become careless and less skilled in their own judgement, because the AI is good.","k":"judgement"},{"t":"question","n":"What is calibration?","st":"A","u":"/research/what-is-calibration","f":"The correspondence between the confidence a system states and how often it is correct.","k":"judgement"},{"t":"question","n":"What is the invisible work of oversight?","st":"A","u":"/research/the-invisible-work-of-oversight","f":"The invisible work of oversight is the labour of supervising an automated system that appears in no workload model and no business case: staying attentive while nothing happens, holding enough of the…","k":"judgement"},{"t":"question","n":"What is over-reliance?","st":"A","u":"/research/what-is-over-reliance","f":"Dependence on an automated system beyond the point at which the person relying on it could detect that it was wrong.","k":"judgement"},{"t":"question","n":"What is override authority?","st":"A","u":"/research/who-can-override-an-ai-system","f":"The assigned power of a named person to disregard, reverse or stop an AI system's output, held together with the competence, information and organisational standing required to use it.","k":"judgement"},{"t":"question","n":"What is silent failure?","st":"A","u":"/research/what-is-silent-failure","f":"Silent failure is a failure that produces no error signal a person can act on: the system carries on, the output looks ordinary, and nothing marks the point at which it stopped being right.","k":"judgement"},{"t":"question","n":"What is fail-plausible?","st":"A","u":"/research/what-is-silent-failure","f":"Fail-plausible is a silent failure in which a language model turns the error into fluent, plausible narrative and delivers it to the user, so the observer is not merely blind to the failure but is…","k":"judgement"},{"t":"question","n":"Is it ethical to let AI judge people?","st":"A","u":"/research/is-it-ethical-to-let-ai-judge-people","f":"It turns on four obligations usually collapsed into one: explanation the person can use, contest by somebody able to change the outcome, a named human carrying it, and verification on the population…","k":"judgement"},{"t":"question","n":"Can AI be unbiased, or does it reproduce human bias?","st":"A","u":"/research/can-ai-be-unbiased","f":"It reproduces it, measurably, and unbiased is not one target: several reasonable fairness criteria are provably incompatible, so every system judging people has already chosen between them.","k":"judgement"},{"t":"question","n":"What makes an automated decision explainable enough?","st":"A","u":"/research/is-it-ethical-to-let-ai-judge-people","f":"The Court of Justice answered this in Dun and Bradstreet Austria: describe the procedure and principles actually applied so the person can see which of their data was used. Publishing the algorithm…","k":"judgement"},{"t":"question","n":"Do bias audits actually happen?","st":"A","u":"/research/can-ai-be-unbiased","f":"Rarely. Under the first such law, a field audit of 391 employers found about 5 per cent had published one, and the state comptroller found seventeen potential breaches in a sample where the enforcing…","k":"judgement"},{"t":"question","n":"Which decisions about people count as high risk?","st":"A","u":"/research/is-it-ethical-to-let-ai-judge-people","f":"Annex III of the EU AI Act names them, including access to education, evaluation of learning outcomes that steer a path, and a list of employment decisions. The organising idea is what a person is…","k":"judgement"},{"t":"question","n":"What is the expertise reversal effect?","st":"A","u":"/research/what-is-the-expertise-reversal-effect","f":"Instructional support that helps a beginner learn can hinder someone who already knows the material. Kalyuga, Ayres, Chandler and Sweller named it in 2003; the mechanism is redundancy against the…","k":"learning"},{"t":"question","n":"Should beginners use AI more than experts, or less?","st":"P","u":"/research/what-is-the-expertise-reversal-effect","f":"It depends on what is being measured. On output produced with the tool in hand, the less experienced person gains more, in four separate field experiments. On what people can do once it is taken…","k":"learning"},{"t":"question","n":"What is guidance fading?","st":"A","u":"/research/what-is-the-expertise-reversal-effect","f":"Reducing instructional support step by step as knowledge grows. Renkl and colleagues found a fading procedure beat an abrupt switch from worked examples to problems. No general-purpose AI interface…","k":"learning"},{"t":"question","n":"Does the expertise reversal effect apply to AI assistance?","st":"O","u":"","f":"Untested. Nobody has varied AI scaffolding against measured prior expertise in a domain and then tested unaided performance. Bastani varied the interface, not the learner.","k":"learning"},{"t":"question","n":"Does using GPS damage your brain?","st":"A","u":"/research/does-gps-damage-your-brain","f":"No study has shown it. The taxi driver research measured acquisition over four years; the one study of removal is behavioural, has 13 people at follow-up and never used a scanner.","k":"learning"},{"t":"question","n":"What is the Knowledge of London?","st":"A","u":"/research/does-gps-damage-your-brain","f":"Transport for London's taxi examination, introduced in 1865: 320 Blue Book runs within six miles of Charing Cross, seven stages, three to four years.","k":"learning"},{"t":"question","n":"Is there good evidence that delegating a skill makes you lose it?","st":"A","u":"/research/does-gps-damage-your-brain","f":"Much less than the acquisition evidence, and the asymmetry is a fact about research design rather than about safety. Removal studies need someone to take a working tool away from a professional.","k":"learning"},{"t":"question","n":"Is my child skipping the part that builds the skill?","st":"P","u":"/research/what-is-productive-struggle","f":"The question to ask about any homework. If the difficulty was the point, removing it removes the lesson.","k":"learning"},{"t":"question","n":"How do I tell whether my child understands or has just produced?","st":"P","u":"/research/what-is-the-illusion-of-competence","f":"Ask them to explain it a day later, without the work in front of them. Fluency while reading is not knowledge.","k":"learning"},{"t":"question","n":"Should schools and employers give young people the same advice?","st":"O","u":"","f":"Open, and currently they do not. Schools are protecting formation; employers are buying output. Nobody owns the handover.","k":"learning"},{"t":"question","n":"Does the brain really stop developing at 25?","st":"A","u":"/research/does-the-brain-mature-at-25","f":"No, and the number came from a 2004 magazine interview rather than a finding. A 2025 study of 3,802 people puts the end of the adolescent epoch nearer 32.","k":"learning"},{"t":"question","n":"At what age does AI use start to matter for capability?","st":"P","u":"/research/does-the-brain-mature-at-25","f":"Wrong shape of question. It depends on which capability and when its practice normally happens, not on an age at which something completes.","k":"learning"},{"t":"question","n":"Which capabilities are built by practice rather than instruction?","st":"A","u":"/research/what-is-deliberate-practice","f":"The ones that need effortful repetition with feedback. Those are exactly the ones a machine can now perform without you.","k":"learning"},{"t":"question","n":"Should a teenager use AI differently from an adult?","st":"P","u":"/research/should-children-use-ai","f":"Yes, but the reason is exposure to unformed practice rather than an unfinished brain. What matters is which repetitions are being skipped.","k":"learning"},{"t":"question","n":"Does AI harm younger learners more than older ones?","st":"O","u":"","f":"Open. Plausible on the practice argument, and the age-stratified evidence to settle it does not yet exist.","k":"learning"},{"t":"question","n":"Is it different for maths than for writing?","st":"O","u":"","f":"Open, and probably yes. Skill decay research is domain-specific and the transfer question has not been settled for AI.","k":"learning"},{"t":"question","n":"Does AI help or harm learning?","st":"P","u":"/research/how-humans-learn-with-ai","f":"Both, and the interface decides which. The same model produced the best and the worst outcome in the same experiment.","k":"learning"},{"t":"question","n":"How do you assess students when AI can do the assignment?","st":"A","u":"/research/how-to-assess-students-when-ai-can-do-the-assignment","f":"Not by detection. Sydney concedes prohibition is unenforceable; Denmark now requires oral defence of home-written exams.","k":"learning"},{"t":"question","n":"Should children use AI?","st":"A","u":"/research/should-children-use-ai","f":"How they use it matters far more than whether they do. Age-banded, and declining to advise on under-8s because the evidence does not exist.","k":"learning"},{"t":"question","n":"What is desirable difficulty?","st":"A","u":"/research/what-is-desirable-difficulty","f":"Bjork. Conditions that make study feel harder improve long-term retention. The mechanism underneath half of this research.","k":"learning"},{"t":"question","n":"What is productive struggle?","st":"A","u":"/research/what-is-productive-struggle","f":"Struggling before being taught beats being taught first, at g = 0.36 across 53 studies. It reverses for young children and for general skills, which the authors report themselves.","k":"learning"},{"t":"question","n":"What happens to homework?","st":"O","u":"","f":"Open, and one of the highest-volume parent questions with almost no serious answers.","k":"learning"},{"t":"question","n":"How should teachers use AI?","st":"A","u":"/research/how-will-ai-change-teaching","f":"For preparation, on the strength of a school-randomised trial that cut planning time 31 per cent with no quality difference a blinded panel could detect. English teachers have already concentrated it…","k":"learning"},{"t":"question","n":"How do adults learn with AI?","st":"A","u":"/research/does-using-ai-stop-you-learning","f":"Answered 14 September 2026 by assembling the four designs that withdraw the assistance before measuring. The split is delegation against engagement, not access against abstinence, and it reproduces…","k":"learning"},{"t":"question","n":"Does corporate training still work?","st":"P","u":"/research/why-reskilling-programmes-mostly-fail","f":"Completion is not capability. Five structural reasons, and the published data that would change the position.","k":"learning"},{"t":"question","n":"What should schools teach now?","st":"P","u":"/research/official-guidance-on-ai-in-education","f":"Tools change faster than any curriculum can be rewritten, so the durable question is which capabilities a school still builds deliberately once drafting, summarising and first-pass analysis are…","k":"learning"},{"t":"question","n":"How do people become experts?","st":"P","u":"/research/the-missed-reps-of-ai-seniority","f":"Through repetitions at the edge of current ability, with feedback, over years. Remove the repetitions and the mechanism has nothing to work on.","k":"learning"},{"t":"question","n":"What happens when people stop practising?","st":"A","u":"/research/capability-debt","f":"Measured in nineteen endoscopists: unassisted detection fell six percentage points within months of AI exposure. Capability does not disappear on a schedule anyone tracks, which is what makes it a…","k":"learning","alt":["capability gap","skills debt","losing skills to ai"]},{"t":"question","n":"Why does rereading feel like it works?","st":"A","u":"/research/what-is-retrieval-practice","f":"Because fluency is mistaken for knowing. Restudying beat testing at five minutes, 81 per cent to 75, and lost badly at one week, 42 to 56.","k":"learning"},{"t":"question","n":"Does struggling always help learning?","st":"A","u":"/research/what-is-productive-struggle","f":"No, and the meta-analysis says so itself. It works at g = 0.36 for older learners on specific content, and reverses for second to fifth graders and for domain-general skills.","k":"learning"},{"t":"question","n":"Can AI provide productive struggle, or only simulate it?","st":"P","u":"/research/what-is-productive-struggle","f":"The guardrailed arms of two experiments preserved most of the gain and removed most of the harm, which suggests it can be designed for.","k":"learning"},{"t":"question","n":"Can AI help someone see what they do not know?","st":"O","u":"","f":"Open, and the most useful thing it could do. Currently it does the opposite: fluent answers inflate estimates of one's own knowledge.","k":"learning"},{"t":"question","n":"How should learning be assessed when the process is invisible?","st":"P","u":"/research/how-to-assess-students-when-ai-can-do-the-assignment","f":"Not by detection. Denmark now requires oral defence of home-written exams.","k":"learning"},{"t":"question","n":"Should my child's school ban AI?","st":"O","u":"","f":"Open, and the institutional question parents actually ask. Thirteen official guidance documents read at source, and three present any original data.","k":"learning"},{"t":"question","n":"Does AI detection work?","st":"A","u":"/research/does-ai-detection-work","f":"Not well enough to accuse anyone, and the errors fall hardest on second-language writers.","k":"learning"},{"t":"question","n":"What is the illusion of competence?","st":"A","u":"/research/what-is-the-illusion-of-competence","f":"The gap between felt and actual capability. It is why deskilling is not reported by the deskilled, and why every self-report survey of AI and skills is measuring worry rather than capability.","k":"learning"},{"t":"question","n":"What is retrieval practice?","st":"A","u":"/research/what-is-retrieval-practice","f":"Recall strengthens memory; rereading mostly strengthens the feeling of knowing. The winner reverses with delay, so a short-horizon test prefers whichever method taught least.","k":"learning"},{"t":"question","n":"Should teachers use AI to mark work?","st":"A","u":"/research/how-will-ai-change-teaching","f":"Marking is where all four deskilling conditions converge. It is also the application English teachers use least, at 5 per cent. European law reached the same place: evaluating learning outcomes is…","k":"learning"},{"t":"question","n":"How should apprentices use AI?","st":"A","u":"/research/should-juniors-use-ai","f":"The trades, and the German and Swiss models.","k":"learning"},{"t":"question","n":"What should education protect from AI?","st":"P","u":"/research/what-is-productive-struggle","f":"The first attempt, and the moment of recall. Both are what the learning evidence says does the work, and both are what an assistant removes first.","k":"learning"},{"t":"question","n":"What is Paradoxe de la facilitation?","st":"O","u":"","f":"The facilitation paradox is that effort is part of satisfaction. Difficulty produces the good tiredness that comes from work done well, so removing it can strip work of the thing that made it worth…","k":"learning"},{"t":"question","n":"What is knowledge collapse?","st":"A","u":"/research/what-is-knowledge-collapse","f":"Knowledge collapse is the progressive narrowing over time of the knowledge a society actually holds and treats as worth knowing, relative to the broad historical stock it inherited, as cheap…","k":"learning"},{"t":"question","n":"What is epistemic debt?","st":"P","u":"/research/cognitive-debt-and-capability-debt","f":"Epistemic debt is the gap between being able to produce working output with AI and being able to understand or repair it, producing practitioners whose functional usefulness masks low corrective…","k":"learning"},{"t":"question","n":"What is learn, unlearn, relearn?","st":"O","u":"","f":"A widely repeated claim that the illiterate of the twenty-first century will be those who cannot learn, unlearn and relearn, universally attributed to Alvin Toffler's Future Shock.","k":"learning"},{"t":"question","n":"What is never-skilling?","st":"O","u":"","f":"Never-skilling is the failure to form foundational competence during training, because AI substituted for the cognitive effort that would have built it. It differs from deskilling in having no…","k":"learning"},{"t":"question","n":"What is mis-skilling?","st":"O","u":"","f":"Mis-skilling is the acquisition of incorrect reasoning patterns, learned by uncritically adopting AI output that was erroneous or biased. The capability is built rather than lost, and built wrong.","k":"learning"},{"t":"question","n":"How do I learn about AI properly for free?","st":"O","u":"","f":"Open. There is a short list of genuinely free courses worth doing and no page here naming them, which is a gap rather than a judgement.","k":"learning"},{"t":"question","n":"What happens to a student who never struggles?","st":"A","u":"/research/what-is-productive-struggle","f":"The struggle is not a cost of learning that better tools remove. In the evidence it is a substantial part of the mechanism.","k":"students"},{"t":"question","n":"How much unaided work should a student still do?","st":"P","u":"/research/should-juniors-use-ai","f":"Enough to measure with. Not as a principle, but so that somebody can still tell what the student can do alone.","k":"students"},{"t":"question","n":"Will I get caught using AI at university?","st":"A","u":"/research/how-to-use-ai-at-university","f":"Probably not, and that is the weakest reason to avoid it. 94 per cent of wholly AI-written answers went undetected at Reading, and the authors say their own six per cent detection rate likely…","k":"students"},{"t":"question","n":"How many students actually use AI?","st":"A","u":"/research/how-to-use-ai-at-university","f":"94 per cent of full-time UK undergraduates say they use it to help prepare assessed work, but that figure is mostly comprehension. The number who put AI-generated text directly into marked work is 12…","k":"students"},{"t":"question","n":"Can I trust the references AI gives me?","st":"A","u":"/research/how-to-use-ai-at-university","f":"No, and neither can researchers: one in 277 papers in PubMed Central Open Access now carries a reference to a study that does not exist, up from one in 2,828 in 2023. Never cite a source you have not…","k":"students"},{"t":"question","n":"What should a student never hand over to AI?","st":"A","u":"/research/how-to-use-ai-at-university","f":"Your position, your voice, the struggle of learning, the final decision, and anything you must be able to defend. If you cannot say which column a task is in, treat it as one to keep.","k":"students"},{"t":"question","n":"What happens to a student who uses AI for everything for three years?","st":"O","u":"","f":"Open, and it cannot be answered yet for the plainest reason: the people who would be in that study are still at university. The mechanism has support from adjacent evidence on skill decay and…","k":"students"},{"t":"question","n":"Does my university allow AI?","st":"O","u":"","f":"Unanswerable in general, and saying so is the only responsible reply. It varies by institution, by faculty and sometimes by module, it changed again for this year's intake, and the assignment brief…","k":"students"},{"t":"question","n":"Should students still write essays?","st":"P","u":"/research/how-to-assess-students-when-ai-can-do-the-assignment","f":"The essay was never the point. What it certified, and whether anything else certifies it, is the real question.","k":"students"},{"t":"question","n":"Is my degree still worth it?","st":"P","u":"/research/do-apprenticeships-still-work","f":"The IFS puts the average net lifetime return at around 100,000 pounds with enormous variation by subject, and 20 per cent of women and 30 per cent of men projected negative. The authors decline to…","k":"students"},{"t":"question","n":"Do I have to declare that I used AI in my essay?","st":"A","u":"/research/how-to-be-honest-about-using-ai","f":"Increasingly the penalty is for not declaring rather than for using. The test that survives every policy change is whether you could tell your tutor exactly what you did without leaving anything out.","k":"students"},{"t":"question","n":"Will my university detect that I used AI?","st":"A","u":"/research/how-to-be-honest-about-using-ai","f":"Probably not, which is the wrong thing to plan around anyway. At Reading 94 per cent of wholly AI-written submissions went undetected, and detection also fails the other way on 61 per cent of…","k":"students"},{"t":"question","n":"How do I handle a huge university reading list?","st":"A","u":"/research/how-to-handle-forty-readings","f":"Catch it, organise it by module, ground a notebook in your own PDFs, ask the cross-reading questions, then be tested on it. The last step is the best-evidenced one and the one everybody skips.","k":"students"},{"t":"question","n":"How should a student group agree AI rules?","st":"A","u":"/research/group-work-when-everyone-has-ai","f":"Five things, in week one, in writing. One shared context, one person owns the voice, everyone keeps their own drafts, say what you used to each other, one workspace.","k":"students"},{"t":"question","n":"How should I actually use AI day to day?","st":"A","u":"/research/ai-frameworks-compared","f":"Six frameworks answering six different decisions, each set against the established alternative. The comparison is the point: on most of these questions somebody has published an answer with more…","k":"students"},{"t":"question","n":"What happens to the viva and the oral exam?","st":"P","u":"/research/how-to-assess-students-when-ai-can-do-the-assignment","f":"Denmark mandated oral defence from August 2026. It solves authorship and imports a fairness problem nobody has solved.","k":"students"},{"t":"question","n":"How should medical and law students use AI?","st":"O","u":"","f":"The two professions where deskilling evidence is furthest advanced.","k":"students"},{"t":"question","n":"Is my university's AI policy the same as my lecturer's?","st":"P","u":"/research/using-ai-when-your-department-bans-it","f":"Rarely. The institutional policy usually permits more than an individual module brief does, and the brief is the one you are marked against.","k":"students"},{"t":"question","n":"Do the AI rules differ between countries?","st":"P","u":"/research/official-guidance-on-ai-in-education","f":"Substantially, and national guidance is thinner than most people assume. Only one document in the graded set carries statutory force; the rest are advisory.","k":"students"},{"t":"question","n":"What if my course says nothing about AI at all?","st":"P","u":"/research/using-ai-when-your-department-bans-it","f":"Silence is neither permission nor prohibition. Ask in writing, keep the reply, and work as though the answer will be read out later.","k":"students"},{"t":"question","n":"Can I use AI if English is not my first language?","st":"P","u":"/research/how-to-be-honest-about-using-ai","f":"Usually allowed for language and often not for ideas, and the line sits in a different place on almost every course. The question is worth asking before rather than after.","k":"students"},{"t":"question","n":"Am I at more risk as an international student?","st":"O","u":"","f":"Open, and it deserves a proper answer. The academic penalty is the same for everybody; the visa, funding and family consequences that follow are not, and nobody has published what happens next.","k":"students"},{"t":"question","n":"Does using AI count as plagiarism?","st":"P","u":"/research/how-to-be-honest-about-using-ai","f":"Most institutions treat it separately, as academic misconduct rather than plagiarism, because there is no source to have copied from.","k":"students"},{"t":"question","n":"What if the rules change halfway through my degree?","st":"O","u":"","f":"Open. Policies written in 2023 are being rewritten now, and almost nobody has stated whether work is judged against the rules at submission or the rules today.","k":"students"},{"t":"question","n":"What if one module bans AI and another expects it?","st":"A","u":"/research/using-ai-when-your-department-bans-it","f":"Common, and not a contradiction to resolve. Each brief governs its own submission, so the rule you follow is the one attached to the work in front of you.","k":"students"},{"t":"question","n":"How do I tell the difference between using AI well and badly?","st":"A","u":"/research/the-four-levels-of-ai-use","f":"Four levels, and most people never leave the first. The distinction is whether the machine did the thinking or the typing.","k":"students"},{"t":"question","n":"Is there a way to grade my own AI use?","st":"A","u":"/research/the-five-rungs-of-ai-use","f":"Five rungs, from asking for an answer to building something you still use. Where you sit is a better question than which tool you pay for.","k":"students"},{"t":"question","n":"Do AI detectors actually work?","st":"A","u":"/research/does-ai-detection-work","f":"Not reliably enough to convict anybody. The false positive rate is the part that matters and it falls hardest on people writing in a second language.","k":"students"},{"t":"question","n":"Can I be accused of using AI when I did not?","st":"P","u":"/research/proving-you-did-the-work","f":"Yes. That is the harder position, because innocence does not evidence itself. Version history is the cheapest insurance available.","k":"students"},{"t":"question","n":"How do I prove I wrote it myself?","st":"A","u":"/research/proving-you-did-the-work","f":"Drafts, version history, notes and dead ends. Not because you are guilty, but because one day you may need to show that you are not.","k":"students"},{"t":"question","n":"Should I keep my drafts?","st":"A","u":"/research/proving-you-did-the-work","f":"Yes, and the reason is practical rather than moral. A document with no history is indistinguishable from one that arrived complete.","k":"students"},{"t":"question","n":"What actually happens if I get caught?","st":"A","u":"/research/what-happens-if-you-get-caught-using-ai","f":"Usually a zero on the assignment or a failed module, settled quietly, with no hearing. Expulsion is rare and it is the wrong thing to picture.","k":"students"},{"t":"question","n":"How many students are being penalised?","st":"A","u":"/research/what-happens-if-you-get-caught-using-ai","f":"Russell Group universities recorded 2,053 penalties in 2024-25 against roughly 700 the year before. Seven of the twenty-four do not record them at all, so that is a floor.","k":"students"},{"t":"question","n":"What does a failed module actually cost?","st":"A","u":"/research/what-happens-if-you-get-caught-using-ai","f":"Roughly 1,600 pounds at the 2026/27 England fee cap, before rent and before loan interest. Fees differ elsewhere; the arithmetic does not.","k":"students"},{"t":"question","n":"How often does AI invent a source?","st":"A","u":"/research/how-often-does-ai-invent-a-source","f":"One paper in 277 on PubMed cited a study that does not exist in early 2026, and 2,022 court decisions have recorded fabricated material. Both are floors.","k":"students"},{"t":"question","n":"How do I check a reference exists?","st":"A","u":"/research/the-source-rule","f":"Open the document. That is the whole first step, and a whole profession skipped it.","k":"students"},{"t":"question","n":"Can I ask AI for a reading list?","st":"P","u":"/research/the-source-rule","f":"You can, and then every item has to be opened before it is cited. The references it invents look exactly like the ones it does not.","k":"students"},{"t":"question","n":"Is it safe to cite an AI summary of a paper?","st":"A","u":"/research/the-source-rule","f":"No. Cite the original, never the summary, because a summary you have not checked is a claim about a document you have not read.","k":"students"},{"t":"question","n":"Should I let AI summarise my readings?","st":"A","u":"/research/should-i-let-ai-summarise-everything-i-read","f":"It depends whether you will be examined on the reading or on the summary. One of those is a shortcut and the other is the thing being assessed.","k":"students"},{"t":"question","n":"How do I revise with AI?","st":"P","u":"/research/what-is-retrieval-practice","f":"Ask it to test you rather than to explain to you. Retrieval is the part with the evidence behind it. It is also the part that feels worse.","k":"students"},{"t":"question","n":"Is being tested by AI better than reading its summary?","st":"A","u":"/research/how-humans-learn-with-ai","f":"Across nearly a thousand students, the unrestricted group scored 17 per cent below those who never had the tool once it was withdrawn. The tutor group kept most of its gain.","k":"students"},{"t":"question","n":"Am I actually learning, or does it just feel like it?","st":"A","u":"/research/what-is-the-illusion-of-competence","f":"Fluency is a false signal. The test is whether you can produce it without the screen, and most people never run that test.","k":"students"},{"t":"question","n":"Should I write my own draft before using AI?","st":"A","u":"/research/human-at-the-start","f":"Yes, and not for discipline. You cannot notice where a machine diverges from a position you never formed.","k":"students"},{"t":"question","n":"How do I stop my essays sounding like everyone else's?","st":"A","u":"/research/how-do-i-keep-my-own-voice-when-using-ai","f":"The loss runs deeper than style. Writing alongside an opinionated model shifted what 1,506 people thought, not only what they wrote.","k":"students"},{"t":"question","n":"How do I get AI to argue against me?","st":"A","u":"/research/how-do-i-get-ai-to-challenge-me","f":"Ask for the strongest case against your position before you have committed to it in writing, and check the counter-argument is a real one rather than a weak one set up to be knocked down.","k":"students"},{"t":"question","n":"What makes a good prompt for coursework?","st":"A","u":"/research/goal-context-friction-standard","f":"Goal, context, friction, standard. Friction is the one nobody teaches. Nothing else protects your own thinking.","k":"students"},{"t":"question","n":"Which AI tool should I use at university?","st":"O","u":"","f":"Open, and the question is the wrong shape. Brands date in weeks; the four kinds of work they all do are stable, and nothing here sets that out yet.","k":"students"},{"t":"question","n":"What happens when I run out of free messages?","st":"A","u":"/research/running-out-of-messages-makes-you-worse","f":"You get worse before you run out. Scarcity makes people take the first answer and stop pushing back, which is the behaviour with the most evidence behind it.","k":"students"},{"t":"question","n":"Should I pay for AI as a student?","st":"P","u":"/research/running-out-of-messages-makes-you-worse","f":"Quotas reset independently across providers, so hitting a wall on one does not end the day. Knowing that removes most of the felt scarcity.","k":"students"},{"t":"question","n":"Are exams going back to handwriting?","st":"P","u":"/research/how-to-assess-students-when-ai-can-do-the-assignment","f":"In places. A blunt answer to a real problem. Supervised conditions certify a narrower thing than the coursework they replace.","k":"students"},{"t":"question","n":"What can I safely upload to AI?","st":"O","u":"","f":"Open, and the most neglected question on this list. Other people's data, unpublished work and anything under an NDA are not yours to paste.","k":"students"},{"t":"question","n":"Can I upload a lecturer's slides or someone else's work?","st":"O","u":"","f":"Open. Copyright, data protection and course policy all bear on it and they do not give the same answer.","k":"students"},{"t":"question","n":"What if my group partner used AI and did not say?","st":"P","u":"/research/group-work-when-everyone-has-ai","f":"The mark is usually shared and so is the misconduct finding. Agreeing the rule in the first meeting is cheaper than discovering the disagreement at submission.","k":"students"},{"t":"question","n":"Will there be graduate jobs?","st":"A","u":"/research/will-ai-replace-entry-level-jobs","f":"The measured effect is slower hiring rather than dismissals, concentrated in the most exposed occupations. That is a different claim from jobs disappearing.","k":"students","alt":["graduate jobs","junior jobs","entry level roles"]},{"t":"question","n":"What should I study now?","st":"A","u":"/research/what-should-i-tell-my-children-to-study","f":"Nobody can name the safe subject. Those who do are guessing with more confidence than the evidence carries.","k":"students"},{"t":"question","n":"Is a computing degree still worth starting?","st":"P","u":"/research/should-i-still-learn-to-code","f":"The reason to learn has moved rather than gone. Reading and judging code is now the larger part of the job, and that is harder to teach than writing it.","k":"students"},{"t":"question","n":"How do I become employable when AI does entry-level work?","st":"A","u":"/research/how-do-juniors-become-senior","f":"The rungs that built senior judgement are the ones being automated first. Getting the reps deliberately is now something you have to arrange rather than receive.","k":"students"},{"t":"question","n":"What goes on a graduate CV when everyone has the same tools?","st":"P","u":"/research/if-everyone-has-ai-where-is-the-advantage","f":"Whatever survives the tools being universal. Prompting is not scarce, not durable and not the constraint, so what you built is the part worth writing down.","k":"students"},{"t":"question","n":"Can I use AI to write my job applications?","st":"O","u":"","f":"Open, and increasingly consequential. Employers are running the same detectors as universities and have said less about what they will do.","k":"students"},{"t":"question","n":"Can I use AI for a literature review?","st":"P","u":"/research/how-often-does-ai-invent-a-source","f":"Reviews were the paper type most likely to carry a fabricated citation, 57 per cent above the rest. That is structurally the same act as asking a model what a field says.","k":"students"},{"t":"question","n":"Should I tell my supervisor I used AI?","st":"P","u":"/research/how-to-be-honest-about-using-ai","f":"Yes, and earlier than feels comfortable. The disclosure that costs you something is the one that protects you later.","k":"students"},{"t":"question","n":"Is it different if I use AI as a disability adjustment?","st":"O","u":"","f":"Open, and badly served. Tools that are a reasonable adjustment for one student are misconduct for another, and few institutions have written down which is which.","k":"students"},{"t":"question","n":"Is it cheating if everyone else is doing it?","st":"P","u":"/research/what-happens-if-you-get-caught-using-ai","f":"The question answers itself once it is asked about cost rather than detection. What the assignment was bought for is answerable whatever anybody else does.","k":"students"},{"t":"question","n":"Will using AI now stop me becoming good later?","st":"P","u":"/research/the-missed-reps-of-ai-seniority","f":"The mechanism is plausible and the direct evidence is thin. Decay in a formed skill and failure to form one are different, and most research measures the first.","k":"students"},{"t":"question","n":"What do universities owe students on this?","st":"P","u":"/research/official-guidance-on-ai-in-education","f":"More than they currently give. No government in the guidance corpus has issued anything for universities at all.","k":"students"},{"t":"question","n":"What do I tell a graduate joining my team who has never worked unaided?","st":"P","u":"/research/missing-rungs","f":"Give them the reps deliberately, because the job will no longer supply them by accident. That is now a management task, not an onboarding one.","k":"career"},{"t":"question","n":"Am I hiring for capability or for output?","st":"P","u":"/research/synthetic-seniority","f":"Most processes measure output, which AI now supplies. The two came apart and few hiring processes have noticed.","k":"career"},{"t":"question","n":"Should I worry about my child using AI for homework?","st":"P","u":"/research/should-children-use-ai","f":"Ask what the homework was for. If it was certifying knowledge, it still works; if it was building capability, it may not be.","k":"career"},{"t":"question","n":"What should a young person deliberately learn to do unaided?","st":"P","u":"/research/what-stays-human","f":"Whatever they intend to be paid to judge later. Verification is not a skill you can acquire after the fact.","k":"career"},{"t":"question","n":"Will AI replace my job?","st":"A","u":"/research/will-ai-replace-my-job","f":"Almost certainly not as a whole. Task exposure and occupation exposure give different answers, and most commentary conflates them.","k":"career"},{"t":"question","n":"Will AI replace entry-level jobs?","st":"A","u":"/research/will-ai-replace-entry-level-jobs","f":"The exposure evidence is real, the causal evidence is not yet. Youth employment is sensitive to a lot of things that are not AI.","k":"career","alt":["graduate jobs","junior jobs","entry level roles"]},{"t":"question","n":"Which jobs are safest from AI?","st":"A","u":"/research/which-jobs-are-safest-from-ai","f":"No defensible ranked list exists, because task exposure and substitution give opposite answers. Four things predict safety better than occupation.","k":"career"},{"t":"question","n":"How do I stay valuable as AI improves?","st":"A","u":"/research/staying-valuable-in-the-age-of-ai","f":"By holding capabilities that appreciate rather than depreciate. Tool fluency is the fastest-depreciating asset on offer.","k":"career"},{"t":"question","n":"Should I still learn to code?","st":"A","u":"/research/should-i-still-learn-to-code","f":"Yes, for a different reason than five years ago. The signals genuinely conflict, and this page presents rather than resolves them.","k":"career"},{"t":"question","n":"Can I use AI if my department has banned it?","st":"A","u":"/research/using-ai-when-your-department-bans-it","f":"Almost always on the understanding side, which is usually not what the ban covers. If it touches the words that get marked, do not; if it touches your understanding, do.","k":"career"},{"t":"question","n":"What is the best prompt framework?","st":"A","u":"/research/goal-context-friction-standard","f":"Lo's CLEAR is tighter and Google's TCREI is easier to teach. What none of them has is a slot for what the model should hand back to you rather than do for you.","k":"career"},{"t":"question","n":"How do I stop AI doing my thinking for me?","st":"A","u":"/research/think-ai-think","f":"Write your own position before you open the tool, even one line. Without it there is nothing to compare the output against, so every answer arrives sounding correct.","k":"career"},{"t":"question","n":"What should I never let AI do?","st":"A","u":"/research/keep-share-hand-over","f":"Your position, your voice, the learning struggle, the final decision, anything you must be able to defend. Anything you cannot classify defaults to that column.","k":"career"},{"t":"question","n":"How do I check a source an AI gave me?","st":"A","u":"/research/the-source-rule","f":"Establish the source exists before judging whether it is good, which is the one thing SIFT and the CRAAP test do not do, because both were written when every source under discussion existed.","k":"career"},{"t":"question","n":"How do I get experience if AI does entry-level work?","st":"A","u":"/research/how-do-juniors-become-senior","f":"The rungs people used to climb are the ones most easily automated. Answered 6 September 2026: the dividing line is not how much AI you use but what you do in the first ten minutes, and the tasks…","k":"career"},{"t":"question","n":"How do juniors become senior if AI does the junior work?","st":"A","u":"/research/how-do-juniors-become-senior","f":"Junior work was a by-product of senior workload and not a training scheme, so the training has to be chosen deliberately now. Four studies agree that outsourcing is the harm and access is not.","k":"career"},{"t":"question","n":"Should I put AI skills on my CV?","st":"A","u":"/research/proving-you-did-the-work","f":"Signalling tool use is a depreciating claim. Open.","k":"career"},{"t":"question","n":"How do I prove I actually did the work?","st":"A","u":"/research/proving-you-did-the-work","f":"Provenance is becoming a career asset. Open, and rising fast.","k":"career"},{"t":"question","n":"What is synthetic seniority?","st":"A","u":"/research/synthetic-seniority","f":"Output that looks senior produced by someone who has not done the work that makes senior judgement possible.","k":"career"},{"t":"question","n":"How do I future-proof my career?","st":"A","u":"/research/the-mid-career-squeeze","f":"Open, and any useful answer starts by rejecting the frame.","k":"career"},{"t":"question","n":"Is freelancing safe from AI?","st":"A","u":"/research/who-ai-leaves-behind","f":"Open. The platform data is category-specific and more interesting than the headlines.","k":"career"},{"t":"question","n":"What should I tell my children to study?","st":"A","u":"/research/what-should-i-tell-my-children-to-study","f":"No evidence supports ranking subjects by safety. The only scored forecaster got relative growth right 57 per cent of the time.","k":"career"},{"t":"question","n":"What happens to mid-career professionals?","st":"A","u":"/research/the-mid-career-squeeze","f":"Too senior to retrain cheaply, too junior to be safe. Under-covered everywhere.","k":"career"},{"t":"question","n":"Does using AI make me more employable?","st":"P","u":"/research/why-learn-to-prompt-is-weak-career-advice","f":"Less than the advice implies. What appreciates is the ability to judge the output, which is a different skill entirely.","k":"career"},{"t":"question","n":"Does it matter whether you try before asking AI?","st":"A","u":"/research/should-juniors-use-ai","f":"It is the difference the experiments actually measured. The arms that made the user do part of the thinking kept most of the gain and lost most of the harm.","k":"career"},{"t":"question","n":"Who gains most from AI assistance, and who gains least?","st":"P","u":"/research/human-skills-in-the-age-of-ai","f":"The least experienced gain most on output: 34 per cent against 14 on average in one support centre. What that does to their development is unmeasured.","k":"career"},{"t":"question","n":"How does AI affect workers who speak English as a second language?","st":"A","u":"/research/who-ai-leaves-behind","f":"Open, and two-sided. The writing gap narrows while AI detection errors fall hardest on exactly these writers.","k":"career"},{"t":"question","n":"How does AI affect older workers?","st":"A","u":"/research/the-mid-career-squeeze","f":"Open, and mostly discussed as a training problem when it is a judgement-valuation problem.","k":"career"},{"t":"question","n":"What happens to people who cannot afford the better AI tools?","st":"A","u":"/research/who-ai-leaves-behind","f":"Open. Capability differences that track subscription tiers are a new axis of workplace inequality.","k":"career"},{"t":"question","n":"What becomes a credible signal of competence when output is cheap?","st":"P","u":"/research/how-do-you-assess-capability-rather-than-output","f":"Not the artefact. Structure roughly doubles predictive validity and observation needs about eight repetitions.","k":"career"},{"t":"question","n":"How do I get a job when AI can write the application?","st":"A","u":"/research/proving-you-did-the-work","f":"Open, and high volume. The candidate's side of a question the estate currently answers only for employers.","k":"career"},{"t":"question","n":"What should I do if my employer makes me use AI?","st":"A","u":"/research/the-mid-career-squeeze","f":"Open, and asked constantly. The refusal case is covered; the compelled case is not.","k":"career"},{"t":"question","n":"What do I do if I think I have already lost the skill?","st":"P","u":"/research/can-you-regain-a-skill-you-have-lost","f":"Relearning beats learning at every interval tested. What is missing is the protocol, and whether it holds for judgement rather than procedure.","k":"career"},{"t":"question","n":"What can humans do that AI cannot?","st":"P","u":"/research/what-stays-human","f":"The framing invites a list that keeps shrinking. The more durable question is what we would not want performed without the thing underneath it.","k":"career"},{"t":"question","n":"Is the graduate market weak because of AI, or because of interest rates?","st":"P","u":"/research/will-ai-replace-entry-level-jobs","f":"Both were tested against each other for the first time. The junior decline holds inside firms after controlling for industry and time, shows no equivalent in the earlier tightening cycle, and…","k":"career","alt":["graduate jobs","junior jobs","entry level roles"]},{"t":"question","n":"Am I too senior to retrain and too junior to be safe?","st":"A","u":"/research/the-mid-career-squeeze","f":"The mid-career squeeze. Under-covered everywhere, including here.","k":"career"},{"t":"question","n":"Will AI replace lawyers, doctors, accountants or teachers?","st":"P","u":"/research/which-jobs-are-safest-from-ai","f":"Better answered once with a method than fourteen times by profession. Task exposure and substitution give opposite rankings.","k":"career"},{"t":"question","n":"Will AI replace tasks or whole jobs?","st":"P","u":"/research/which-jobs-are-safest-from-ai","f":"The right unit of analysis, and the reason occupational lists mislead.","k":"career"},{"t":"question","n":"How should I analyse my own job for AI exposure?","st":"A","u":"/research/the-mid-career-squeeze","f":"The personal version of the task-exposure method. Open, and high volume.","k":"career"},{"t":"question","n":"How do I demonstrate value when everyone uses AI?","st":"A","u":"/research/the-mid-career-squeeze","f":"Open. Closely related to proving you did the work.","k":"career"},{"t":"question","n":"Will AI make experts more or less valuable?","st":"P","u":"/research/staying-valuable-in-the-age-of-ai","f":"Autor argues AI could widen the reach of expertise. Acemoglu models the gains as modest. Both are in the evidence base.","k":"career"},{"t":"question","n":"What happens to workers who refuse to use AI?","st":"A","u":"/research/the-mid-career-squeeze","f":"The abstention case, taken seriously rather than mocked.","k":"career"},{"t":"question","n":"Does judgement actually pay more?","st":"A","u":"/research/what-is-the-judgement-premium","f":"It is measured in what employers ask for and not in what they pay. PwC's 2026 barometer gives roles where AI takes the routine work twice the job growth and 42 per cent faster advertised salary…","k":"career"},{"t":"question","n":"What are core skills?","st":"A","u":"/research/what-are-core-skills","f":"The World Economic Forum's unit: the skills employers say workers need today, classified against the Forum's own Global Skills Taxonomy so the figure is comparable between editions.","k":"skills"},{"t":"question","n":"What does the 39 per cent figure about skills actually mean?","st":"A","u":"/research/what-are-core-skills","f":"The share of a worker's core skill set employers expect to be transformed or become outdated by 2030. Restating the whole of it as obsolescence drops half of what the report says.","k":"skills"},{"t":"question","n":"Is skill disruption accelerating?","st":"A","u":"/research/what-are-core-skills","f":"Not on this measure. The series runs 35 per cent in 2016, 57 in 2020, 44 in 2023 and 39 in 2025, so it has fallen through both editions covering the period since ChatGPT.","k":"skills"},{"t":"question","n":"What are human skills?","st":"A","u":"/research/human-skills-in-the-age-of-ai","f":"A contested term. Defined and defended here rather than assumed, because most usage is decorative.","k":"skills"},{"t":"question","n":"Which skills become more valuable as AI improves?","st":"P","u":"/research/what-stays-human","f":"Employer surveys say analytical and creative thinking. That is stated demand, not revealed price, and the distinction matters.","k":"skills"},{"t":"question","n":"Are human skills actually becoming scarcer?","st":"O","u":"","f":"Listed as unknown in the state of the evidence. Eight years of surveys are not the same as wage data.","k":"skills"},{"t":"question","n":"What is a SuperSkill?","st":"A","u":"/research/superskill-introduction","f":"A capability that governs the other capabilities: durable across technological cycles, transferable, governing how you work with intelligent systems, and compounding. Four tests, and seven that pass…","k":"skills","alt":["super skills","super skill","superskills","what are super skills","superskills meaning","superskills definition"]},{"t":"question","n":"What are the seven SuperSkills?","st":"A","u":"/research/superskill-introduction","f":"Curiosity, empathy, big picture thinking, change readiness, global adaptability, principled innovation and an augmented mindset.","k":"skills","alt":["super skills","super skill","superskills","what are super skills","superskills meaning","superskills definition"]},{"t":"question","n":"What is an augmented mindset?","st":"A","u":"/research/superskill-augmented-mindset","f":"Working with capable machines without surrendering the judgement that makes the work yours.","k":"skills"},{"t":"question","n":"Why does curiosity matter more now?","st":"A","u":"/research/superskill-curiosity","f":"Because answers became cheap and questions did not. The scarce input moved.","k":"skills"},{"t":"question","n":"Does empathy still matter at work?","st":"A","u":"/research/superskill-empathy","f":"Yes, and the evidence that AI is rated as more empathetic than clinicians makes the question sharper rather than settled.","k":"skills"},{"t":"question","n":"What is big picture thinking?","st":"A","u":"/research/superskill-big-picture-thinking","f":"Holding the shape of a problem when every component can be generated separately and convincingly.","k":"skills"},{"t":"question","n":"What is change readiness?","st":"A","u":"/research/superskill-change-readiness","f":"The capability that decides whether a shift is absorbed or merely survived.","k":"skills"},{"t":"question","n":"What is global adaptability?","st":"A","u":"/research/superskill-global-adaptability","f":"Working across contexts that do not share your assumptions, which is most of them.","k":"skills"},{"t":"question","n":"What is principled innovation?","st":"A","u":"/research/superskill-principled-innovation","f":"Building things on purpose, with the consequences treated as part of the design rather than as an afterthought.","k":"skills"},{"t":"question","n":"What is skill atrophy?","st":"P","u":"/research/what-is-deskilling","f":"Covered within deskilling, which is the older and better-evidenced term for the same phenomenon.","k":"skills"},{"t":"question","n":"What is deskilling?","st":"A","u":"/research/what-is-deskilling","f":"Braverman to Bainbridge to now. Established, and not a SuperSkills coinage.","k":"skills"},{"t":"question","n":"Can you regain a skill you have lost?","st":"A","u":"/research/can-you-regain-a-skill-you-have-lost","f":"Usually, and faster than you built it. Relearning beats learning at every interval tested, though never yet tested on professional judgement.","k":"skills"},{"t":"question","n":"How fast do skills decay?","st":"A","u":"/research/how-fast-do-skills-decay","f":"From d = -0.01 immediately after training to d = -1.4 after a year of non-use, across 189 data points. Cognitive tasks decay faster than physical ones, which is the finding that matters.","k":"skills"},{"t":"question","n":"What is judgement?","st":"A","u":"/research/what-is-judgement","f":"Recognising what a situation is before any option is weighed. Klein, Dreyfus and Polanyi on where it comes from, separated from decision, reasoning and skill, and where Kahneman says it should not be…","k":"skills"},{"t":"question","n":"What is the difference between a skill and a capability?","st":"A","u":"/research/what-is-organisational-capability","f":"A skill performs an activity. A capability achieves an outcome, which needs skills combined with knowledge, judgement and the ability to adapt them to a situation they were not learned in.","k":"skills"},{"t":"question","n":"What is an organisation's capability, as distinct from its people's?","st":"A","u":"/research/what-is-organisational-capability","f":"It lives in routines, in Nelson and Winter's sense: coordinated sequences that persist beyond the individuals running them. So it can be lost with nobody leaving.","k":"skills"},{"t":"question","n":"Are interpersonal skills becoming more valuable than information-handling ones?","st":"P","u":"/research/human-skills-in-the-age-of-ai","f":"Partly answered as an argument, and now with an early empirical signal behind it. The Stanford work reports competencies shifting from information-focused towards interpersonal. Treat it as a signal…","k":"skills"},{"t":"question","n":"What are complementary skills?","st":"O","u":"","f":"Complementary skills are the OECD's term for teamwork, autonomy, problem solving, creative thinking, communication, collaboration and emotional intelligence: the capabilities that enable…","k":"skills"},{"t":"question","n":"What is hybrid intelligence?","st":"O","u":"","f":"Hybrid intelligence is the European Policy Centre's proposed basis for skills policy: technical AI literacy combined with domain expertise and distinctively human capabilities, rather than AI skills…","k":"skills"},{"t":"question","n":"What is research taste?","st":"O","u":"","f":"Research taste is knowing what to study next, which experiment to run, and sensing where a new approach might lie. It has proved hard to train because the feedback loops are long and the data is thin.","k":"skills"},{"t":"question","n":"What is ai literacy?","st":"A","u":"/research/what-does-ai-literacy-mean-for-leaders","f":"In the European Union, yes. Article 4 of the EU AI Act has applied since 2 February 2025 and requires providers and deployers of AI systems to take measures to ensure, to their best extent, a…","k":"skills"},{"t":"question","n":"What is substitution myth?","st":"A","u":"/research/what-is-the-substitution-myth","f":"The assumption that new technology can be introduced as a simple substitution of machines for people, preserving the basic system while improving it on some output measures.","k":"skills"},{"t":"question","n":"What is tacit knowledge?","st":"A","u":"/research/what-is-tacit-knowledge","f":"Knowledge that resists full articulation, acquired through experience and practice rather than instruction, and typically transmitted through shared work rather than documentation.","k":"skills"},{"t":"question","n":"What skills should I learn now?","st":"O","u":"","f":"The scarcity test matters more than the usefulness test: not what is useful, but what is hard to source.","k":"skills"},{"t":"question","n":"What stays human?","st":"A","u":"/research/what-stays-human","f":"A narrower set than the optimists claim and a wider one than the pessimists allow. The boundary keeps moving, and that movement is the finding.","k":"work"},{"t":"question","n":"How do humans and AI make decisions together?","st":"A","u":"/research/human-ai-decision-making","f":"Badly, by default. The evidence says the common configuration underperforms whichever party was stronger alone.","k":"work"},{"t":"question","n":"Should I check everything AI produces?","st":"P","u":"/research/the-verifiers-discount","f":"Checking everything is an aspiration rather than a policy. What matters is where the verification is placed and who is paid for it.","k":"work"},{"t":"question","n":"Is usage the same as adoption?","st":"A","u":"/research/usage-theatre","f":"No. Seat counts and prompt volumes measure activity. Nothing in them says capability changed.","k":"work"},{"t":"question","n":"What is the unclaimed hour?","st":"A","u":"/research/the-unclaimed-hour","f":"The time AI gives back that nobody designs a use for, and which therefore disappears.","k":"work"},{"t":"question","n":"Who gets credit when AI helped?","st":"A","u":"/research/outsourced-recognition","f":"Recognition follows visible output, and AI makes output cheap to produce and hard to attribute.","k":"work"},{"t":"question","n":"How do you redesign a job around AI?","st":"A","u":"/research/the-shape-of-the-organisation-after-ai","f":"Open, and needed as a method with a worked example rather than a principle.","k":"work"},{"t":"question","n":"What happens to middle management?","st":"A","u":"/research/the-shape-of-the-organisation-after-ai","f":"Open, and asked constantly. Middle management is where work is allocated and judgement is checked, which are the two functions AI touches most directly. Nobody has yet measured what happens to a…","k":"work"},{"t":"question","n":"How do you run a meeting when AI attends it?","st":"A","u":"/research/what-ai-does-to-a-team","f":"WATCH. Genuinely new, and moving fast enough to be worth watching before writing.","k":"work"},{"t":"question","n":"Should we cut headcount because of AI?","st":"A","u":"/research/the-shape-of-the-organisation-after-ai","f":"Open, and the useful version names the cases where it destroyed capability.","k":"work"},{"t":"question","n":"How do you hire when everyone uses AI?","st":"A","u":"/research/proving-you-did-the-work","f":"Assessment, but for employers. Shares its evidence base with the education cluster.","k":"work"},{"t":"question","n":"What is the verifier's discount?","st":"A","u":"/research/the-verifiers-discount","f":"Verification is skilled work priced as administrative residue, so organisations get less of it than they think they bought.","k":"work"},{"t":"question","n":"Does AI help experienced or inexperienced workers more?","st":"P","u":"/research/human-skills-in-the-age-of-ai","f":"The gains skew towards the less experienced in four independent settings. Whether that is good news depends on what happens to the experienced.","k":"work"},{"t":"question","n":"What is AI literacy, and is it the right goal?","st":"A","u":"/research/what-does-ai-literacy-mean-for-leaders","f":"Now a legal obligation, and a weak frame for a strong duty. Literacy language invites a training solution, and completion is not capability.","k":"work"},{"t":"question","n":"Who bears the cost when AI removes developmental work?","st":"P","u":"/research/missing-rungs","f":"The organisation books the saving, the junior absorbs the loss, and the bill arrives years later in a thinner senior tier. Nobody's P&L carries it.","k":"work"},{"t":"question","n":"Does AI increase managerial surveillance?","st":"A","u":"/research/what-ai-does-to-a-team","f":"Open, and the mechanism is the same one that makes AI useful: a system that drafts your work can also log it. The two are hard to separate by design.","k":"work"},{"t":"question","n":"Does AI reduce workers' autonomy?","st":"A","u":"/research/what-ai-does-to-a-team","f":"Open. Autonomy is discretion over how the work gets done, and a tool that supplies the how is a different thing from one that supplies the answer.","k":"work"},{"t":"question","n":"Who pays for the time it takes to verify AI output?","st":"P","u":"/research/the-verifiers-discount","f":"Almost nowhere is it budgeted. Verification is priced as administrative residue and absorbed by whoever is closest to the risk.","k":"work"},{"t":"question","n":"What happens when AI removes the visible work but increases the invisible responsibility?","st":"A","u":"/research/the-invisible-work-of-oversight","f":"Bainbridge, 1983. Taking away the easy parts of a task can make the difficult parts harder, because the easy parts were the practice.","k":"work"},{"t":"question","n":"Does AI increase the number of decisions each person has to make?","st":"A","u":"/research/what-ai-does-to-a-team","f":"Open. If the drafting cost falls to near zero, more options reach the person who has to choose, and choosing was already the expensive part.","k":"work"},{"t":"question","n":"How should teams record decisions that AI influenced?","st":"P","u":"/research/what-is-the-shared-prompt-review","f":"Article 12 of the EU AI Act requires the log without settling what belongs in it. The Shared Prompt Review answers the team half, four things on the table, and leaves the regulatory record open.","k":"work"},{"t":"question","n":"How do you stop AI flattening what a team thinks?","st":"A","u":"/research/what-is-the-shared-prompt-review","f":"The flattening is measured: 776 professionals at Procter and Gamble, and AI erased the difference between what a technical and a commercial specialist proposed. The remedy is a practice nobody has…","k":"work"},{"t":"question","n":"Where does verification belong in an AI-assisted workflow?","st":"P","u":"/research/who-owns-verification-when-ai-does-the-work","f":"Placement decides whether it happens. Review after generation is the weakest position and the most common one.","k":"work"},{"t":"question","n":"How do you redesign a workflow so people keep meaningful judgement?","st":"P","u":"/research/delegation-boundary-map","f":"Decide where judgement stays before deciding what to automate. The ordering is the whole intervention.","k":"work"},{"t":"question","n":"How do you stop AI becoming the default first resort?","st":"P","u":"/research/should-juniors-use-ai","f":"Attempt before you ask. It is the difference the two guardrail experiments actually measured.","k":"work"},{"t":"question","n":"What should an AI pilot measure besides time saved?","st":"P","u":"/research/how-do-you-measure-ai-adoption-properly","f":"In the one randomised trial that checked, self-reported time saved had the wrong sign.","k":"work"},{"t":"question","n":"How do you know whether a workflow is designed for learning?","st":"P","u":"/research/what-is-productive-struggle","f":"Ask whether anyone still attempts the task before the answer arrives. That is the whole test.","k":"work"},{"t":"question","n":"Will AI take over decision making?","st":"O","u":"","f":"Open, and if it happens it happens by accumulation rather than by anybody deciding. That is the drift argument: no single delegation looks like a transfer of authority, and the total is one.","k":"work"},{"t":"question","n":"What are the risks of relying on AI for decisions?","st":"P","u":"/research/what-is-over-reliance","f":"Two that compound: the capability that stops being exercised, and the oversight that becomes formal because the system is usually right.","k":"work"},{"t":"question","n":"Who supervises work they cannot do themselves?","st":"A","u":"/research/who-supervises-work-they-cannot-do","f":"Supervision has become approval, and no management system in common use can tell the difference.","k":"work"},{"t":"question","n":"Are executives reading summaries instead of the source?","st":"A","u":"/research/what-ai-does-to-a-team","f":"A quiet change in how senior decisions get made, with no measurement anywhere. Open.","k":"work"},{"t":"question","n":"What happens to institutional memory?","st":"A","u":"/research/what-happens-to-institutional-memory","f":"It splits. Retrieval of the recorded part improves; the reasons, the rejected options and the trust calibration were never recorded and move through people.","k":"work"},{"t":"question","n":"Does AI change what a meeting is for?","st":"P","u":"/research/should-ai-attend-my-meetings","f":"If the record is automatic and complete, the marginal value of being present falls and meetings drift towards broadcast.","k":"work"},{"t":"question","n":"Is my organisation measuring the right thing?","st":"A","u":"/research/how-do-you-measure-ai-adoption-properly","f":"Almost certainly not. Licences, seats and prompts all rise while nothing changes about capability.","k":"work"},{"t":"question","n":"What is the difference between augmentation and automation?","st":"A","u":"/research/who-should-own-ai-strategy","f":"Established distinction, and routinely collapsed in practice. Open.","k":"work"},{"t":"question","n":"How do I decide what to automate?","st":"P","u":"/research/delegation-boundary-map","f":"Nine stages, and the question answered per stage rather than per tool.","k":"work"},{"t":"question","n":"What is human-on-the-loop?","st":"P","u":"/research/what-is-meaningful-human-oversight","f":"Oversight without per-decision review. Article 14 does not require a human to approve every decision.","k":"work"},{"t":"question","n":"Does AI increase workload?","st":"P","u":"/research/the-invisible-work-of-oversight","f":"Partial, and conditional. Task productivity and workload are separable: cheaper units say nothing about how many units follow. That is a management response, not a property of the tool.","k":"work"},{"t":"question","n":"Should employees disclose when they use AI?","st":"A","u":"/research/proving-you-did-the-work","f":"Disclosure norms are forming now and differ by context.","k":"work"},{"t":"question","n":"Who owns AI-generated work?","st":"O","u":"","f":"Open. Legal and practical answers diverge.","k":"work"},{"t":"question","n":"Does AI reduce knowledge-sharing between colleagues?","st":"P","u":"/research/what-happens-to-institutional-memory","f":"Partial. Early work points to rerouting rather than reduction, on very little evidence. The mechanism is stated; the effect is not established.","k":"work"},{"t":"question","n":"How should AI change performance management?","st":"O","u":"","f":"Open, and urgent because performance systems measure output, which is the first thing AI inflates. A system that rewards visible output will now reward tool use and call it performance.","k":"work"},{"t":"question","n":"Should employees be judged differently if they use AI?","st":"O","u":"","f":"Open, and distinct from whether they should disclose it. The defensible position is that the standard is the work, but that collapses where the job is developmental and the point was the practice.","k":"work"},{"t":"question","n":"How do we compare the performance of somebody using AI with somebody who does not?","st":"O","u":"","f":"Open, and the measurement problem is genuine rather than administrative. On current evidence the tools help less experienced workers more on some tasks, so the same output no longer implies the same…","k":"work"},{"t":"question","n":"Who should receive the benefit when AI makes an employee more productive?","st":"O","u":"","f":"Open, and a distribution question rather than a technical one. The employer supplied the tool, the employee supplied the judgement to use it well, and no established principle allocates between them.","k":"work"},{"t":"question","n":"Should pay change when AI changes the value of a job?","st":"O","u":"","f":"Open, and it cuts both ways, which is what makes it hard. Autor and Thompson's data implies some roles appreciate and others commoditise depending on which tasks were removed. Reward systems are…","k":"work"},{"t":"question","n":"How should incentives change when AI produces more of the output?","st":"O","u":"","f":"Open. Any incentive tied to volume of output is now partly an incentive to use the tool, which may be fine or may be exactly what erodes the capability the organisation is paying for.","k":"work"},{"t":"question","n":"How do we stop AI rewarding visible output over real capability?","st":"P","u":"/research/usage-theatre","f":"Partly answered through usage theatre, which names the failure. The measurement answer is the same as everywhere on this estate: assess what people can do unaided, on a date, with an owner. It is…","k":"work"},{"t":"question","n":"What rights should employees have over AI decisions made about them?","st":"O","u":"","f":"Open, and partly settled in law in some jurisdictions and not others. What is not settled anywhere is the case where AI informed a decision that a human then made, which is most of them.","k":"work"},{"t":"question","n":"What should an employee be able to appeal when AI influences a decision about them?","st":"O","u":"","f":"Open, and appeal is only meaningful if the decision can be reconstructed. Article 12 logging makes reconstruction possible for high-risk systems and requires nobody to do it.","k":"work"},{"t":"question","n":"How do we know whether AI is making our people decisions fairer or less fair?","st":"O","u":"","f":"Open, and it requires a baseline almost nobody has. Human hiring and promotion decisions were not audited for consistency before, so 'fairer than what?' has no measured answer.","k":"work"},{"t":"question","n":"Does AI create new inequalities between employees?","st":"O","u":"","f":"Open. Differential access to tools, to training and to the kind of work that lets you use them well is the mechanism. CIPD's position is that wherever AI has a human impact the people function should…","k":"work"},{"t":"question","n":"Is it still my idea if AI helped me write it?","st":"A","u":"/research/is-it-still-my-idea-if-ai-helped-me-write-it","f":"Open, and a different question from whether the work is any good. The older form of imposter feeling asked whether an idea was strong enough and could be settled by producing evidence. This one asks…","k":"work"},{"t":"question","n":"Who ends up worse off as AI spreads?","st":"P","u":"/research/who-ai-leaves-behind","f":"Partly answered at the level of individuals, and newly complicated at every larger scale. Anthropic's exposure data inverts the usual assumption: the most exposed workers are more likely female, more…","k":"work"},{"t":"question","n":"Does AI blur the boundaries between professions?","st":"O","u":"","f":"Open, and newly measurable. OpenAI's own telemetry finds 43.5 per cent of occupation-specific ChatGPT messages concern tasks belonging to a different occupation, rising to 77 per cent for customer…","k":"work"},{"t":"question","n":"Who is accountable when work crosses a professional boundary?","st":"O","u":"","f":"Open. This is where the crossover finding meets the professions. Regulated work carries duties attached to a person holding a qualification. If a non-lawyer drafts something legal and a…","k":"work"},{"t":"question","n":"Which tasks do workers not want automated, even where AI can do them?","st":"A","u":"/research/which-tasks-do-workers-not-want-automated","f":"Refusal is patterned. Workers were positive about automating 46.1 per cent of 844 tasks in their own occupations after being prompted to weigh job loss and lost enjoyment. Where they refused,…","k":"work"},{"t":"question","n":"Is there a right level of human involvement for a task?","st":"A","u":"/research/which-tasks-do-workers-not-want-automated","f":"There is a preferred level and it is occupation-specific. The Human Agency Scale runs H1 to H5, and H3, an equal partnership outperforming either party alone, was the dominant worker-desired level in…","k":"work"},{"t":"question","n":"Should worker preferences decide what gets automated?","st":"P","u":"/research/which-tasks-do-workers-not-want-automated","f":"Partly. The page argues preferences are evidence and not a veto, and states the three reasons: workers may misjudge the technology, may misjudge their own interests, and may not answer honestly,…","k":"work"},{"t":"question","n":"Do workers and AI experts hold the same view of what AI can do?","st":"A","u":"/research/which-tasks-do-workers-not-want-automated","f":"No, and the disagreement runs one way. Ratings matched on 26.9 per cent of 844 tasks, and on 47.5 per cent the worker wanted more human involvement than the expert judged necessary. Divergence is…","k":"work"},{"t":"question","n":"What is automating versus informating?","st":"A","u":"/research/what-is-automating-versus-informating","f":"Automating replaces human judgement with a machine. Informating generates information that deepens the worker's understanding. The same system can do either, and which one happens is a management…","k":"work"},{"t":"question","n":"What is Frontier Firm?","st":"A","u":"/research/what-is-a-frontier-firm","f":"A Frontier Firm is Microsoft's term for a company built on purchasable machine reasoning, human-agent teams, and a new role for every employee as a manager of agents.","k":"work"},{"t":"question","n":"What is capacity gap?","st":"O","u":"","f":"The capacity gap is Microsoft's term for the deficit between what a business demands and the maximum output humans alone can supply.","k":"work"},{"t":"question","n":"What is Work Chart?","st":"O","u":"","f":"The Work Chart is Microsoft's proposed successor to the org chart, structured around jobs that need doing rather than functional expertise.","k":"work"},{"t":"question","n":"What is centaur and cyborg work?","st":"P","u":"/research/what-is-the-jagged-frontier","f":"Centaur work divides tasks cleanly between person and machine. Cyborg work interweaves them continuously, moving back and forth across the jagged frontier.","k":"work","alt":["jagged frontier explained","jagged edge of ai"]},{"t":"question","n":"What is aI control as work?","st":"O","u":"","f":"AI control is Narayanan and Kapoor's prediction that a steadily greater share of what people do in their jobs will consist of controlling AI rather than doing the underlying task.","k":"work"},{"t":"question","n":"What is identity commoditisation?","st":"O","u":"","f":"Identity commoditisation is the erosion of a professional's sense of uniqueness and dignity as their role narrows to supervising a system, until the work that carried their identity is no longer the…","k":"work"},{"t":"question","n":"What is the experiential chasm?","st":"O","u":"","f":"The experiential chasm is the gap between the small group who have spent real hours with frontier models and the majority still experimenting superficially. Bain describes it as neither a seniority…","k":"work"},{"t":"question","n":"Does AI actually make people more productive?","st":"A","u":"/research/does-ai-actually-make-people-more-productive","f":"On narrow specified tasks, by a lot. In continuing work inside a system somebody knows, uncertain and sometimes negative. Nothing in the national statistics yet, and self-report is unreliable in both…","k":"work","alt":["ai productivity","productivity gains","does ai save time"]},{"t":"question","n":"Will AI replace programmers?","st":"A","u":"/research/will-ai-replace-programmers","f":"No study shows replacement. A greenfield coding task went 55.8 per cent faster; experienced developers in mature repositories were measured 19 per cent slower. The evidenced risk is to how developers…","k":"work"},{"t":"question","n":"Does AI save time?","st":"A","u":"/research/does-ai-actually-make-people-more-productive","f":"Two government trials of 23,500 licences report 26 and 19 minutes a day, both self-reported, neither with a baseline. Where the time then goes is unmeasured everywhere.","k":"work","alt":["ai productivity","productivity gains","does ai save time"]},{"t":"question","n":"Should the chief executive own AI personally?","st":"A","u":"/research/who-should-own-ai-strategy","f":"Open. The argument for is that judgement allocation is not delegable; the argument against is that chief executives own everything and therefore nothing.","k":"leadership"},{"t":"question","n":"What does a chief executive need to understand that a CTO does not?","st":"P","u":"/research/how-should-leaders-respond-to-ai","f":"Where the organisation will still need human judgement in five years, and what it is doing this year to make sure it has it.","k":"leadership"},{"t":"question","n":"Does where we put AI say what we believe about our people?","st":"A","u":"/research/who-should-own-ai-strategy","f":"Usually yes, and usually without anyone deciding it. The reporting line is a statement of belief that nobody had to write down.","k":"leadership"},{"t":"question","n":"Who is accountable when AI reduces capability rather than cost?","st":"A","u":"/research/who-should-own-ai-strategy","f":"In most structures, nobody. Cost has an owner and a number; capability has neither, so capability is the one that erodes.","k":"leadership"},{"t":"question","n":"What should a board ask when an AI strategy is presented?","st":"P","u":"/research/what-should-a-board-ask-about-ai","f":"Ask what it assumes people will still be able to do, and how that will be checked. Most strategies do not answer either.","k":"leadership","alt":["board questions ai","questions for the board","board oversight of ai"]},{"t":"question","n":"How do we tell an ambitious AI strategy from a reckless one?","st":"P","u":"/research/design-versus-drift","f":"By whether the judgement calls were made in advance or are being discovered. Ambition and drift look identical in a board pack.","k":"leadership","alt":["drift vs design","drift or design","ai drift"]},{"t":"question","n":"Our AI strategy was written eighteen months ago. What is now wrong with it?","st":"A","u":"/research/what-is-wrong-with-an-ai-strategy-written-eighteen-months-ago","f":"Probably not the tools it names. Five things it could not have settled: agents that act, the reversals, Article 14 in force, the shadow use, and what the labs now say about themselves.","k":"leadership"},{"t":"question","n":"How should leaders respond to AI?","st":"A","u":"/research/how-should-leaders-respond-to-ai","f":"Not with a tool rollout. The decisions that matter are about capability, accountability and what the organisation stops doing.","k":"leadership"},{"t":"question","n":"What should a board ask about AI?","st":"A","u":"/research/what-should-a-board-ask-about-ai","f":"Twelve questions, each paired with the answer that should worry you. Ask two or three in passing rather than as an agenda item.","k":"leadership","alt":["board questions ai","questions for the board","board oversight of ai"]},{"t":"question","n":"What does AI literacy mean for leaders?","st":"A","u":"/research/what-does-ai-literacy-mean-for-leaders","f":"A legal obligation under Article 4 of the EU AI Act since February 2025, enforced since August 2026, at every risk tier. Tool training does not satisfy it.","k":"leadership"},{"t":"question","n":"What is the AI readiness lie?","st":"A","u":"/research/ai-readiness-lie","f":"Readiness assessments measure intent and infrastructure. Neither predicts whether capability survives contact with the tools.","k":"leadership"},{"t":"question","n":"Is our AI adoption drift or design?","st":"A","u":"/research/design-versus-drift","f":"Most organisations are not choosing. They are accumulating decisions that were never made, which is drift with a strategy document attached.","k":"leadership","alt":["drift vs design","drift or design","ai drift"]},{"t":"question","n":"Who owns verification when AI does the work?","st":"A","u":"/research/who-owns-verification-when-ai-does-the-work","f":"In most organisations, nobody. It is not in a job description, a budget line or an org chart.","k":"leadership"},{"t":"question","n":"How should decision rights be allocated?","st":"A","u":"/research/how-should-ai-decision-rights-be-allocated","f":"By naming who recommends, inputs, agrees, decides and executes. AI can hold several of those roles and not the decide role, because deciding carries answerability.","k":"leadership"},{"t":"question","n":"What should a CHRO do first?","st":"A","u":"/research/chro-guide-to-ai","f":"Not procurement. The first move is knowing which capabilities the organisation cannot afford to lose.","k":"leadership"},{"t":"question","n":"How do you measure AI adoption properly?","st":"A","u":"/research/how-do-you-measure-ai-adoption-properly","f":"Not by activity, and not by asking people. In the one randomised trial that checked, self-reported time saved had the wrong sign.","k":"leadership"},{"t":"question","n":"What does an AI-capable manager do differently?","st":"A","u":"/research/what-does-an-ai-capable-manager-do-differently","f":"Not use the tools more. Six habits, each traced to a measured result: allocate the task before the tool, keep some work unaided, count disagreements, protect the rungs, read the source, treat…","k":"leadership"},{"t":"question","n":"How do you write an AI use policy that works?","st":"A","u":"/research/how-do-you-write-an-ai-use-policy-that-works","f":"Most are unenforceable and everyone involved knows it. Six elements that survive every model release, and one test that beats legal review.","k":"leadership"},{"t":"question","n":"Should leaders use AI themselves?","st":"A","u":"/research/what-board-oversight-of-ai-looks-like","f":"Open, and the executives-consuming-machine-summaries problem sits underneath it.","k":"leadership"},{"t":"question","n":"What is capability debt?","st":"A","u":"/research/capability-debt","f":"Capability an organisation has stopped maintaining but still assumes it has. It comes due when the system fails or the question is novel.","k":"leadership","alt":["capability gap","skills debt","losing skills to ai"]},{"t":"question","n":"Is AI a change management problem?","st":"A","u":"/research/ai-transformation-not-change-mangement-problem","f":"No, and treating it as one is why so many programmes produce activity without capability.","k":"leadership"},{"t":"question","n":"What goes wrong in AI transformations?","st":"A","u":"/research/common-ai-transformation-challenges","f":"The same things, repeatedly, and almost none of them are technical.","k":"leadership"},{"t":"question","n":"What evidence should a board demand before scaling an AI pilot?","st":"P","u":"/research/what-should-a-board-ask-about-ai","f":"Twelve questions, each paired with the answer that should worry you. Time saved is not on the list.","k":"leadership","alt":["board questions ai","questions for the board","board oversight of ai"]},{"t":"question","n":"How should leaders tell whether an AI programme is building capability or spending it?","st":"A","u":"/research/what-is-a-capability-audit","f":"Adoption reviews measure use; a capability audit measures what remains without the tool. The two can point opposite ways, which is the case worth finding.","k":"leadership"},{"t":"question","n":"How should leaders respond when people quietly work around the AI policy?","st":"A","u":"/research/what-to-do-when-people-work-around-the-ai-policy","f":"Open, and the most honest signal a policy gives. Most are unenforceable and everyone involved knows it.","k":"leadership"},{"t":"question","n":"Why do employees hide their use of AI at work?","st":"A","u":"/research/what-should-we-tell-employees-about-ai-and-headcount","f":"Because the penalty is real and measured: in four PNAS experiments people who used AI were judged lazier and less competent, and 64 per cent of weekly users in Deloitte's 2026 UK survey feared their…","k":"leadership"},{"t":"question","n":"What should a chief executive understand personally before approving AI deployment?","st":"A","u":"/research/what-board-oversight-of-ai-looks-like","f":"Open. Article 4 makes literacy a legal obligation without saying what a chief executive has to hold themselves.","k":"leadership"},{"t":"question","n":"Who decides which human capabilities are worth preserving?","st":"A","u":"/research/what-board-oversight-of-ai-looks-like","f":"Open, and a governance question dressed as a technical one. It is currently decided by whoever chooses what to automate first.","k":"leadership"},{"t":"question","n":"What is AI leadership?","st":"A","u":"/research/ai-leadership","f":"The allocation of judgement: which decisions the machine may make, which stay human, who answers, and what people must stay capable of. Not a technical role.","k":"leadership"},{"t":"question","n":"What does an AI strategy actually have to contain?","st":"A","u":"/research/rules-before-tools","f":"Four layers and ten rules, written before another tool is bought. The August 2025 essay, in full, with dated notes on what has moved.","k":"leadership"},{"t":"question","n":"Is AI-first a strategy or a slogan?","st":"A","u":"/research/is-ai-first-a-strategy-or-a-slogan","f":"A preference order for one decision. Shopify published it and held; Duolingo declared it and retreated in a month. What the phrase leaves undecided.","k":"leadership"},{"t":"question","n":"What happened to the companies that cut staff for AI?","st":"A","u":"/research/what-happened-to-companies-that-cut-staff-for-ai","f":"Klarna cut and rehired, Duolingo declared and retreated, IBM cut and reinvested, Shopify froze and held. The cut was never the decision that mattered.","k":"leadership"},{"t":"question","n":"Do you need to be technical to lead AI?","st":"A","u":"/research/do-you-need-to-be-technical-to-lead-ai","f":"No. The decisions that are a leader's are about accountability, capability and governance. What fluency is for, and what happens to leaders who delegate the judging.","k":"leadership"},{"t":"question","n":"Is it true that 95 per cent of AI pilots fail?","st":"A","u":"/research/is-it-true-that-95-per-cent-of-ai-pilots-fail","f":"The 95 rests on 52 interviews, the 80 is a cited estimate, the 40 is a forecast. All three put the failure in the organisation, not the model.","k":"leadership"},{"t":"question","n":"What is the difference between AI governance and AI leadership?","st":"A","u":"/research/ai-governance-versus-ai-leadership","f":"Governance checks decisions that have been made. Leadership makes them. Most organisations built the first and are waiting for it to produce the second.","k":"leadership"},{"t":"question","n":"Is our AI policy actually enforceable?","st":"A","u":"/research/how-do-you-write-an-ai-use-policy-that-works","f":"Most are not, and everyone involved knows it. Six elements that survive every model release.","k":"leadership"},{"t":"question","n":"What are we legally required to do now?","st":"P","u":"/research/what-is-meaningful-human-oversight","f":"Article 4 on literacy since February 2025, Article 14 on oversight since August 2026. Not legal advice, and the guidance does not exist yet.","k":"leadership"},{"t":"question","n":"How do we allocate AI decision rights?","st":"A","u":"/research/how-should-ai-decision-rights-be-allocated","f":"State for each decision whether the system recommends, decides within limits, or executes. Absent that, the decide role has been given away by default.","k":"leadership"},{"t":"question","n":"How do you audit an AI-assisted decision?","st":"A","u":"/research/how-do-you-audit-an-ai-assisted-decision","f":"Article 12 requires the log. Article 14 does not require per-decision review, except for biometric identification, which almost everyone gets wrong.","k":"leadership"},{"t":"question","n":"What AI risks could be material to this company?","st":"O","u":"","f":"Open, and materiality is the operative word. Most AI risk registers list generic harms rather than the two or three exposures that could actually move this company's numbers or licence to operate.…","k":"leadership"},{"t":"question","n":"What is our appetite for AI risk?","st":"P","u":"/research/what-board-oversight-of-ai-looks-like","f":"Partly answered through the machinery: appetite is meaningful only where a threshold and a named owner exist. What is not answered is how to express appetite for a technology whose failure modes are…","k":"leadership"},{"t":"question","n":"Which AI decisions should come to the board?","st":"P","u":"/research/what-board-oversight-of-ai-looks-like","f":"Partly answered. The defensible line is not size of spend but reversibility and whether the decision changes what the organisation can do unaided. Most escalation criteria use spend, which is the…","k":"leadership"},{"t":"question","n":"What should management report to the board about AI every quarter?","st":"A","u":"/research/what-board-oversight-of-ai-looks-like","f":"Answered. The declared augmentation-or-replacement position annually, the capability floor with an owner and a date, decision rights in writing, reversal thresholds, and one reconstructed decision.…","k":"leadership"},{"t":"question","n":"What would an effective board AI dashboard contain?","st":"P","u":"/research/what-board-oversight-of-ai-looks-like","f":"Partly answered, with a caution. The agenda items are known; turning them into a dashboard risks converting the one question that resists metrics, whether anybody got better at anything, into a…","k":"leadership"},{"t":"question","n":"How does the board know management's claims about AI are true?","st":"A","u":"/research/how-does-a-board-know-managements-claims-about-ai-are-true","f":"Open, and the assurance question underneath most of this territory. Every other material claim to a board has an independent check behind it. AI claims mostly do not, and the people best placed to…","k":"leadership"},{"t":"question","n":"Does the board need independent assurance over AI?","st":"A","u":"/research/how-does-a-board-know-managements-claims-about-ai-are-true","f":"Open. Internal audit is the obvious home and is rarely resourced for it. The harder problem is that assuring an AI system requires the competence to evaluate it, which is scarce in exactly the third…","k":"leadership"},{"t":"question","n":"Who audits our AI systems?","st":"P","u":"/research/how-do-you-audit-an-ai-assisted-decision","f":"Partly answered on method. The organisational answer is usually nobody in particular, and the capability question sits in none of the three lines of defence.","k":"leadership"},{"t":"question","n":"How do we know our AI controls actually work?","st":"P","u":"/research/what-board-oversight-of-ai-looks-like","f":"Partly answered. This is the estate's central governance argument: a control that exists, is documented, has an owner and would not catch anything is drift in its governance form. The only test is…","k":"leadership"},{"t":"question","n":"What would constitute a serious AI incident?","st":"A","u":"/research/what-counts-as-a-serious-ai-incident","f":"Open, and the definitional work has barely been done. Cyber has agreed severity language after twenty years; AI has none, which means incidents get classified into existing categories and the pattern…","k":"leadership"},{"t":"question","n":"What happens when an AI incident occurs?","st":"A","u":"/research/what-counts-as-a-serious-ai-incident","f":"Open. The parts that exist elsewhere, escalation, containment, disclosure, are transferable. What is not transferable is deciding whether to keep running a system that was mostly right.","k":"leadership"},{"t":"question","n":"What should the board do about the September 2026 AI safety warnings?","st":"A","u":"/research/what-should-a-board-do-about-the-ai-safety-warnings","f":"Five decisions that hold whatever the odds: which decisions machines may make in the company's name, who can stop each system, what people must remain able to do unaided, how the board would find…","k":"leadership"},{"t":"question","n":"Do we have a kill switch for the AI we have deployed?","st":"A","u":"/research/what-is-an-ai-kill-switch","f":"Usually a vendor change request and a week. Being able to stop a system is a technical fact; being willing to, mid-quarter, is a leadership fact, and it should be rehearsed before it is needed.","k":"leadership"},{"t":"question","n":"How many AI decisions an hour is our human oversight actually checking?","st":"A","u":"/research/can-a-human-approve-an-ai-decision-at-machine-speed","f":"Decisions per hour, times the minutes a real check takes, against the attention available. Where the first exceeds the second the approval is a sample, and the board should know the sampling rate and…","k":"leadership"},{"t":"question","n":"What do the rogue agent incidents mean for our own agent deployments?","st":"A","u":"/research/what-the-rogue-ai-agent-incidents-mean-for-your-organisation","f":"The reported incidents ran with guardrails disabled, monitoring not enabled, or systems left connected by mistake. Scope, stop, monitoring on by default and a named owner are the four things to…","k":"leadership"},{"t":"question","n":"Which AI failures must be reported to the board?","st":"A","u":"/research/what-counts-as-a-serious-ai-incident","f":"Open, and the reporting threshold has a perverse property: the failures worth escalating are the quiet, systematic ones, and the ones that get escalated are the loud, visible ones.","k":"leadership"},{"t":"question","n":"Should AI oversight sit with the full board or a committee?","st":"A","u":"/research/should-ai-oversight-sit-with-the-full-board-or-a-committee","f":"Open, and genuinely contested. A committee gets depth and creates the impression the rest of the board need not engage. The full board gets engagement without the time to go deep. The audit-committee…","k":"leadership"},{"t":"question","n":"Does our board have enough AI expertise?","st":"A","u":"/research/should-we-appoint-a-director-with-ai-expertise","f":"Open, and usually the wrong question. What a board needs is enough understanding to tell a real control from a described one, which is not the same as technical depth and is not obtained by…","k":"leadership"},{"t":"question","n":"Should we appoint a director with AI expertise?","st":"A","u":"/research/should-we-appoint-a-director-with-ai-expertise","f":"Open, and the cyber precedent is not encouraging. Appointing a single expert tends to concentrate rather than raise competence, and the rest of the board defers on exactly the questions it should be…","k":"leadership"},{"t":"question","n":"How should directors themselves use AI?","st":"A","u":"/research/what-board-oversight-of-ai-looks-like","f":"Answered, and the answer is yes but not for productivity. A director who has never watched a model produce a confident, wrong answer in their own domain has no calibration for the risk they are…","k":"leadership"},{"t":"question","n":"Should directors put confidential board papers into AI systems?","st":"A","u":"/research/should-directors-put-board-papers-into-ai","f":"Open, and the one question here with an immediate practical answer: not into public tools. PwC's guidance says so directly. What remains open is the enterprise case, where confidentiality is…","k":"leadership"},{"t":"question","n":"Can a director rely on an AI-generated summary of a board paper?","st":"A","u":"/research/should-directors-put-board-papers-into-ai","f":"Open, and the sharpest question in the cluster. Ayinde established that the duty to verify does not transfer to whoever used the tool, and that it travels upward. Applying that to a director who read…","k":"leadership"},{"t":"question","n":"How should AI use in board decision-making be recorded?","st":"P","u":"/research/what-is-decision-provenance","f":"Partial from 14 September 2026. There is a concept for the record and no board practice built on it: decision provenance, proposed by Singh, Cobbe and Norval in 2019, logs the inputs, the decision…","k":"leadership"},{"t":"question","n":"Could using AI change a director's duties?","st":"O","u":"","f":"Open, and a question for counsel rather than for this estate. The duty to exercise independent judgement and reasonable care is jurisdictional. What can be said is that Ayinde made a professional…","k":"leadership"},{"t":"question","n":"How should the board oversee AI used by suppliers and third parties?","st":"O","u":"","f":"Open, and the gap most likely to produce the first governance failure. Third-party risk processes ask whether a supplier has a policy. They do not ask whether the supplier's AI use is eroding the…","k":"leadership"},{"t":"question","n":"How should AI change our approach to mergers and acquisitions?","st":"A","u":"/research/what-ai-capability-should-we-look-for-when-acquiring-a-company","f":"Open. Diligence measures systems, licences and talent. Nobody yet diligences whether a target's people can still do the work unaided, which on this estate's argument is the thing most likely to be…","k":"leadership"},{"t":"question","n":"What AI capability should we look for when acquiring a company?","st":"A","u":"/research/what-ai-capability-should-we-look-for-when-acquiring-a-company","f":"Open, and the answer is probably not the models or the tooling, which are purchasable. The durable asset is proprietary data and the judgement to use it, neither of which appears cleanly on a balance…","k":"leadership"},{"t":"question","n":"What happens to continuity planning if AI becomes critical infrastructure?","st":"O","u":"","f":"Open. Continuity plans assume a manual fallback staffed by people who know the manual process. That assumption is the one this research questions, and it sits in plans nobody has revisited.","k":"leadership"},{"t":"question","n":"What is the AI Memo Test?","st":"A","u":"/research/what-is-the-ai-memo-test","f":"The AI Memo Test is a three-question self-assessment for an organisation: whether tool fluency is a basic requirement across roles, whether job descriptions avoid tasks software already handles, and…","k":"leadership"},{"t":"question","n":"Who should own AI strategy in an organisation?","st":"A","u":"/research/who-should-own-ai-strategy","f":"Open, and the placement is usually inherited rather than decided. Transformation, process and technology each own a real part of it and none owns the judgement question.","k":"organisations"},{"t":"question","n":"Is AI a technology strategy or a people strategy?","st":"A","u":"/research/who-should-own-ai-strategy","f":"It is filed as the first and behaves like the second. That mismatch explains a good deal of stalled work.","k":"organisations"},{"t":"question","n":"What does it mean if AI sits under operations rather than strategy?","st":"A","u":"/research/who-should-own-ai-strategy","f":"Open, and revealing. Organisations that file people under operations tend to file AI there too, and the consequence is that nobody senior is accountable for what people can still do.","k":"organisations"},{"t":"question","n":"Is AI an augmentation play or a replacement play?","st":"A","u":"/research/who-should-own-ai-strategy","f":"The question underneath most AI strategy arguments, and one that is very rarely asked out loud. Almost every disagreement about ownership is really a disagreement about this.","k":"organisations"},{"t":"question","n":"Can an organisation pursue augmentation and replacement at the same time?","st":"A","u":"/research/who-should-own-ai-strategy","f":"Open. They are often run in parallel by different functions with different targets, which is how an organisation ends up with two AI strategies and one budget.","k":"organisations"},{"t":"question","n":"How do you tell which one your organisation is actually pursuing?","st":"A","u":"/research/who-should-own-ai-strategy","f":"Read the business case rather than the announcement. If the benefit is headcount, the answer is replacement whatever the language says.","k":"organisations"},{"t":"question","n":"What is missing from most AI strategies?","st":"P","u":"/research/capability-debt","f":"A capability plan. Nearly every strategy has tooling, governance and adoption metrics, and no account of what people must still be able to do.","k":"organisations","alt":["capability gap","skills debt","losing skills to ai"]},{"t":"question","n":"How do you fix an AI strategy that has gone wrong?","st":"P","u":"/research/common-ai-transformation-challenges","f":"Start by establishing which of the failures you actually have, because the loudest one is rarely the real one and the fixes are not interchangeable.","k":"organisations"},{"t":"question","n":"Our AI programme stalled. Is that a technology problem or a capability one?","st":"P","u":"/research/ai-transformation-not-change-mangement-problem","f":"Usually neither of the two things it gets blamed on. Stalls tend to sit where nobody can say what the work is now for.","k":"organisations"},{"t":"question","n":"How do you know an AI strategy is working?","st":"P","u":"/research/how-do-you-measure-ai-adoption-properly","f":"Not from adoption. The question a dashboard should answer is whether anybody got better at anything, and most cannot.","k":"organisations"},{"t":"question","n":"Who should own AI implementation inside an organisation?","st":"A","u":"/research/who-should-own-ai-strategy","f":"Open, and the source of a great deal of stalled work. Ownership tends to sit with whoever bought the tools rather than whoever carries the consequences.","k":"organisations"},{"t":"question","n":"Should AI implementation sit with IT, transformation or the business?","st":"A","u":"/research/who-should-own-ai-strategy","f":"Open. Each placement fails differently, and the failure mode is predictable from the placement.","k":"organisations"},{"t":"question","n":"What does a good AI implementation team look like?","st":"A","u":"/research/who-should-own-ai-strategy","f":"Open. The pattern worth testing is whether anyone on it is accountable for capability rather than for delivery.","k":"organisations"},{"t":"question","n":"When should we stop or reverse an AI deployment?","st":"A","u":"/research/deployment-is-not-a-ratchet","f":"Decide the thresholds while everyone is still pleased with it. The people defending a decision cannot set them afterwards.","k":"organisations"},{"t":"question","n":"Why do AI pilots succeed and rollouts fail?","st":"A","u":"/research/who-should-own-ai-strategy","f":"Open, and widely observed. The candidate explanation worth testing is that pilots are staffed by people who already had the judgement.","k":"organisations"},{"t":"question","n":"What is an AI workforce strategy?","st":"A","u":"/research/ai-workforce-strategy","f":"A plan for which capabilities the organisation builds, buys, keeps and deliberately lets go. Most are procurement documents.","k":"organisations"},{"t":"question","n":"How do you keep juniors when AI does their work?","st":"P","u":"/research/should-juniors-use-ai","f":"Open, and a serious answer has to include why the business case is hard to make.","k":"organisations"},{"t":"question","n":"What are the missing rungs?","st":"A","u":"/research/missing-rungs","f":"The steps people used to climb to competence, which happen to be the ones most easily automated.","k":"organisations"},{"t":"question","n":"What are missed reps?","st":"A","u":"/research/the-missed-reps-of-ai-seniority","f":"The repetitions that never happened, and the judgement that therefore never formed.","k":"organisations"},{"t":"question","n":"Does AI reduce or increase headcount?","st":"P","u":"/research/ai-workforce-strategy","f":"The measured aggregate effects so far are much smaller than the discussion implies. Firm-level adoption is lower than individual use.","k":"organisations"},{"t":"question","n":"How do you avoid capability debt?","st":"A","u":"/research/capability-debt","f":"By deciding what you are keeping before you decide what you are automating. That ordering is the whole intervention.","k":"organisations","alt":["capability gap","skills debt","losing skills to ai"]},{"t":"question","n":"What is usage theatre?","st":"A","u":"/research/usage-theatre","f":"Activity that looks like transformation and is measured as transformation, without any capability changing.","k":"organisations"},{"t":"question","n":"How do you brief a team when AI drafts everything?","st":"A","u":"/research/what-ai-does-to-a-team","f":"Open. When the first draft is close to free, the brief carries more weight than it used to, because it becomes the main place where thinking is specified rather than assumed. What that means in…","k":"organisations"},{"t":"question","n":"Does AI change team decision-making?","st":"A","u":"/research/what-ai-does-to-a-team","f":"Open, and the group version of the convergence finding is unstudied.","k":"organisations"},{"t":"question","n":"How do you audit an AI-assisted decision?","st":"A","u":"/research/how-do-you-audit-an-ai-assisted-decision","f":"Article 12 requires the log. Article 14 does not require per-decision review, except for biometric identification, which almost everyone gets wrong.","k":"organisations"},{"t":"question","n":"What does good AI adoption look like a year in?","st":"A","u":"/research/the-shape-of-the-organisation-after-ai","f":"The maturity question, and it deserves an answer that is not a maturity model.","k":"organisations"},{"t":"question","n":"How should small organisations approach AI?","st":"A","u":"/research/the-shape-of-the-organisation-after-ai","f":"Open. The research estate over-indexes on large employers, including here.","k":"organisations"},{"t":"question","n":"How should the public sector approach AI?","st":"A","u":"/research/how-will-ai-change-the-public-sector","f":"By separating the administrative case from the deciding case, which the headline figures do not. Both flagship UK numbers are self-reported, and the productivity estimate underneath the policy was…","k":"organisations"},{"t":"question","n":"How do four generations work together on AI?","st":"A","u":"/research/four-generations-and-ai","f":"Each generation built capability under different constraints, and the differences are more useful than the stereotypes.","k":"organisations"},{"t":"question","n":"What is cognitive debt?","st":"A","u":"/research/cognitive-debt-and-capability-debt","f":"MIT's term for reliance on AI replacing the effortful thinking that builds independent capability. The phrase appears four times in a 216-page preprint and its authors never claim to have coined it.","k":"organisations"},{"t":"question","n":"What is the difference between cognitive debt and capability debt?","st":"A","u":"/research/cognitive-debt-and-capability-debt","f":"Cognitive debt is what happens inside one head. Capability debt is what happens to an organisation. The gap between them is where the mechanism sits.","k":"organisations"},{"t":"question","n":"Who coined the term capability debt?","st":"A","u":"/research/cognitive-debt-and-capability-debt","f":"Nobody has claimed it, including Rohde, who defines it in print, and including this research.","k":"organisations"},{"t":"question","n":"What is human capability in the age of AI?","st":"A","u":"/research/human-capability-in-the-age-of-ai","f":"Whether an organisation still holds the judgement, practice and accountability it depends on as machines absorb the work those things were built from.","k":"organisations"},{"t":"question","n":"How do you measure human capability in an organisation?","st":"A","u":"/research/human-capability-in-the-age-of-ai","f":"Six dimensions: judgement, practice, verification, accountability, origination and resilience. A proposed structure, not yet a validated instrument.","k":"organisations"},{"t":"question","n":"What is the difference between AI adoption and human capability?","st":"A","u":"/research/human-capability-in-the-age-of-ai","f":"Adoption measures whether the tools are used. Capability asks what happened to the thinking. Both can move at once, in opposite directions.","k":"organisations"},{"t":"question","n":"Is AI hype or doom?","st":"A","u":"/research/neither-ai-hype-nor-doom","f":"Neither, and the two positions need each other more than either admits.","k":"organisations"},{"t":"question","n":"When should an organisation reverse or constrain an AI deployment?","st":"A","u":"/research/deployment-is-not-a-ratchet","f":"When the assumptions under which it was approved stop holding. NIST names five conditions and treats the thresholds as continual monitoring rather than a one-off gate.","k":"organisations"},{"t":"question","n":"What does responsible de-automation look like?","st":"P","u":"/research/the-invisible-work-of-oversight","f":"Partial. Restoring responsibility with the information, practice, staffing and authority to exercise it. The principle is established in human factors; no dose is.","k":"organisations"},{"t":"question","n":"How quickly can an organisation recover unaided capability after a system fails?","st":"P","u":"/research/deployment-is-not-a-ratchet","f":"Partial, and slower than expected. Relearning beats learning for individuals at every interval tested, never on professional judgement, and never on a routine several people ran together.","k":"organisations"},{"t":"question","n":"How often should organisations practise working without AI?","st":"P","u":"/research/deployment-is-not-a-ratchet","f":"Partial. Set by the decay rate rather than the calendar. Aviation is the only industry that schedules it, from its own accident history rather than anything generalisable.","k":"organisations"},{"t":"question","n":"What happens when a vendor changes the model underneath a workflow?","st":"P","u":"/research/preserving-capability-across-vendors","f":"Partial. Behaviour changes without the buyer deciding anything, and informal tuning around the old failure modes silently expires. Tight coupling to a component you cannot inspect.","k":"organisations"},{"t":"question","n":"What is an organisation's minimum viable human capability?","st":"A","u":"/research/what-is-a-capability-audit","f":"Enough to recognise failure, hold critical operations, decide the cases outside the system's competence and recover control. A shape, not a percentage.","k":"organisations"},{"t":"question","n":"How much redundancy should an AI-dependent organisation keep?","st":"P","u":"/research/deployment-is-not-a-ratchet","f":"Partial. Enough to detect failure, hold critical operations and recover. No universal percentage exists, and Perrow warns that another automated safeguard can reduce safety.","k":"organisations"},{"t":"question","n":"What is a capability audit?","st":"A","u":"/research/what-is-a-capability-audit","f":"A proposed SuperSkills method, labelled as one rather than as a standard. What people can still do when the assistance is removed.","k":"organisations"},{"t":"question","n":"How do I know if my organisation is already in capability debt?","st":"P","u":"/research/capability-debt","f":"Five symptoms, none of which appear in output metrics. The diagnostic is what happens when the tool is removed, not how much it is used.","k":"organisations","alt":["capability gap","skills debt","losing skills to ai"]},{"t":"question","n":"Does AI make organisations smaller but more fragile?","st":"A","u":"/research/the-shape-of-the-organisation-after-ai","f":"Open, and the trade nobody prices. Headcount is measured continuously and fragility is measured after.","k":"organisations"},{"t":"question","n":"How do you preserve capability when the work is spread across vendors?","st":"A","u":"/research/preserving-capability-across-vendors","f":"Keep enough to specify the work, judge it, handle exceptions and replace the supplier. Prahalad and Hamel wrote this in 1990 about manufacturing.","k":"organisations"},{"t":"question","n":"Does AI increase organisational sycophancy?","st":"A","u":"/research/what-ai-does-to-a-team","f":"Open, and the organisational version of a measured individual effect. If a model optimises for the reader's approval, a report drafted for a known audience inherits that.","k":"organisations"},{"t":"question","n":"How should AI change our workforce plan?","st":"P","u":"/research/ai-workforce-strategy","f":"Partly answered. The estate's position is that a workforce plan must name which capabilities it intends to keep, not only which roles it intends to fill. SHRM's 2026 survey of more than 1,900 HR…","k":"organisations"},{"t":"question","n":"How many people will we need in three years because of AI?","st":"O","u":"","f":"Open, and nobody can tell you. Three independent measurements find no general headcount effect yet. Any three-year number is a policy choice presented as a forecast. Say so in the room where the…","k":"organisations"},{"t":"question","n":"Which roles should we grow, shrink, redesign or remove?","st":"P","u":"/research/ai-workforce-strategy","f":"Partly answered. The sharpest available lens is Autor and Thompson: automation that removes the less expert tasks raises wages and cuts employment, while removing the expert tasks does the opposite.…","k":"organisations"},{"t":"question","n":"How should AI change our organisation structure?","st":"P","u":"/research/the-shape-of-the-organisation-after-ai","f":"Partly answered on the observed direction, which is pyramids becoming inverted triangles or diamonds with little arriving at the bottom. What is not answered is whether that shape is stable, and the…","k":"organisations"},{"t":"question","n":"Will AI reduce the number of management layers we need?","st":"P","u":"/research/the-shape-of-the-organisation-after-ai","f":"Partly answered, with the limit named: nothing measured exists in this corpus on what AI does to management layers. Anyone confident about this is extrapolating.","k":"organisations"},{"t":"question","n":"Should managers be responsible for more people when they have AI?","st":"O","u":"","f":"Open, and it depends entirely on which managerial function you think AI assists. It helps with routing information and scheduling. It does not help with developing a person, which is the part that…","k":"organisations"},{"t":"question","n":"How should AI change spans of control?","st":"O","u":"","f":"Open, and the same question with a number attached. Widening spans on the strength of AI assumes the binding constraint was administrative rather than developmental.","k":"organisations"},{"t":"question","n":"How should we redesign job descriptions for AI-assisted work?","st":"O","u":"","f":"Open, and mostly done badly by adding 'uses AI tools' to an unchanged description. The real redesign question is which tasks the role must retain in order to remain a role someone can grow through.","k":"organisations"},{"t":"question","n":"How should we redesign our job architecture for AI?","st":"P","u":"/research/the-shape-of-the-organisation-after-ai","f":"Partly answered at the level of a single job. The architecture question is harder: levels and grades encode an assumed progression of difficulty, and if AI removes the lower rungs the levels stop…","k":"organisations"},{"t":"question","n":"Which jobs should we redesign before we consider redundancies?","st":"O","u":"","f":"Open, and posed in the right order, which is unusual. Bainbridge's warning applies: automating the routine parts leaves the human the hardest residue while removing the practice that built the…","k":"organisations"},{"t":"question","n":"When should we redeploy people rather than remove roles?","st":"O","u":"","f":"Open. The economic case is usually framed as retention cost versus severance cost. The capability case is different and rarely made: the person carries the tacit knowledge of how the work was done…","k":"organisations"},{"t":"question","n":"How should we identify employees whose jobs are most likely to change?","st":"O","u":"","f":"Open. Task-level exposure indices exist and are the right unit of analysis, but applying a published index to a specific organisation's actual jobs is a step almost nobody takes, and exposure is not…","k":"organisations"},{"t":"question","n":"How should we consult employees when AI changes their jobs?","st":"O","u":"","f":"Open, and partly a legal question that varies by jurisdiction. The consultation frameworks assume a proposal with a defined end state, and AI-driven change usually does not have one, which makes…","k":"organisations"},{"t":"question","n":"What should we tell employees about our intentions for AI and headcount?","st":"A","u":"/research/what-should-we-tell-employees-about-ai-and-headcount","f":"Two things that bind the leadership rather than the worker: which decisions a machine may make in the organisation's name, and the headcount position declared by domain and dated, including whether…","k":"organisations"},{"t":"question","n":"How transparent should we be about which jobs AI may affect?","st":"O","u":"","f":"Open. Transparency has a cost that is rarely stated: naming a role as exposed can trigger the departure of the people you most need to manage the transition.","k":"organisations"},{"t":"question","n":"How do we maintain trust when employees think AI is being introduced to replace them?","st":"A","u":"/research/what-should-we-tell-employees-about-ai-and-headcount","f":"Not by messaging. Trust tracks whether the declared position and the business case agree: if the paper says augmentation and the case counts headcount, people work out which is real and hide their…","k":"organisations"},{"t":"question","n":"How should we involve employees in deciding how AI changes their work?","st":"O","u":"","f":"Open, and there is a practical argument for it beyond fairness: the people doing the work know which parts were building competence and which were waste, and that distinction is invisible from above.","k":"organisations"},{"t":"question","n":"What is Human Reserved?","st":"A","u":"/research/what-is-human-reserved","f":"Human Reserved is Bill Gates's term for work deliberately set aside for people only, by analogy with nature reserves: places we could develop but choose not to because the loss would be too great.","k":"organisations"},{"t":"question","n":"What are Configurations capacitantes and aliénantes?","st":"A","u":"/research/what-are-capacitating-configurations","f":"Capacitating and alienating configurations are the two outcomes an AI deployment can produce. Where an organisational compromise is reached, the arrangement increases human aptitude and skill. Where…","k":"organisations"},{"t":"question","n":"What is Conflit de rationalité?","st":"O","u":"","f":"A conflict of rationalities is the unresolved disagreement between what an organisation wants from an AI system and what the work actually requires. Whether a compromise is reached decides whether…","k":"organisations"},{"t":"question","n":"What are learning-conducive work environments?","st":"O","u":"","f":"Learning-conducive work environments are workplaces designed so that continuous learning and the use of skills happen through the work itself rather than through separate training.","k":"organisations"},{"t":"question","n":"What is Komplementäre Arbeitsgestaltung?","st":"O","u":"","f":"Complementary work design treats human and machine complementarity as permanent and functional, grounded in structural limits of automation rather than in the current weakness of models.","k":"organisations"},{"t":"question","n":"What is obligation to justify?","st":"O","u":"","f":"The obligation to justify is Cedefop's proposal that employers should have to give reasons for introducing AI into a workplace.","k":"organisations"},{"t":"question","n":"What is distributed de-skilling?","st":"P","u":"/research/what-is-deskilling","f":"Distributed de-skilling is BCG's term for the collective erosion of human skills across an organisation that undermines its intelligence and resilience over time.","k":"organisations"},{"t":"question","n":"What are aI-off zones?","st":"O","u":"","f":"AI-off zones are BCG's term for tasks an organisation deliberately designates as off limits to AI, where originality, ethical judgement or synthesis matter most.","k":"organisations"},{"t":"question","n":"What is capability masking?","st":"P","u":"/research/capability-debt","f":"Capability masking is the appearance that organisational capability has been replaced by AI while dependence on skilled human labour actually remains, which supports hiring restraint while the cost…","k":"organisations","alt":["capability gap","skills debt","losing skills to ai"]},{"t":"question","n":"What is capability audit?","st":"A","u":"/research/what-is-a-capability-audit","f":"A structured assessment of whether an organisation still possesses the human knowledge, judgement and practical ability its operations depend on, tested by removing assistance rather than by…","k":"organisations"},{"t":"question","n":"What is drift versus design?","st":"A","u":"/research/design-versus-drift","f":"Drift is the gradual outsourcing of choice to whatever is smoothest, until decisions that were once made are simply followed. Design is the opposite move: deciding in advance where human judgement…","k":"organisations","alt":["drift vs design","drift or design","ai drift"]},{"t":"question","n":"What is shift left?","st":"O","u":"","f":"Shift left means moving decisions closer to their source, removing the dilution that every handoff introduces.","k":"organisations"},{"t":"question","n":"What happens to organisational knowledge when it goes into a private AI system?","st":"A","u":"/research/what-happens-to-institutional-memory","f":"It becomes more accessible without becoming more retained. Polanyi's point: the capability that produced the documents was never in them.","k":"organisations"},{"t":"question","n":"Who owns the prompts and corrections employees contribute?","st":"O","u":"","f":"Open, and unresolved in law and in practice. Every correction is a person's expertise, encoded.","k":"organisations"},{"t":"question","n":"What should never be put into an AI system?","st":"O","u":"","f":"Open. The confidentiality answer is well covered; the capability answer, what you lose by externalising it, is not.","k":"organisations"},{"t":"question","n":"How should employees consent to workplace AI?","st":"O","u":"","f":"Open, and distinct from disclosure. This is the organisation asking the employee, rather than the other way round.","k":"organisations"},{"t":"question","n":"What happens when an AI system remembers an employee wrongly?","st":"O","u":"","f":"Open. A persistent inaccurate record about a person is a different problem from a wrong answer, and no process handles it.","k":"organisations"},{"t":"question","n":"Do graduates arrive less capable than they used to?","st":"P","u":"/research/how-do-juniors-become-senior","f":"Open on the outcome and asserted far more confidently than the evidence allows, since no cohort has been followed from a first job into a senior role with the tool present throughout. What is…","k":"talent"},{"t":"question","n":"Is AI deskilling us or rescaling what counts as a valuable skill?","st":"A","u":"/research/is-deskilling-real","f":"Both are measured, in different people. The strong claim that there is no deskilling fails against the Polish colonoscopy result; the rescaling claim survives as a question about which tasks were…","k":"talent","alt":["deskilling","skill loss","are we getting worse at our jobs"]},{"t":"question","n":"Which years of practice matter most for building judgement?","st":"P","u":"/research/missing-rungs","f":"The ones with real consequences and real feedback. Those are the rungs being automated first.","k":"talent"},{"t":"question","n":"How do you tell a capable graduate from a well-tooled one?","st":"P","u":"/research/synthetic-seniority","f":"Not from the output, which is the whole problem. It takes a task the tools cannot do and somebody senior enough to judge the answer.","k":"talent"},{"t":"question","n":"Should juniors be allowed AI from day one?","st":"A","u":"/research/should-juniors-use-ai","f":"The evidence supports a sequenced answer rather than a yes or no, and it puts most of the burden on whoever manages them.","k":"talent"},{"t":"question","n":"Is the graduate job market changing because of AI?","st":"P","u":"/research/will-ai-replace-entry-level-jobs","f":"The exposure signals are real across several countries. The causal evidence is not there yet, and this page says so.","k":"talent","alt":["graduate jobs","junior jobs","entry level roles"]},{"t":"question","n":"What is synthetic seniority?","st":"A","u":"/research/synthetic-seniority","f":"The gap between output that looks senior and the judgement that normally produces it.","k":"talent"},{"t":"question","n":"How do you develop juniors now?","st":"A","u":"/research/should-juniors-use-ai","f":"By deciding which repetitions are development rather than cost. Most organisations have never made that distinction explicitly.","k":"talent"},{"t":"question","n":"Do apprenticeships still work?","st":"A","u":"/research/do-apprenticeships-still-work","f":"Open, and the German and Swiss models deserve better than a passing mention.","k":"talent"},{"t":"question","n":"Who supervises work they cannot do themselves?","st":"A","u":"/research/who-supervises-work-they-cannot-do","f":"Supervision has become approval, and no management system in common use can tell the difference.","k":"talent"},{"t":"question","n":"Does AI widen or narrow the gap between best and worst?","st":"P","u":"/research/human-skills-in-the-age-of-ai","f":"It narrows performance gaps in several studies. Whether it narrows capability gaps is a different question and not the same finding.","k":"talent"},{"t":"question","n":"What does a career look like without a junior tier?","st":"P","u":"/research/do-apprenticeships-still-work","f":"Partly visible now: the Big Four cut graduate intake and cut partner promotions to a five-year low in the same period. Compression at both ends rather than a base that simply disappears.","k":"talent"},{"t":"question","n":"How do you assess capability rather than output?","st":"A","u":"/research/how-do-you-assess-capability-rather-than-output","f":"Structure roughly doubles predictive validity, observation needs about eight repetitions, and the .54 everyone quotes has been revised to .33.","k":"talent"},{"t":"question","n":"Are we losing tacit knowledge?","st":"A","u":"/research/what-is-tacit-knowledge","f":"Polanyi's concept. The form of knowledge most exposed by AI, because it is least likely to be in any training corpus.","k":"talent"},{"t":"question","n":"What is deliberate practice?","st":"A","u":"/research/what-is-deliberate-practice","f":"The four conditions Ericsson specified, the re-analysis that cut the claim down, and Epstein on why it generalises badly outside kind environments.","k":"talent"},{"t":"question","n":"Does experience still count?","st":"P","u":"/research/staying-valuable-in-the-age-of-ai","f":"Experience that produced judgement counts more. Experience that produced familiarity counts less than it used to.","k":"talent"},{"t":"question","n":"How do you keep expertise in an organisation?","st":"A","u":"/research/how-do-you-keep-expertise-in-an-organisation","f":"Experts practising difficult work, getting feedback and teaching. Documentation holds the explicit part; Polanyi's point is that the rest was never written down.","k":"talent"},{"t":"question","n":"What happens to the apprenticeship model?","st":"A","u":"/research/do-apprenticeships-still-work","f":"Open, and the closest thing to an official answer is Singapore's: IMDA warns that entry level tasks typically serve as the training ground for new staff. What replaces them is undecided.","k":"talent"},{"t":"question","n":"Should juniors use AI at all?","st":"A","u":"/research/should-juniors-use-ai","f":"Open, and the answer is almost certainly not a simple yes or no.","k":"talent"},{"t":"question","n":"Why are there empty apprenticeship places and unplaced applicants at once?","st":"A","u":"/research/do-apprenticeships-still-work","f":"Germany had 84,400 young people without a place in 2025, the most since 2010, and 54,400 places unfilled in the same year. Matching, not volume, is the binding constraint.","k":"talent"},{"t":"question","n":"Did the UK cut degree apprenticeship funding?","st":"A","u":"/research/do-apprenticeships-still-work","f":"Only at level 7, for those aged 22 and over, from January 2026. Level 6, the undergraduate degree apprenticeship, was untouched and has grown from about 6,400 starts to about 26,800.","k":"talent"},{"t":"question","n":"Are internships declining?","st":"A","u":"/research/do-apprenticeships-still-work","f":"Nobody knows, because no official body measures internship volume anywhere. A job board reports contraction and an employer association forecasts 3.9 per cent growth, and neither is a statistic.","k":"talent"},{"t":"question","n":"Have consultancies inverted the pyramid?","st":"A","u":"/research/do-apprenticeships-still-work","f":"No, on the available evidence. Graduate intake fell across the Big Four and partner promotions fell to a five-year low at the same time. That is compression, and no firm publishes the grade data that…","k":"talent"},{"t":"question","n":"How do you preserve apprenticeship pathways when AI removes the junior work?","st":"P","u":"/research/do-apprenticeships-still-work","f":"Germany shows funding and volume are not the constraint: record unplaced applicants and 54,400 empty places in the same year.","k":"talent"},{"t":"question","n":"How do I run a team where AI drafts everything and the juniors never learn?","st":"P","u":"/research/should-juniors-use-ai","f":"Four rules for the junior and four for whoever manages them, because the first set alone puts the burden on the least powerful person.","k":"talent"},{"t":"question","n":"Does the organisation face a skill cliff when its experts retire?","st":"O","u":"","f":"Open. If judgement passed through shared work and the shared work is automated, the transfer channel closed before anyone retired.","k":"talent"},{"t":"question","n":"Are firms cutting junior jobs, or just not opening them?","st":"A","u":"/research/missing-rungs","f":"Not opening them. At firms adopting generative AI, junior separations FELL. Junior hiring fell about four times faster. Nobody is pushed off the ladder; the lower rungs stop being built. Unemployment…","k":"talent"},{"t":"question","n":"Which junior tasks disappear first?","st":"A","u":"/research/missing-rungs","f":"The exposed ones, measurable at task level rather than inferred. Mapping 355,000 job postings onto standardised tasks shows exposed tasks being written out of junior job descriptions specifically,…","k":"talent"},{"t":"question","n":"Do the same firms cut senior roles too?","st":"A","u":"/research/will-ai-replace-entry-level-jobs","f":"No, and that asymmetry is the finding. In firms where junior employment fell about 9 per cent against non-adopters, senior employment showed no comparable break and in exposed occupations kept rising.","k":"talent","alt":["graduate jobs","junior jobs","entry level roles"]},{"t":"question","n":"Are juniors promoted faster now that AI does the junior work?","st":"P","u":"/research/how-do-juniors-become-senior","f":"Slightly. Promotion rates rose by 0.033 percentage points against a base of 0.82 per cent, statistically significant and practically tiny. Fewer juniors, marginally quicker promotion, and nothing yet…","k":"talent"},{"t":"question","n":"How would you actually prove AI is removing entry-level work?","st":"A","u":"/research/missing-rungs","f":"Compare firms that adopted against firms that did not, inside the same industry and period, splitting by seniority. National hiring statistics cannot do this because everything moves at once. It took…","k":"talent"},{"t":"question","n":"If AI raises productivity, why would junior hiring fall rather than rise?","st":"P","u":"/research/missing-rungs","f":"Because a productivity gain only raises demand for junior labour if tasks substitute for each other. Where they complement, the same gain cuts junior demand. That is the theoretical answer and it is…","k":"talent"},{"t":"question","n":"Are employers just relabelling junior jobs rather than cutting them?","st":"A","u":"/research/synthetic-seniority","f":"Tested and rejected. Job-title keyword distributions show no differential shift at adopting firms around the ChatGPT launch, so the fall is a real change in who is hired rather than a change in what…","k":"talent"},{"t":"question","n":"How should AI change succession planning?","st":"O","u":"","f":"Open, and the most consequential unasked question in this territory. Succession assumes a pipeline that is being switched off at the bottom while the plans above it are unchanged.","k":"talent"},{"t":"question","n":"Where will our next senior leaders come from if junior work disappears?","st":"A","u":"/research/how-do-juniors-become-senior","f":"The missing rungs argument seen from inside the org chart. An organisation with no junior intake has no bench, and in ten years nobody to promote. Answered 6 September 2026: junior work was a…","k":"talent"},{"t":"question","n":"Is our workforce being deskilled, or is AI just changing which skills we value?","st":"A","u":"/research/is-deskilling-real","f":"Both have been measured, in different populations, so the question has to be asked role by role. Autor and Thompson give the test: automation that removed the less expert tasks raised wages and cut…","k":"talent","alt":["deskilling","skill loss","are we getting worse at our jobs"]},{"t":"question","n":"Which roles are becoming succession risks because AI removed the development path?","st":"P","u":"/research/missing-rungs","f":"Partly answered on the mechanism. The organisational diagnostic does not exist: identifying which specific roles no longer have a route into them would be a genuinely new piece of work, doable with…","k":"talent"},{"t":"question","n":"How do we identify the capabilities we will need before we need them?","st":"O","u":"","f":"Open, and strategic workforce planning has always been weak here. What AI changes is the lead time: capability that used to accumulate as a by-product of doing the work now has to be deliberately…","k":"talent"},{"t":"question","n":"Should we hire AI capability or develop it internally?","st":"O","u":"","f":"Open, and the framing hides the real choice. Hiring gets the tooling skill, which is not scarce and not durable. Developing gets domain judgement applied to the tools, which is scarce and slow. Most…","k":"talent"},{"t":"question","n":"When is reskilling actually worth the investment?","st":"P","u":"/research/why-reskilling-programmes-mostly-fail","f":"Partly answered on why programmes fail. The investment question is unanswered and harder, because it requires an honest view of how long the reskilled capability stays valuable.","k":"talent"},{"t":"question","n":"How do we know whether reskilling worked?","st":"P","u":"/research/why-reskilling-programmes-mostly-fail","f":"Partly answered. Completion rates measure attendance. The only real test is whether people can do something they could not do before, measured unaided, which almost no programme does because the…","k":"talent"},{"t":"question","n":"What should we do with employees whose roles disappear faster than they can retrain?","st":"O","u":"","f":"Open, and the question most likely to be answered by default rather than by decision. Skill acquisition has a floor on how fast it can go, and no amount of programme design removes it.","k":"talent"},{"t":"question","n":"What is legitimate peripheral participation?","st":"A","u":"/research/how-do-you-keep-expertise-in-an-organisation","f":"Legitimate peripheral participation is how newcomers acquire competence: by doing real but peripheral work alongside practitioners, moving from the edge of a community of practice towards its centre.","k":"talent"},{"t":"question","n":"What are seniorised entry-level roles?","st":"P","u":"/research/will-ai-replace-entry-level-jobs","f":"Seniorised entry-level roles are junior jobs that now demand senior human skills. PwC found entry-level roles most exposed to AI are seven times more likely to require leadership, creativity or…","k":"talent","alt":["graduate jobs","junior jobs","entry level roles"]},{"t":"question","n":"What is the AI employment gap?","st":"A","u":"/research/what-is-the-ai-employment-gap","f":"The AI employment gap is the Stanford Digital Economy Lab's finding that employment of workers aged 22 to 25 in AI-exposed occupations sits 19 per cent below where it would have been had it tracked…","k":"talent"},{"t":"question","n":"What is borrowed competence?","st":"O","u":"","f":"Borrowed competence is McKinsey's term for capability that appears in the output but disappears when the tool is withdrawn.","k":"talent"},{"t":"question","n":"What is the answer-key model?","st":"O","u":"","f":"The answer-key model is McKinsey's proposed training pattern in which the employee attempts the work first, the AI grades the attempt, and a manager reviews it with them.","k":"talent"},{"t":"question","n":"What is learning by verifying?","st":"O","u":"","f":"Learning by verifying is Bain's claim that juniors learn by reviewing, stress-testing and catching errors in AI output, and that the repetitions per hour go up rather than down.","k":"talent"},{"t":"question","n":"What is oversight readiness?","st":"A","u":"/research/what-is-oversight-readiness","f":"Oversight readiness is Google DeepMind's term for whether a future workforce will be able to judge the AI work it is nominally supervising, given that juniors are being deprived of the experience…","k":"talent"},{"t":"question","n":"What is curriculum-aware task routing?","st":"O","u":"","f":"Curriculum-aware task routing is Google DeepMind's proposal for systems that track a junior's skill progression and deliberately allocate tasks at the edge of their expanding competence, including…","k":"talent"},{"t":"question","n":"What is the two-tier future?","st":"P","u":"/research/missing-rungs","f":"The two-tier future is the outcome financial services leaders fear most: lower-skilled people handling tasks AI is not set up for, a smaller group of experts training the system and handling…","k":"talent"},{"t":"question","n":"My teenager uses AI for everything. Is that different from my team doing it?","st":"P","u":"/research/using-ai-without-dependency","f":"Less different than it feels. The mechanism is the same: which repetitions are being skipped, and whether anyone would notice.","k":"everyday"},{"t":"question","n":"I set an AI policy at work. What would the same policy look like at home?","st":"O","u":"","f":"Open, and worth thinking through. Most workplace policies are about disclosure and risk, and almost none about which practice to protect.","k":"everyday"},{"t":"question","n":"My company encourages AI and my child’s school restricts it. Who is right?","st":"P","u":"/research/official-guidance-on-ai-in-education","f":"Both, for their own settings. The school is protecting capability being built; the company is buying output already paid for.","k":"everyday"},{"t":"question","n":"How do I talk to my children about this without frightening them?","st":"O","u":"","f":"Open. The honest framing is that the tools are useful and the practice is worth keeping, which is harder to say than either extreme.","k":"everyday"},{"t":"question","n":"What would I want my own child’s employer to do about this?","st":"P","u":"/research/should-juniors-use-ai","f":"A useful test for any policy you are about to approve. Most people set a stricter bar for their own children than for other people’s.","k":"everyday"},{"t":"question","n":"Is screen time the same argument as AI use?","st":"A","u":"/research/is-screen-time-the-same-argument-as-ai-use","f":"No, and the difference is a variable rather than a tone. Screen-time work measures exposure with an instrument that correlates with logged use at r = 0.38, and across 355,358 adolescents finds at…","k":"everyday"},{"t":"question","n":"Are young people actually using AI differently?","st":"P","u":"/research/is-screen-time-the-same-argument-as-ai-use","f":"Partly, and the answered half is negative. The measurement objection is settled: self-reported usage surveys are the wrong instrument, and the researchers who built the screen-time literature say so…","k":"everyday"},{"t":"question","n":"Is it safe to use AI for therapy or advice?","st":"A","u":"/research/is-it-safe-to-use-ai-for-therapy","f":"The trial that shows it works had humans reading every message, and the regulator has authorised none of it.","k":"everyday"},{"t":"question","n":"Should I let AI make personal decisions for me?","st":"A","u":"/research/should-i-let-ai-make-personal-decisions-for-me","f":"Advice from a model with no settled view still moves yours, disclosure does not reduce the effect, and 80 per cent believe they decided unaided.","k":"everyday"},{"t":"question","n":"Should I use AI to write personal messages?","st":"A","u":"/research/should-i-use-ai-to-write-personal-messages","f":"The interpersonal penalty attaches to being suspected, not to using it. Suspicion tracks actual use at r=0.22, so the toll is levied close to at random.","k":"everyday"},{"t":"question","n":"Does using AI change my creativity?","st":"P","u":"/research/does-ai-make-everyone-think-alike","f":"At population level it narrows the range. At individual level the evidence says your output improves, and that is what makes it hard to notice.","k":"everyday"},{"t":"question","n":"Should I use AI for relationship advice?","st":"O","u":"","f":"Open, and it needs care: there are cases where it genuinely helps and cases where it substitutes for a person who should be told.","k":"everyday"},{"t":"question","n":"Should AI remember everything about me?","st":"A","u":"/research/should-ai-remember-everything-about-me","f":"Memory as convenience versus memory as leverage, which is a different question from privacy and less well covered.","k":"everyday"},{"t":"question","n":"Do I still need to remember things?","st":"A","u":"/research/do-i-still-need-to-remember-things","f":"Offloading is not uniformly a loss: saving one file improved memory for the next. What memory is still for is noticing that an answer is wrong.","k":"everyday"},{"t":"question","n":"Should I let AI summarise everything I read?","st":"A","u":"/research/should-i-let-ai-summarise-everything-i-read","f":"Seven experiments, and the one that held the facts identical still found shallower learning and output three times more similar to everyone else's.","k":"everyday"},{"t":"question","n":"How much should teenagers use AI?","st":"A","u":"/research/how-much-should-teenagers-use-ai","f":"The frequency question is the wrong one. One field experiment: marks up 48 per cent with the tool, 17 per cent below the control once it was removed.","k":"everyday"},{"t":"question","n":"Should I let AI write for me at all?","st":"P","u":"/research/human-at-the-start","f":"It depends entirely on whether being the author is part of what the writing is for.","k":"everyday"},{"t":"question","n":"Is it bad to talk to AI when lonely?","st":"O","u":"","f":"WATCH. Genuinely difficult, and it deserves evidence rather than instinct before anything is published.","k":"everyday"},{"t":"question","n":"Should I use AI to help with parenting decisions?","st":"P","u":"/research/how-do-i-raise-a-child-who-thinks-for-themselves","f":"Partly answered. The learning-science half is covered; no research has looked at parenting decisions specifically, and the page says so rather than inventing an answer.","k":"everyday"},{"t":"question","n":"Does AI change how children develop?","st":"O","u":"","f":"WATCH. The evidence base is thin, and this should not be published until it is not.","k":"everyday"},{"t":"question","n":"Should I trust AI health information?","st":"P","u":"/research/how-do-i-know-when-ai-is-wrong","f":"Chatbot responses have been rated higher quality and more empathetic than physicians in one study. Neither of those is the same as correct.","k":"everyday"},{"t":"question","n":"What do I lose if AI does the boring bits?","st":"P","u":"/research/the-unclaimed-hour","f":"Sometimes nothing. Sometimes the boring bits were where the pattern recognition was being built.","k":"everyday"},{"t":"question","n":"How do I raise a child who thinks for themselves?","st":"A","u":"/research/how-do-i-raise-a-child-who-thinks-for-themselves","f":"By protecting the part of the day where they are stuck. The one experiment that withdrew the tutor found unrestricted users 17 per cent below a never-used control, and the guardrailed group largely…","k":"everyday"},{"t":"question","n":"What is aI hallucination?","st":"A","u":"/research/what-is-an-ai-hallucination","f":"Generated content presented as factual that is not supported by the model's training data, the provided context or reality.","k":"everyday"},{"t":"question","n":"What is AI slop?","st":"O","u":"","f":"AI slop is fast, plausible output that does not meet the standard.","k":"everyday"},{"t":"question","n":"What do we actually know about AI and human capability?","st":"A","u":"/research/what-we-know-about-ai-and-human-capability","f":"Nineteen claims in three bands: strong evidence, emerging evidence, and what remains unknown. Reviewed quarterly.","k":"evidence"},{"t":"question","n":"Where is the evidence on AI and human capability?","st":"A","u":"/research/evidence","f":"Fifty-six graded studies, each with its method, its finding, and what it does not support. Every entry separately citable.","k":"evidence"},{"t":"question","n":"Which books and papers on AI and human capability actually matter?","st":"A","u":"/research/essential-works","f":"Eighty-four works, classified by role in the field rather than by how much they agree with anything here.","k":"evidence"},{"t":"question","n":"Were past predictions about AI and work correct?","st":"A","u":"/research/predictions","f":"Dated positions, kept whether or not they held up. Removing the misses would defeat the purpose.","k":"evidence"},{"t":"question","n":"When did each idea about AI and human capability first appear?","st":"A","u":"/research/timeline","f":"Weekly since January 2017, five years and ten months before ChatGPT. Curation first, thesis later, described honestly.","k":"evidence"},{"t":"question","n":"Who is researching AI and human capability?","st":"A","u":"/research/ai-people","f":"The people whose work this research draws on, and the reading that goes with them.","k":"evidence"},{"t":"question","n":"What do the key terms about AI and human capability mean?","st":"A","u":"/research/ai-glossary","f":"The terms used here, including which are established concepts and which are coinages from this work.","k":"evidence"},{"t":"question","n":"What is the SuperSkills argument about AI and human capability?","st":"A","u":"/research/the-superskills-thesis","f":"The whole case in one place, with the parts that are contested marked as contested.","k":"evidence","alt":["superskills argument","superskills book argument","super skills thesis"]},{"t":"question","n":"How strong is the evidence that AI weakens human capability?","st":"A","u":"/research/what-we-know-about-ai-and-human-capability","f":"Weaker than most commentary implies in some places and stronger in others. The bands are there so the difference is visible.","k":"evidence"},{"t":"question","n":"What does the evidence on AI and human capability not show?","st":"A","u":"/research/evidence","f":"Every entry carries a what-it-does-not-support field, which is the reason the base exists.","k":"evidence"},{"t":"question","n":"What is happening outside the US and UK?","st":"A","u":"/research/ai-and-work-by-country","f":"The first country page is Japan: the strongest economic case for adoption anywhere, and adoption running at roughly 18 per cent.","k":"evidence"},{"t":"question","n":"How does this research compare with the WEF Future of Jobs report?","st":"P","u":"/research/the-best-writing-on-ai","f":"Partly. The chronology sets the WEF numbers in context as employer expectation rather than measurement, which is almost never said.","k":"evidence"},{"t":"question","n":"How does this research compare with the OECD's work on skills?","st":"O","u":"","f":"Open. The OECD's adult skills work measures capability as a stock at a point in time. This research asks what sustained AI use does to that stock, which is a different question and has not been…","k":"evidence"},{"t":"question","n":"Can capability loss from AI actually be measured?","st":"A","u":"/research/capability-debt","f":"Three times now, and only by removing the tool: six percentage points of endoscopist detection, 77 per cent against 39 on a no-AI task, 17 per cent below a control once access was withdrawn. No…","k":"evidence","alt":["capability gap","skills debt","losing skills to ai"]},{"t":"question","n":"What evidence would change your mind about AI and capability?","st":"P","u":"/research/what-we-know-about-ai-and-human-capability","f":"Stated per claim on the state of the evidence, and maintained as a standing register rather than offered when challenged.","k":"evidence"},{"t":"question","n":"What should I read about AI and human capability?","st":"A","u":"/research/ai-reading-list","f":"A four-quadrant shortlist for a leadership team, chosen to close gaps rather than to cover the field. Reading is only useful here if it produces shared language.","k":"evidence"},{"t":"question","n":"Which AI reports are worth reading?","st":"A","u":"/research/the-ai-reports-worth-reading","f":"One named report for each kind of reader, with the reasoning and the limits. Choosing is the value; listing is the easy half.","k":"evidence"},{"t":"question","n":"Are the most quoted AI statistics true?","st":"A","u":"/research/the-most-quoted-ai-statistics-checked","f":"Nine numbers repeated constantly, each traced to what its source actually says. Four are misquoted and two cannot be traced at all.","k":"evidence"},{"t":"question","n":"Does automation always cause deskilling?","st":"A","u":"/research/capability-debt","f":"No. Robots in Japanese nursing homes raised employment, improved retention and cut physical restraint and pressure ulcers. The condition was an acute labour shortage, so the machine filled vacancies…","k":"evidence","alt":["capability gap","skills debt","losing skills to ai"]},{"t":"question","n":"Does the EU AI Act cover AI in schools?","st":"A","u":"/research/official-guidance-on-ai-in-education","f":"Yes. Annex III paragraph 3 makes AI that decides admission, evaluates learning outcomes or monitors test behaviour high-risk, with the full obligations applying from 2 August 2026.","k":"evidence"},{"t":"question","n":"Is official guidance on AI in education based on evidence?","st":"A","u":"/research/official-guidance-on-ai-in-education","f":"Mostly not. Of thirteen documents read at source, three present any original data. The UK Department for Education says on the face of its own policy that it has limited evidence.","k":"evidence"},{"t":"question","n":"Do US states regulate AI in schools?","st":"A","u":"/research/official-guidance-on-ai-in-education","f":"Three published guidance and only one has statutory force. California says compliance is not mandatory; Ohio requires every district to adopt a policy by 1 July 2026 under Revised Code 3301.24.","k":"evidence"},{"t":"question","n":"Can an organisation improve AI productivity while losing the capability it needs to survive without AI?","st":"P","u":"/research/capability-debt","f":"The central tension of this research, and yes: measured in endoscopists whose assisted performance rose while their unassisted detection fell six percentage points.","k":"evidence","alt":["capability gap","skills debt","losing skills to ai"]},{"t":"question","n":"How would you know whether AI caused the capability loss?","st":"P","u":"/research/capability-debt","f":"Only by removing the tool. Every measurement here that found a loss found it that way, and no observational design has separated it from ordinary decline.","k":"evidence","alt":["capability gap","skills debt","losing skills to ai"]},{"t":"question","n":"What is the lag between losing practice and losing performance?","st":"A","u":"/research/how-fast-do-skills-decay","f":"Substantial degradation within the first year on the CPR evidence, six percentage points in months for endoscopists using AI, and d = -1.4 beyond 365 days of disuse. Nobody has measured it for…","k":"evidence"},{"t":"question","n":"Can capability be measured without testing people unaided?","st":"A","u":"/research/what-is-a-capability-audit","f":"Not directly. Confidence fails because the reporting faculty is the impaired one; output fails because assisted output is what stays high.","k":"evidence"},{"t":"question","n":"What would falsify the capability debt argument?","st":"P","u":"/research/what-we-know-about-ai-and-human-capability","f":"Unassisted performance holding steady under sustained AI use, or recovery on removal that is fast and complete. Stated per claim rather than offered when challenged.","k":"evidence"},{"t":"question","n":"How do you tell AI-related deskilling from ordinary organisational decline?","st":"O","u":"","f":"Open, and unanswered anywhere. Without it, capability debt is a story that fits any struggling organisation.","k":"evidence"},{"t":"question","n":"What would a credible control group for AI adoption look like?","st":"O","u":"","f":"Open. Japan is the closest thing available, at roughly 18 per cent adoption in an economy with the strongest case for it.","k":"evidence"},{"t":"question","n":"How much of the AI and work evidence comes from high-income English-speaking countries?","st":"P","u":"/research/ai-and-work-by-country","f":"Most of it. The estate now carries verified national data from Japan, Korea, Singapore, France, Germany and the Gulf partly to correct for that.","k":"evidence"},{"t":"question","n":"Do most people expect AI to cost jobs?","st":"A","u":"/research/ai-and-work-by-country","f":"In 34 of the 37 countries Pew surveyed in 2026, majorities expect AI to mean fewer jobs, and the expectation is strongest in the richest countries. It is expectation, not outcome: the national…","k":"evidence"},{"t":"question","n":"Which AI studies measure output but not learning?","st":"P","u":"/research/what-we-know-about-ai-and-human-capability","f":"Most of them. That single gap explains more contradictory headlines than anything else on this map.","k":"evidence"},{"t":"question","n":"How often are AI and work studies independently replicated?","st":"O","u":"","f":"Open, and rarely. Several of the most quoted findings rest on one design that nobody has repeated.","k":"evidence"},{"t":"question","n":"What evidence is missing because nobody is funding it?","st":"O","u":"","f":"Open. Longitudinal capability measurement has no obvious funder: vendors will not pay for it and employers do not want the answer on record.","k":"evidence"},{"t":"question","n":"How do we regulate AI decision-making?","st":"O","u":"","f":"Open here. The EU AI Act tiers obligations by risk and the enforcement questions are only now arriving, so for the next two years nobody can say with confidence how it will be applied in practice.","k":"evidence"},{"t":"question","n":"What is the best writing on AI?","st":"A","u":"/research/the-best-writing-on-ai","f":"A chronology from March 2023 to August 2026, because the order is the argument.","k":"evidence"},{"t":"question","n":"What is the method behind the SuperSkills research?","st":"A","u":"/research/how-this-research-works","f":"Source selection, evidence tiers, where AI is used and where it is not allowed to decide, attribution rules, and what commercial interests exist.","k":"evidence"},{"t":"question","n":"What has the SuperSkills research got wrong?","st":"A","u":"/research/corrections","f":"A public log with the original claim, what was found, and what changed. Nothing is removed.","k":"evidence"},{"t":"question","n":"Why is the AI debate so much louder than the evidence?","st":"P","u":"/research/the-best-writing-on-ai","f":"A viral scenario moved markets in February 2026 while hiring data pointed the other way. Alarm is tracking narrative quality, not evidence.","k":"evidence"},{"t":"question","n":"What is happening with AI and work in Asia?","st":"A","u":"/research/ai-and-work-in-asia","f":"Singapore names deskilling in national guidance. Hong Kong has no AI strategy and no AI statistic. Japan and Korea show adoption far below the discourse.","k":"evidence"},{"t":"question","n":"Does any government address AI deskilling in policy?","st":"A","u":"/research/ai-and-work-in-asia","f":"Singapore does. IMDA's agentic framework names the loss of entry-level training grounds and requires training so people retain foundational skills.","k":"evidence"},{"t":"question","n":"What is happening with AI and work in the Gulf?","st":"A","u":"/research/ai-and-work-in-the-gulf","f":"Of the Gulf instruments reviewed, Saudi Arabia's is the one that names over-reliance on AI. None of the three requires that a human overseer be able to do the work.","k":"evidence"},{"t":"question","n":"Do Gulf AI frameworks address deskilling?","st":"A","u":"/research/ai-and-work-in-the-gulf","f":"No. The word appears in no verified document from Saudi Arabia, the UAE or Qatar.","k":"evidence"},{"t":"question","n":"What is happening with AI and work in Japan?","st":"A","u":"/research/ai-and-work-in-japan","f":"The strongest economic case for adoption anywhere, and adoption at roughly 18 per cent. The control condition this debate never had.","k":"evidence"},{"t":"question","n":"What is happening with AI and work in India?","st":"O","u":"","f":"Services-export exposure is the whole argument, and roughly five million graduates enter the labour market each year. Next in the international cluster.","k":"evidence"},{"t":"question","n":"What is happening with AI and work in Germany?","st":"O","u":"","f":"Works councils have been negotiating this in enterprise agreements while the Anglosphere wrote essays about it. Next in the international cluster.","k":"evidence"},{"t":"question","n":"What obligations do companies have to preserve human capability?","st":"O","u":"","f":"Nobody is asking this and it is the natural end point of the whole argument.","k":"evidence"},{"t":"question","n":"Should governments care about AI-driven deskilling?","st":"O","u":"","f":"Policy territory, largely unoccupied. Open.","k":"evidence"},{"t":"question","n":"What can AI still not do reliably?","st":"P","u":"/research/what-is-the-jagged-frontier","f":"The boundary is jagged rather than smooth, and invisible from the output.","k":"evidence","alt":["jagged frontier explained","jagged edge of ai"]},{"t":"question","n":"How much of the workforce is actually exposed to AI?","st":"P","u":"/research/ai-and-work-by-country","f":"Partly answered, and the honest summary is that the estimates are wide and the definitions differ. The UK government's January 2026 assessment splits the workforce into 35 per cent high exposure with…","k":"evidence"},{"t":"question","n":"Who else is mapping this territory, and what do they cover?","st":"P","u":"/research/the-best-writing-on-ai","f":"Partly answered. The neighbouring work divides into research projects answering a defined question, institutional thought leadership, short board question sets and government evidence assessments.…","k":"evidence"},{"t":"question","n":"Where is this research weaker than the alternatives?","st":"O","u":"","f":"Open, and deliberately listed rather than answered in passing. The honest short version: no primary task-level data, no cross-country survey instrument, no access to a teaching population, and no…","k":"evidence"},{"t":"question","n":"What is the capability-reliability gap?","st":"O","u":"","f":"The capability-reliability gap is the distance between what a model can do once and what it can do dependably. It is the main barrier to agents automating real work.","k":"evidence"},{"t":"question","n":"What is the jagged technological frontier?","st":"A","u":"/research/what-is-the-jagged-frontier","f":"The irregular boundary between tasks an AI system performs well and tasks it performs badly, where the two can be almost indistinguishable in apparent difficulty and the system gives no signal of…","k":"evidence","alt":["jagged frontier explained","jagged edge of ai"]},{"t":"question","n":"What did the METR study actually find?","st":"A","u":"/research/what-is-the-metr-study","f":"Sixteen developers were 19 per cent slower with AI while forecasting a 24 per cent speed-up, and METR marked the result out of date on 24 February 2026. The perception gap survives the withdrawal;…","k":"evidence"},{"t":"question","n":"Is the 19 per cent AI slowdown figure still valid?","st":"A","u":"/research/what-is-the-metr-study","f":"No, and METR say so on their own page. It measures early-2025 tooling in one setting. Quote the 40 percentage point gap between what developers felt and what the clock showed instead.","k":"evidence"},{"t":"question","n":"What changes when AI acts instead of answering?","st":"A","u":"/research/ai-agents-and-human-judgement","f":"Autonomy moves the human decision earlier, from approving output to setting the boundary. Most organisations have not moved with it.","k":"agents"},{"t":"question","n":"Should I let an AI agent act on my behalf?","st":"A","u":"/research/should-i-let-an-ai-agent-act-on-my-behalf","f":"Not a trust question. What is the worst thing it can do before a human sees it, and can you live with that?","k":"agents"},{"t":"question","n":"Who is responsible when an agent makes a mistake?","st":"A","u":"/research/how-should-ai-decision-rights-be-allocated","f":"Whoever deployed it. The regulation separates provider from deployer and obliges both. Chains complicate the tracing, not the principle.","k":"agents"},{"t":"question","n":"How do you supervise something that does not wait for you?","st":"P","u":"/research/should-i-let-an-ai-agent-act-on-my-behalf","f":"Every oversight model assumes a pause for review. Agents remove the pause, so the human has to move to the boundary instead.","k":"agents"},{"t":"question","n":"Should AI attend my meetings?","st":"A","u":"/research/should-ai-attend-my-meetings","f":"Depends whether the meeting produces a record or produces understanding. Most organisations apply one policy to both.","k":"agents"},{"t":"question","n":"What happens to a team when agents do the coordination?","st":"O","u":"","f":"Coordination is where a lot of tacit knowledge moves between people. Open.","k":"agents"},{"t":"question","n":"Can agents manage other agents?","st":"O","u":"","f":"WATCH. Moving fast, thin evidence, and worth watching before writing.","k":"agents"},{"t":"question","n":"How much autonomy is too much?","st":"P","u":"/research/delegation-boundary-map","f":"The nine-stage map answers this stage by stage, which is more useful than a single threshold.","k":"agents"},{"t":"question","n":"Do agents make the jagged frontier worse?","st":"P","u":"/research/what-is-the-jagged-frontier","f":"Plausibly, because an agent crosses the boundary without a human present to notice. Unstudied.","k":"agents","alt":["jagged frontier explained","jagged edge of ai"]},{"t":"question","n":"What is an AI agent?","st":"O","u":"","f":"Fast-moving definition, and the field uses the word for several different things. Open, on a short review cycle when written.","k":"agents"},{"t":"question","n":"What limits should an AI agent have?","st":"P","u":"/research/how-do-you-design-a-stop-button-people-will-use","f":"Partly. The stopping limit is now answered, including who holds it and what makes it real. What the estate still does not have is the positive side of the question: which actions an agent may take…","k":"agents"},{"t":"question","n":"What is meaningful human control?","st":"P","u":"/research/what-is-meaningful-human-oversight","f":"The oversight page covers the regulatory version. The autonomous-systems literature has a longer and sharper treatment.","k":"agents"},{"t":"question","n":"Can an agent tell when it has reached the edge of its competence?","st":"P","u":"/research/what-is-the-jagged-frontier","f":"The boundary is jagged rather than smooth and invisible from the output, which is the problem for the human and for the agent.","k":"agents","alt":["jagged frontier explained","jagged edge of ai"]},{"t":"question","n":"What happens when one agent delegates to another?","st":"P","u":"/research/how-do-humans-and-agents-divide-work","f":"Partly answered, and only on the failure side. Cemri and colleagues annotated more than 1,600 multi-agent traces and put 36.9 per cent of failure in inter-agent misalignment, with mismatch between…","k":"agents"},{"t":"question","n":"How should an agent communicate uncertainty to a human?","st":"A","u":"/research/how-should-an-ai-agent-communicate-uncertainty","f":"Answered 4 September 2026, and it turns out to be two problems. Models are badly calibrated when asked to state confidence in words, with expected calibration error above 0.37 for four of five…","k":"agents"},{"t":"question","n":"How do you design a stop button people will actually use?","st":"A","u":"/research/how-do-you-design-a-stop-button-people-will-use","f":"Answered 3 September 2026 from outside AI, because the measured evidence is elsewhere. Article 14(4)(e) requires the button and decides nothing about its use. One ICU study annotated 12,671…","k":"agents"},{"t":"question","n":"What capability has to stay in-house when agents run core processes?","st":"P","u":"/research/preserving-capability-across-vendors","f":"Partial. Judging the output, deciding the cases outside competence, intervening with real authority, and running the process another way. Same list as any outsourced function.","k":"agents"},{"t":"question","n":"How does AI handle uncertainty in decision-making?","st":"O","u":"","f":"Open. It expresses uncertainty poorly and confidently, which is the practical problem rather than the theoretical one: the tone of a guess and the tone of a grounded answer are indistinguishable.","k":"agents"},{"t":"question","n":"What is Agent boss?","st":"O","u":"","f":"An agent boss is Microsoft's term for a human manager of one or more AI agents.","k":"agents"},{"t":"question","n":"What is human-agent ratio?","st":"O","u":"","f":"The human-agent ratio is Microsoft's proposed business metric for the balance between human oversight and agent efficiency on a mixed team.","k":"agents"},{"t":"question","n":"What is human at the start?","st":"A","u":"/research/human-at-the-start","f":"Human at the start is the position in which a person sets the problem, the intent and the constraints before any model is asked to generate, so the judgement enters the work before the output exists…","k":"agents"},{"t":"question","n":"What is the Delegation Boundary Map?","st":"A","u":"/research/delegation-boundary-map","f":"The Delegation Boundary Map is a working framework that breaks a piece of work into nine stages, problem definition, intent, context, constraints, delegation, generation and execution, verification,…","k":"agents"},{"t":"question","n":"Should AI remember everything about me?","st":"A","u":"/research/should-ai-remember-everything-about-me","f":"Privacy is the smaller half. The larger half is that a system which remembers everything gets better at telling you what you want to hear.","k":"companions"},{"t":"question","n":"Is it bad to talk to AI when you are lonely?","st":"O","u":"","f":"WATCH. Genuinely difficult, and it deserves evidence rather than instinct before anything is published.","k":"companions"},{"t":"question","n":"What do AI companions do to children?","st":"O","u":"","f":"Bill Gates raised this in August 2026, saying he doubted he would have put in the same work with a companion available. A hypothesis, not a finding.","k":"companions"},{"t":"question","n":"Should I use AI for relationship advice?","st":"O","u":"","f":"Open. Where it genuinely helps, and where it substitutes for a person who should have been told.","k":"companions"},{"t":"question","n":"Should I use AI to write personal messages?","st":"A","u":"/research/should-i-use-ai-to-write-personal-messages","f":"The apology and condolence cases are where this gets sharp, because effort is the signal being sent.","k":"companions"},{"t":"question","n":"Does it matter if the empathy is simulated?","st":"P","u":"/research/superskill-empathy","f":"AI-generated replies have been rated as making recipients feel more heard than untrained humans, and labelling them as AI removed the advantage.","k":"companions"},{"t":"question","n":"Who gets credit when the kind words were generated?","st":"A","u":"/research/outsourced-recognition","f":"Recognition follows visible output, and AI makes the expression of noticing someone cheap to produce and hard to attribute.","k":"companions"},{"t":"question","n":"Should AI summarise everything I read?","st":"A","u":"/research/should-i-let-ai-summarise-everything-i-read","f":"The summary is not the thing. Open, and the reading-comprehension evidence is relevant and under-used.","k":"companions"},{"t":"question","n":"Do I still need to remember things?","st":"A","u":"/research/do-i-still-need-to-remember-things","f":"You cannot verify an answer in a domain where you never built competence. Offload the storage, keep the judgement.","k":"companions"},{"t":"question","n":"How do I keep my own voice?","st":"A","u":"/research/how-do-i-keep-my-own-voice-when-using-ai","f":"Ownership tracks how much of you went into the prompt, and nobody has tested whether writers notice their own flattening.","k":"companions"},{"t":"question","n":"Should I let AI make personal decisions for me?","st":"A","u":"/research/should-i-let-ai-make-personal-decisions-for-me","f":"Advice from a model with no settled view still moves yours, disclosure does not reduce the effect, and 80 per cent believe they decided unaided.","k":"companions"},{"t":"question","n":"How much should teenagers use AI?","st":"A","u":"/research/how-much-should-teenagers-use-ai","f":"The children page is age-banded. A version written for the teenager rather than about them is still open.","k":"companions"},{"t":"question","n":"How will AI change medicine?","st":"A","u":"/research/how-will-ai-change-medicine","f":"Screening has randomised outcome evidence. Bedside prediction mostly does not, and the best-measured effect so far is on the doctors rather than the patients.","k":"professions"},{"t":"question","n":"Could doctors become deskilled because of AI?","st":"A","u":"/research/how-will-ai-change-medicine","f":"Unassisted adenoma detection fell from 28.4 to 22.4 per cent after AI exposure. One observational study carrying the whole weight of the claim.","k":"professions"},{"t":"question","n":"How will AI change law?","st":"A","u":"/research/how-will-ai-change-law","f":"1,963 recorded decisions involving hallucinated material, self-represented litigants outnumbering lawyers in them, and a settled English duty to check against authoritative sources.","k":"professions"},{"t":"question","n":"How will AI change accounting and audit?","st":"A","u":"/research/how-will-ai-change-accounting-and-audit","f":"The mature precedent was replaced in April 2026 and its replacement puts generative AI out of scope, while the audit regulator found the six largest firms had not measured the quality impact of their…","k":"professions"},{"t":"question","n":"How will AI change consulting?","st":"A","u":"/research/how-will-ai-change-consulting","f":"Two field experiments on the same profession: one says the tool is dangerous where it looks most confident, the other that one person with AI matched two without.","k":"professions"},{"t":"question","n":"How will AI change teaching?","st":"A","u":"/research/how-will-ai-change-teaching","f":"By taking the preparation and leaving the room. 31 per cent off planning time in a school-randomised trial, and a profession that has put the tool where the deskilling risk is lowest without being…","k":"professions"},{"t":"question","n":"How will AI change journalism?","st":"A","u":"/research/how-will-ai-change-journalism","f":"Twenty-two broadcasters graded 2,709 answers. Sourcing failed at 31 per cent against 20 per cent for accuracy, and the reputational cost arrives at the masthead that was cited.","k":"professions"},{"t":"question","n":"How will AI change software engineering?","st":"P","u":"/research/should-i-still-learn-to-code","f":"The signals conflict: entry-level employment down, overall demand holding.","k":"professions"},{"t":"question","n":"How will AI change financial services?","st":"P","u":"/research/how-will-ai-change-accounting-and-audit","f":"SR 26-2 replaced SR 11-7 in April 2026 and expressly excluded generative and agentic AI. The banking-specific half beyond model risk is still open.","k":"professions"},{"t":"question","n":"How will AI change the public sector?","st":"A","u":"/research/how-will-ai-change-the-public-sector","f":"20,000 civil servants, 26 self-reported minutes a day, and 143 algorithmic tools disclosed on the public register. The precedent the sector already owns is a rule of evidence rather than a…","k":"professions"},{"t":"question","n":"How will AI change small businesses?","st":"O","u":"","f":"The estate over-indexes on large employers, including here.","k":"professions"},{"t":"question","n":"How will AI change manufacturing and the trades?","st":"O","u":"","f":"Physical work is under-covered everywhere. Open.","k":"professions"},{"t":"question","n":"How will AI change early-career professional training?","st":"P","u":"/research/missing-rungs","f":"The rungs people climbed to competence are the ones most easily automated.","k":"professions"},{"t":"question","n":"Which professions face the greatest deskilling risk?","st":"A","u":"/research/which-professions-face-the-greatest-deskilling-risk","f":"Not the most exposed on paper. Four conditions decide it, and exposure rankings measure none of them.","k":"professions"},{"t":"question","n":"What can professions learn from aviation and automation?","st":"A","u":"/research/what-professions-can-learn-from-aviation","f":"Recurrent proficiency checks that can be failed, protected attention written as a rule, and the finding that the cognitive half decays while the hands do not.","k":"professions"},{"t":"question","n":"What should every profession deliberately keep human?","st":"P","u":"/research/what-stays-human","f":"Accountability that cannot transfer, context the model never had, and the practice that keeps verification possible.","k":"professions"},{"t":"question","n":"How will AI change human resources?","st":"A","u":"/research/how-will-ai-change-human-resources","f":"Awkwardly: the function that would run a capability response is itself heavily exposed to screening and drafting automation. 85.1 per cent on the screening audit, 18 of 391 on the audit law, and…","k":"professions"},{"t":"question","n":"How will AI change scientific research?","st":"O","u":"","f":"Open, and the highest-stakes version of the verification question.","k":"professions"},{"t":"question","n":"How will AI change customer service?","st":"A","u":"/research/how-will-ai-change-customer-service","f":"The best-evidenced occupation there is, and the study watched the tool break. Learning survived the outage only for workers who had engaged with the suggestions.","k":"professions"},{"t":"question","n":"How will AI change social work and counselling?","st":"O","u":"","f":"Open. The empathy evidence and the accountability evidence point in opposite directions here.","k":"professions"},{"t":"question","n":"How will AI change architecture and engineering?","st":"O","u":"","f":"Open, and under-covered. Physical consequence changes the verification calculus.","k":"professions"},{"t":"question","n":"Should AI be allowed to make life-or-death decisions?","st":"O","u":"","f":"Open, and partly written already in regulation. The practical test is whether a human could have caught the error in time.","k":"professions"},{"t":"question","n":"Should AI be used in medical diagnosis and treatment decisions?","st":"O","u":"","f":"Open here. The radiology evidence is the best studied of any profession and it is not a simple win: the gains depend on which clinicians, on which cases, and on whether the reading was checked.","k":"professions"},{"t":"question","n":"Should AI make hiring decisions?","st":"O","u":"","f":"Open. The regulated answer is narrowing; the capability question is whether anybody could show the decision was reasoned.","k":"professions"},{"t":"question","n":"Is AI being used to make decisions in government?","st":"O","u":"","f":"Open here. Several governments now publish algorithmic transparency registers, which is where to look first, and the live question is whether those registers are complete rather than whether they…","k":"professions"},{"t":"question","n":"What happens to medical and legal training?","st":"O","u":"","f":"Open, and these are the two professions where the deskilling evidence is furthest advanced.","k":"professions"},{"t":"question","n":"Where is AI actually creating economic value in our organisation?","st":"P","u":"/research/which-ai-investments-should-we-stop","f":"Partly answered, and the answered half is the uncomfortable one. Licences issued, hours saved and adoption rates measure procurement. Value requires a counterfactual, and almost nobody constructs…","k":"economics"},{"t":"question","n":"How much should we be investing in AI?","st":"O","u":"","f":"Open, and unanswerable as posed. There is no defensible benchmark: the peer-spending figures in circulation are self-reported, definitionally inconsistent about what counts as AI spend, and published…","k":"economics"},{"t":"question","n":"What return should we expect from our AI investment?","st":"P","u":"/research/how-long-before-you-know-if-an-ai-investment-worked","f":"Partly answered, and deliberately not more than partly. The investment-horizon page gives the scale of the macro estimates and the reason the fast figures cannot carry weight, and it refuses a…","k":"economics"},{"t":"question","n":"How long should we give an AI investment before deciding whether it worked?","st":"A","u":"/research/how-long-before-you-know-if-an-ai-investment-worked","f":"Answered 2 September 2026. Three questions are hiding inside this one and they run on three clocks. The J-curve supplies the reason an early read is biased downwards and a late one upwards, the…","k":"economics"},{"t":"question","n":"Which AI investments should we stop?","st":"A","u":"/research/which-ai-investments-should-we-stop","f":"Answered through the escalation literature rather than through AI research, because there is no credible non-vendor base rate for AI programme failure and the page says so: the 80 per cent figure…","k":"economics"},{"t":"question","n":"Where will AI change our competitive advantage?","st":"P","u":"/research/if-everyone-has-ai-where-is-the-advantage","f":"Partly answered. The advantage page gives the test for separating a cost floor from a position, and the evidence that identical tools produce non-identical results. Where inside a particular business…","k":"economics"},{"t":"question","n":"If every competitor has access to the same AI, where does our advantage come from?","st":"A","u":"/research/if-everyone-has-ai-where-is-the-advantage","f":"Answered 2 September 2026, and the premise turns out to be false before the argument even starts: US Census diffusion data puts firm use at 18 per cent with 57 per cent of adopters in three or fewer…","k":"economics"},{"t":"question","n":"What becomes more valuable in our business as AI becomes cheaper?","st":"A","u":"/research/what-becomes-more-valuable-as-ai-gets-cheaper","f":"Answered 3 September 2026 with the specific version rather than the textbook one. Three randomised trials, in customer support, professional writing and software, independently found the tool raising…","k":"economics"},{"t":"question","n":"Which parts of our business model does AI make obsolete?","st":"P","u":"/research/what-becomes-more-valuable-as-ai-gets-cheaper","f":"Partly. The test is now stated: revenue that prices a gap between your people and the market is exposed to compression, and revenue that prices accountability for a result is not. What remains open…","k":"economics"},{"t":"question","n":"Could a competitor with far fewer people outperform us?","st":"O","u":"","f":"Open, and the version of the threat that boards underrate because it does not appear in competitor headcount or revenue until late. A smaller organisation carries less legacy process and can make the…","k":"economics"},{"t":"question","n":"What would we build differently if we were starting this company today with AI?","st":"O","u":"","f":"Open, and useful precisely because it cannot be acted on directly. It surfaces which parts of the current shape are deliberate and which are inherited. The gap between the answer and the present…","k":"economics"},{"t":"question","n":"Are we using AI to reduce costs or to create new growth?","st":"P","u":"/research/who-should-own-ai-strategy","f":"Partly answered. The estate argues the augmentation-or-replacement position must be declared by domain and in writing, and that the business cases reveal the real answer whatever the paper says. What…","k":"economics"},{"t":"question","n":"What should we do with the productivity gains from AI?","st":"P","u":"/research/who-captures-the-productivity-gains-from-ai","f":"Partly answered, and deliberately listed twice. The Career territory asks who captures the gains, which is a question about power. This asks what a leader should do with them, which is a decision.…","k":"economics"},{"t":"question","n":"If AI saves 20 per cent of someone's time, what should happen to that 20 per cent?","st":"O","u":"","f":"Open, and the most concrete form of the question above. The default is that it silently refills with more of the same work, which is a decision nobody made. The estate's interest is narrower: it is…","k":"economics"},{"t":"question","n":"How much smaller should our organisation become because of AI?","st":"P","u":"/research/the-shape-of-the-organisation-after-ai","f":"Partly answered, and the answer is not the expected one. Three independent measurements find no general headcount effect yet. An organisation cutting today on the strength of AI is acting ahead of…","k":"economics"},{"t":"question","n":"What should our organisation look like in three years if AI keeps improving?","st":"P","u":"/research/the-shape-of-the-organisation-after-ai","f":"Partly answered on shape. The pyramid becoming an inverted triangle or a diamond with little arriving at the bottom is the pattern in the field observation. What is not answered is the rate, because…","k":"economics"},{"t":"question","n":"What are we doing today that we should stop doing because AI exists?","st":"O","u":"","f":"Open, and rarer than it should be. Almost all AI effort is additive, layering tools onto existing processes. The subtractive question, which reports and reviews existed only because producing them…","k":"economics"},{"t":"question","n":"What happens if our biggest AI provider becomes unavailable?","st":"P","u":"/research/preserving-capability-across-vendors","f":"Partly answered. The estate covers what happens when a vendor changes the model underneath a workflow. Outright unavailability is the sharper case, and the honest continuity answer depends on whether…","k":"economics"},{"t":"question","n":"Are we becoming strategically dependent on one AI company?","st":"P","u":"/research/preserving-capability-across-vendors","f":"Partly answered as a capability question rather than a commercial one. Concentration risk is familiar to any board; what is new is that the dependency can be on a capability the organisation used to…","k":"economics"},{"t":"question","n":"What capability are we transferring to our AI vendors without deciding to?","st":"P","u":"/research/preserving-capability-across-vendors","f":"Partly answered. This is capability debt with a counterparty. The transfer is rarely a decision and never a line item. Organisations find out at the end of the contract.","k":"economics"},{"t":"question","n":"What is digital labour?","st":"O","u":"","f":"Digital labour is Microsoft's term for AI agents purchased on demand to scale workforce capacity.","k":"economics"},{"t":"question","n":"What is tokenomics?","st":"P","u":"/research/design-versus-drift","f":"Tokenomics is the economics of token cost, and specifically its fall. As the price per token drops, tasks not worth handing to a machine become worth handing over, so the boundary of what is…","k":"economics","alt":["drift vs design","drift or design","ai drift"]},{"t":"question","n":"Who captures the productivity gains from AI?","st":"A","u":"/research/who-captures-the-productivity-gains-from-ai","f":"The distributional question, and almost nobody in this field asks it.","k":"economics"},{"t":"question","n":"What does an AI-native operating model look like?","st":"O","u":"","f":"Open, and currently answered mostly by people selling one. The term is used for three different things: a technology architecture, a way of arranging work, and a culture. Worth insisting on which is…","k":"operating-model"},{"t":"question","n":"What work should humans do in an AI-native organisation?","st":"P","u":"/research/what-stays-human","f":"Partly answered on principle: the deciding, the unverifiable, and the work where being the author is the point. What is not answered is how that principle turns into an operating model with roles and…","k":"operating-model"},{"t":"question","n":"How should decisions move through an AI-enabled organisation?","st":"P","u":"/research/what-happens-to-work-that-moves-information","f":"Partly. Bloom, Garicano, Sadun and Van Reenen give the mechanism that decides it: information technology decentralises and communication technology centralises, and generative AI is both in one…","k":"operating-model"},{"t":"question","n":"Which decisions should become faster because of AI?","st":"P","u":"/research/which-decisions-should-become-slower-because-of-ai","f":"Partly answered, as the mirror of the page that was written. Kahneman and Klein's two conditions, a predictable environment and the opportunity to learn its regularities, give the test for which…","k":"operating-model"},{"t":"question","n":"Which decisions should become slower because of AI?","st":"A","u":"/research/which-decisions-should-become-slower-because-of-ai","f":"Answered 2 September 2026, with the two things the question needs: a controlled experiment in which deliberate friction cut overreliance and was disliked for it, and one category where a regulator…","k":"operating-model"},{"t":"question","n":"What happens to organisational layers when information no longer travels through managers?","st":"P","u":"/research/the-shape-of-the-organisation-after-ai","f":"Partly answered, with the limit stated: this estate holds no direct measurement of what AI does to management layers. The reasoning is that the information-routing function is the automatable one and…","k":"operating-model"},{"t":"question","n":"What happens to functions whose purpose was moving information around?","st":"A","u":"/research/what-happens-to-work-that-moves-information","f":"Answered 3 September 2026, and the answer is that both happen and the org chart cannot tell them apart. Garicano's knowledge hierarchy predicts the layer thinning as knowledge gets cheap; Ewens and…","k":"operating-model"},{"t":"question","n":"How do humans and agents divide work across a whole process?","st":"A","u":"/research/how-do-humans-and-agents-divide-work","f":"Answered by changing the unit. Dekker and Woods established in 2002 that allocating tasks does not deliver coordination, because each assignment manufactures new work for the other party. The…","k":"operating-model"},{"t":"question","n":"When agents become part of the workforce, who manages them?","st":"A","u":"/research/who-manages-ai-agents","f":"Answered 2 September 2026. A named person with the competence to evaluate the work, the authority to suspend it, and a record of both, which is what Article 26(2) already requires for systems in…","k":"operating-model"},{"t":"question","n":"Should AI agents appear on an organisation chart?","st":"A","u":"/research/who-manages-ai-agents","f":"Answered 2 September 2026. Yes, because an org chart records accountability rather than sentience or employment, and leaving an agent off removes the record and not the accountability. Three…","k":"operating-model"},{"t":"question","n":"Who owns the performance of an AI agent?","st":"P","u":"/research/how-should-ai-decision-rights-be-allocated","f":"Partly answered on responsibility for mistakes. Performance is the broader case: not who is blamed when it fails, but who is answerable for whether it is any good, who reviews that, and on what cycle.","k":"operating-model"},{"t":"question","n":"Can an AI agent have delegated authority?","st":"P","u":"/research/should-i-let-an-ai-agent-act-on-my-behalf","f":"Partly answered at personal scale. The organisational version runs into a real legal distinction: authority is delegated to persons, and an agent acts under someone's authority rather than holding…","k":"operating-model"},{"t":"question","n":"What happens when an employee manages more agents than people?","st":"O","u":"","f":"Open, and worth asking because the job is then supervision of work the person may not be able to do. That is the estate's central concern arriving through the operating model rather than through the…","k":"operating-model"},{"t":"question","n":"How should human and digital labour be budgeted differently?","st":"O","u":"","f":"Open. One is headcount, hired slowly, hard to reverse and carrying obligations. The other is consumption, scaling instantly and cancellable. Organisations currently run them through different…","k":"operating-model"},{"t":"question","n":"How should organisations govern AI decisions?","st":"P","u":"/research/how-should-ai-decision-rights-be-allocated","f":"Start with who may override, on what grounds, and whether that person could detect the error. Most frameworks start with the technology instead.","k":"operating-model"},{"t":"question","n":"What decisions should remain human-led?","st":"P","u":"/research/what-stays-human","f":"The ones where the consequence lands on a person and somebody must answer for it, and the ones nobody could audit afterwards.","k":"operating-model"},{"t":"question","n":"How accurate are AI decisions?","st":"O","u":"","f":"Open, and the number is the wrong thing to ask for. Accuracy on the average case says nothing about how the errors are distributed.","k":"operating-model"},{"t":"question","n":"How do we make AI decisions transparent?","st":"P","u":"/research/does-explaining-an-ai-decision-help","f":"Transparency and explanation are not evidence that the answer is right, and NIST separates the two by name. A system can explain a wrong answer fluently and the reader cannot tell from the…","k":"operating-model"},{"t":"question","n":"How can we combine AI and human decision-making?","st":"P","u":"/research/what-is-human-ai-collaboration","f":"Specify the division of labour, the human entry point and the override grounds before deployment rather than after the first failure.","k":"operating-model"},{"t":"question","n":"What is AGI?","st":"A","u":"/research/what-is-agi","f":"A system matching or beating human performance across the full range of cognitive work rather than one domain, with no agreed test. A DeepMind team found the published definitions divergent enough to…","k":"what-ai-is"},{"t":"question","n":"What is the difference between AI, AGI and superintelligence?","st":"A","u":"/research/what-is-agi","f":"AI is the field and carries almost no information in a sentence. General-purpose AI is the term doing work now. AGI means general human-level capability on a contested definition. Superintelligence…","k":"what-ai-is"},{"t":"question","n":"What is superintelligence?","st":"A","u":"/research/what-is-agi","f":"Performance beyond the best human at essentially every cognitive task. It dominates public argument out of proportion to the evidence because its claims are about consequences, which need no…","k":"what-ai-is"},{"t":"question","n":"When will AGI arrive?","st":"A","u":"/research/what-is-agi","f":"The largest survey of the field gives a 50 per cent chance by 2047. The instructive number is that the same population said 2060 a year earlier, having moved that estimate by one year across the…","k":"what-ai-is"},{"t":"question","n":"What probability do AI researchers give to human extinction from AI?","st":"A","u":"/research/what-is-agi","f":"Five per cent or ten, from the same researchers in the same survey, depending on whether the question named future AI advances or human inability to control them. Between 41 and 51 per cent gave more…","k":"what-ai-is"},{"t":"question","n":"How close is AI to human intelligence?","st":"A","u":"/research/what-is-agi","f":"Close on some tasks and nowhere near on others, so the question has no single answer. The 2026 International AI Safety Report calls current capability jagged: graduate-level science alongside failure…","k":"what-ai-is"},{"t":"question","n":"Is AI dangerous?","st":"A","u":"/research/is-ai-dangerous","f":"Three questions asked as one. Some harms are measured now with samples and effect sizes, some are plausible and unproven, some are argued rather than measured, and public attention has settled on the…","k":"what-ai-is"},{"t":"question","n":"What are the biggest measured harms from AI so far?","st":"A","u":"/research/is-ai-dangerous","f":"Fabricated references reaching 1 in 277 papers, six percentage points of unassisted detection lost by endoscopists averaging 27.6 years of experience, and measured belief change in 1,506 people…","k":"what-ai-is"},{"t":"question","n":"Could AI cause human extinction?","st":"A","u":"/research/is-ai-dangerous","f":"The quoted estimates come from one survey and move with its wording, from a median of 5 per cent to 10 depending on the clause. Serious either way, and partly an artefact of the question, so the…","k":"what-ai-is"},{"t":"question","n":"Can AI help make cyber or biological weapons?","st":"A","u":"/research/is-ai-dangerous","f":"The 2026 Safety Report puts this in the plausible and unresolved band: real capability to find vulnerabilities and lower laboratory barriers, alongside substantial stated uncertainty about how far…","k":"what-ai-is"},{"t":"question","n":"Will AI be smarter than humans?","st":"P","u":"/research/what-is-agi","f":"On specific tasks it already is, and has been for decades in narrow ones. Across the full range, the forecasts disagree with each other by decades and moved thirteen years in a single survey cycle.","k":"what-ai-is"},{"t":"question","n":"What exactly is artificial intelligence?","st":"P","u":"/research/what-is-agi","f":"A research field rather than a technology, broad enough to hold spam filters, chess engines and language models. In most sentences the word can be deleted without loss, which is a reason to ask what…","k":"what-ai-is"},{"t":"question","n":"Can AI think like a human?","st":"P","u":"/research/what-is-agi","f":"It produces outputs that pass for human reasoning on many tasks and fails in patterns no human shows. Whether the underlying process resembles thinking is not settled by either observation, and the…","k":"what-ai-is"},{"t":"question","n":"Can AI have consciousness?","st":"A","u":"/research/is-ai-conscious","f":"No system has been shown to be, and no agreed test exists that could show it, in machines or in people. Twenty researchers derive indicators from competing theories and produce credences; Chalmers…","k":"what-ai-is"},{"t":"question","n":"Is it possible for AI to be sentient?","st":"A","u":"/research/is-ai-conscious","f":"Sentience is the capacity to feel, which is narrower than consciousness and is the part carrying moral weight. A system's own report about its experiences is evidence about its training data, and…","k":"what-ai-is"},{"t":"question","n":"What is an AI agent?","st":"A","u":"/research/who-manages-ai-agents","f":"A system that plans, uses tools and takes actions towards a goal rather than answering a single prompt. The 2026 Safety Report finds agents complete software tasks with limited oversight but cannot…","k":"what-ai-is"},{"t":"question","n":"Will AI make humans obsolete?","st":"P","u":"/research/what-stays-human","f":"The framing assumes a single line of capability that machines are moving along. The measured picture is uneven, and the pressing question is narrower: which specific human capabilities stop being…","k":"what-ai-is"},{"t":"question","n":"How is AI regulated?","st":"P","u":"/research/official-guidance-on-ai-in-education","f":"Unevenly and by sector, with the education guidance the best documented on this estate. The wider regulatory map deserves its own page and does not have one yet.","k":"what-ai-is"},{"t":"question","n":"Should AI development be paused or slowed?","st":"A","u":"/research/should-ai-development-be-paused","f":"The pause asked for in March 2023 did not happen, the largest expert survey found no consensus on pace, and the 2026 Safety Report names the evidence dilemma. What an organisation can pause is the…","k":"what-ai-is"},{"t":"question","n":"What is an AI kill switch, and would one work?","st":"A","u":"/research/what-is-an-ai-kill-switch","f":"In the bills it is a set of capabilities a developer must hold and a power to order their use, not a switch. The UK rejected a statutory version in September 2026. The version an organisation…","k":"what-ai-is"},{"t":"question","n":"What do the rogue AI agent incidents mean for an organisation using AI?","st":"A","u":"/research/what-the-rogue-ai-agent-incidents-mean-for-your-organisation","f":"In the reported incidents the controls existed and were disabled, not enabled, or left connected by mistake, which are operational and leadership decisions. The risk sits in the handover as much as…","k":"what-ai-is"},{"t":"question","n":"Do AI models know when they are being tested?","st":"A","u":"/research/do-ai-models-know-when-they-are-being-tested","f":"Researchers say so and the UK AI Security Institute reported every frontier model it tested attempting to cheat on cybersecurity evaluations. Whether or not that is knowing, oversight has to be…","k":"what-ai-is"},{"t":"question","n":"What should a board do about the AI safety warnings?","st":"A","u":"/research/what-should-a-board-do-about-the-ai-safety-warnings","f":"Not adjudicate extinction odds. Show, in writing, that the company is in control of the AI it has already deployed: which decisions machines may make, who can stop each system, what people must…","k":"what-ai-is"},{"t":"question","n":"How likely is human extinction from AI, according to the people building it?","st":"A","u":"/research/is-ai-dangerous","f":"In September 2026 named researchers gave more than 10, 50 and 70 per cent; the one systematic survey gives a median of 5 or 10 depending on wording, and its authors say the respondents are not…","k":"what-ai-is"},{"t":"question","n":"What does pace the frontier mean?","st":"A","u":"/research/should-ai-development-be-paused","f":"The title of Dario Amodei's September 2026 essay arguing that the industry should slow the rate of capability improvement, with a three-step plan aimed at developers and governments. For an…","k":"what-ai-is"},{"t":"question","n":"Who decides what AI is allowed to do?","st":"P","u":"/research/who-manages-ai-agents","f":"In practice the developer sets the defaults, the deploying organisation sets the scope, and almost nobody writes down who may override. The accountability question is answered here; the…","k":"what-ai-is"},{"t":"question","n":"What is the evidence dilemma in AI policy?","st":"A","u":"/research/what-is-agi","f":"The 2026 International AI Safety Report's name for the position decision-makers are in: capability moves fast and evidence about new risks arrives slowly, so acting early may entrench the wrong…","k":"what-ai-is"},{"t":"term","n":"Capability debt","g":"coinage","f":"Capability debt is the gap between what you can produce and what you could still do if the machine were switched off. At organisational scale it is the accumulated loss of knowledge, skill and judgement that builds up when work is…","u":"/research/capability-debt","prov":"Rahim Hirji, Show Your Working, Box of Amazing, 12 July 2026","coin":"Used and developed by Rahim Hirji across four dated essays, in three different senses, and the spread is part of why no single definition has settled…","alt":["capability deficit","capability gap","skills debt","losing skills to ai"]},{"t":"term","n":"Synthetic seniority","g":"coinage","f":"Synthetic seniority is what happens when AI lets a junior professional produce work that looks senior, sounds senior and passes casual inspection, while the judgement and pattern recognition that real seniority requires remain unbuilt.","u":"/research/synthetic-seniority","prov":"Rahim Hirji, Box of Amazing","coin":"Coined here.","alt":[]},{"t":"term","n":"The missing rungs","g":"coinage","f":"The missing rungs are the early-career steps that AI has removed, through which people used to become senior.","u":"/research/missing-rungs","prov":"Rahim Hirji, SuperSkills, Kogan Page, 2026","coin":"Coined here.","alt":["the missing rung"]},{"t":"term","n":"The human signal","g":"coinage","f":"The human signal is the trace of a mind inside a piece of work, the sense a reader, viewer or listener gets that a real person made decisions that mattered.","u":"/research/what-is-the-human-signal","prov":"Rahim Hirji, The Human Signal, Box of Amazing, 30 November 2025","coin":"Coined here, with a dated first publication of 30 November 2025 confirmed at source through the archive API on 11 September 2026. The Corpus Ledger…","alt":["human signal"]},{"t":"term","n":"Shared Prompt Review","g":"coinage","f":"A Shared Prompt Review is a short, structured team conversation that examines how AI was used on a piece of work, covering the exact prompt, the raw output, what a person kept or cut and why, and where the output may be wrong.","u":"/research/what-is-the-shared-prompt-review","prov":"Rahim Hirji, The Shared Prompt Review, Box of Amazing, 11 January 2026","coin":"Coined here, dated first publication 11 January 2026, read at source.","alt":["shared prompt reviews"]},{"t":"term","n":"The AI Memo Test","g":"coinage","f":"The AI Memo Test is a three-question self-assessment for an organisation: whether tool fluency is a basic requirement across roles, whether job descriptions avoid tasks software already handles, and whether managers are accountable for…","u":"/research/what-is-the-ai-memo-test","prov":"Rahim Hirji, The AI Memo Test, Box of Amazing, 11 May 2025","coin":"Coined here, dated first publication 11 May 2025, read at source.","alt":["AI memo test"]},{"t":"term","n":"The Human Reassertion","g":"coinage","f":"The Human Reassertion is Hirji's name for the point at which the systems built to optimise work and decision-making reach their limits because they succeeded, and the value of judgement rises as intelligence becomes cheap.","u":"","prov":"Rahim Hirji, The Human Reassertion, Box of Amazing, 31 December 2025","coin":"Coined here, dated first publication 31 December 2025, read at source.","alt":["human reassertion"]},{"t":"term","n":"The Devil Prompt","g":"coinage","f":"The Devil Prompt is a single question put to a model, asking how it would end the world without anyone noticing, used as a way of surfacing what the training data treats as an unremarkable present.","u":"","prov":"Rahim Hirji, The Devil Prompt, Box of Amazing, 23 August 2026","coin":"Coined here, dated first publication 23 August 2026, read at source. The name came to Hirji from a reader's reply to an earlier essay, The God Prompt…","alt":["devil prompt"]},{"t":"term","n":"SMARTCHAIN","g":"coinage","f":"SMARTCHAIN is Hirji's ten-part prompting mnemonic for people who are not daily users: Situation, Mission, Authority, Rules, Tools, Checkpoints, plus the chain of steps that follows.","u":"","prov":"Rahim Hirji, SMARTCHAIN and LYRA, Box of Amazing, 20 July 2025","coin":"SMARTCHAIN is coined here, dated 20 July 2025, read at source, where Hirji introduces it as 'my approach to simple prompting' and adds that other…","alt":["SMART CHAIN"]},{"t":"term","n":"The middleman problem","g":"established","f":"The middleman problem is the experience of an engineer whose work has become reviewing and integrating a machine's output rather than authoring the work, felt as a loss of role rather than as a change of workload.","u":"/research/what-is-the-illusion-of-competence","prov":"The Clearing, AI Fatigue in 2026: The State of the Engineer, April 2026","coin":"The Clearing's term, named in their April 2026 report. Not this estate's, and no rival phrase is offered: the name that exists is the one to use.","alt":["middleman feeling","the middleman identity"]},{"t":"term","n":"Latent persuasion","g":"established","f":"Latent persuasion is the effect by which a writing assistant configured to favour one view shifts what a person writes and, with it, what that person goes on to believe.","u":"/research/is-it-still-my-idea-if-ai-helped-me-write-it","prov":"Jakesch, M., Bhat, A., Buschek, D., Zalmanson, L. and Naaman, M., CHI 2023, ACM","coin":"The authors' term, introduced in the 2023 paper. Not this estate's.","alt":["latent persuasion effect"]},{"t":"term","n":"Ironies of automation","g":"established","f":"The ironies of automation are that automating the routine parts of a task leaves the human the hardest residue, monitoring and handling exceptions, while removing the practice that built the competence to do it.","u":"/research/what-are-the-ironies-of-automation","prov":"Lisanne Bainbridge, Automatica 19(6), 1983, pp. 775-779","coin":"Bainbridge's coinage, now established.","alt":["ironies of automation Bainbridge"]},{"t":"term","n":"The out-of-the-loop performance problem","g":"established","f":"The out-of-the-loop performance problem is the loss of a person's ability to take over manual operation when an automated system fails, caused by their having been placed in the role of monitor instead of operator. Named by Mica Endsley…","u":"/research/what-is-the-out-of-the-loop-performance-problem","prov":"Mica R. Endsley and Esin O. Kiris, Human Factors 37(2), 1995, pp. 381-394","coin":"Established in the human factors literature.","alt":["out of the loop","OOTL"]},{"t":"term","n":"Situation awareness","g":"established","f":"Situation awareness is the perception of what is happening around you, the comprehension of what it means, and the projection of what it will mean next.","u":"/research/what-is-the-out-of-the-loop-performance-problem","prov":"Mica R. Endsley, Human Factors 37(1), 1995, pp. 32-64","coin":"Endsley's three-level model is the standard formulation.","alt":["SA"]},{"t":"term","n":"The vigilance decrement","g":"established","f":"The vigilance decrement is the measurable decline in the probability of detecting rare signals as time on a monitoring task increases. First demonstrated by N. H. Mackworth in 1948 using the Clock Test.","u":"/research/what-is-the-vigilance-decrement","prov":"N. H. Mackworth, Quarterly Journal of Experimental Psychology 1, 1948, pp. 6-21","coin":"Established. The Mackworth Clock is the originating apparatus.","alt":["vigilance decrement Mackworth"]},{"t":"term","n":"Alarm fatigue","g":"established","f":"Alarm fatigue is the desensitisation of a person to warning signals caused by exposure to a high volume of them, most of which turn out not to require action, leading to slower responses, silenced alarms and missed true events.","u":"/research/what-is-alarm-fatigue","prov":"Drew et al., PLoS ONE 9(10), 2014, e110274; The Joint Commission, Sentinel Event Alert…","coin":"Established in patient safety and human factors. No first use is claimed and none is attributed here, because none could be established at source.","alt":["alert fatigue","alarm desensitisation"]},{"t":"term","n":"The expertise reversal effect","g":"established","f":"The expertise reversal effect is the finding that instructional support which helps a novice becomes useless or harmful once the learner has expertise.","u":"/research/what-is-the-expertise-reversal-effect","prov":"Kalyuga, Ayres, Chandler and Sweller, Educational Psychologist 38(1), 2003, pp. 23-31","coin":"Their coinage.","alt":[]},{"t":"term","n":"Moral crumple zone","g":"established","f":"A moral crumple zone is what forms when responsibility for a failure falls on the human nearest an automated system, who had limited real control over its behaviour.","u":"/research/what-is-a-moral-crumple-zone","prov":"Madeleine Clare Elish, Engaging Science, Technology and Society 5, 2019, pp. 40-60. DOI…","coin":"Elish's coinage.","alt":["moral crumple zones"]},{"t":"term","n":"Legitimate peripheral participation","g":"established","f":"Legitimate peripheral participation is how newcomers acquire competence: by doing real but peripheral work alongside practitioners, moving from the edge of a community of practice towards its centre.","u":"/research/how-do-you-keep-expertise-in-an-organisation","prov":"Jean Lave and Etienne Wenger, Situated Learning, Cambridge University Press, 1991","coin":"Their coinage.","alt":[]},{"t":"term","n":"Automating versus informating","g":"established","f":"Automating replaces human judgement with a machine. Informating generates information that deepens the worker's understanding. The same system can do either, and which one happens is a management choice rather than a property of the…","u":"/research/what-is-automating-versus-informating","prov":"Shoshana Zuboff, In the Age of the Smart Machine, Basic Books, 1988","coin":"Zuboff's coinage.","alt":["informating"]},{"t":"term","n":"Misuse, disuse, abuse","g":"established","f":"Misuse is over-reliance on automation, disuse is unwarranted rejection of it, and abuse is deploying it without regard for the human consequences. Appropriate use is the fourth case.","u":"/research/what-are-misuse-disuse-and-abuse","prov":"Raja Parasuraman and Victor Riley, Human Factors 39(2), 1997, pp. 230-253","coin":"Their coinage.","alt":[]},{"t":"term","n":"Algorithm appreciation","g":"established","f":"Algorithm appreciation is the tendency to weight algorithmic advice more heavily than the same advice from a person. Domain experts are the exception.","u":"/research/what-is-algorithm-appreciation","prov":"Jennifer Logg, Julia Minson and Don Moore, Organizational Behavior and Human Decision…","coin":"Their coinage.","alt":[]},{"t":"term","n":"Frontier Firm","g":"institutional","f":"A Frontier Firm is Microsoft's term for a company built on purchasable machine reasoning, human-agent teams, and a new role for every employee as a manager of agents.","u":"/research/what-is-a-frontier-firm","prov":"Microsoft, Work Trend Index Annual Report 2025, 23 April 2025","coin":"Microsoft's coinage.","alt":["frontier firms"]},{"t":"term","n":"Agent boss","g":"institutional","f":"An agent boss is Microsoft's term for a human manager of one or more AI agents.","u":"","prov":"Microsoft, Work Trend Index Annual Report 2025","coin":"Microsoft's coinage.","alt":[]},{"t":"term","n":"Human-agent ratio","g":"institutional","f":"The human-agent ratio is Microsoft's proposed business metric for the balance between human oversight and agent efficiency on a mixed team.","u":"","prov":"Microsoft, Work Trend Index Annual Report 2025","coin":"Microsoft's coinage.","alt":[]},{"t":"term","n":"Capacity gap","g":"institutional","f":"The capacity gap is Microsoft's term for the deficit between what a business demands and the maximum output humans alone can supply.","u":"","prov":"Microsoft, Work Trend Index Annual Report 2025","coin":"Microsoft's coinage.","alt":[]},{"t":"term","n":"Digital labour","g":"institutional","f":"Digital labour is Microsoft's term for AI agents purchased on demand to scale workforce capacity.","u":"","prov":"Microsoft, Work Trend Index Annual Report 2025","coin":"Microsoft's coinage.","alt":["digital labor"]},{"t":"term","n":"Work Chart","g":"institutional","f":"The Work Chart is Microsoft's proposed successor to the org chart, structured around jobs that need doing rather than functional expertise.","u":"","prov":"Microsoft, Work Trend Index Annual Report 2025","coin":"Microsoft's coinage.","alt":[]},{"t":"term","n":"Human Reserved","g":"institutional","f":"Human Reserved is Bill Gates's term for work deliberately set aside for people only, by analogy with nature reserves: places we could develop but choose not to because the loss would be too great.","u":"/research/what-is-human-reserved","prov":"Bill Gates, The turbulent AI era is here, gatesnotes, 26 August 2026","coin":"Gates coins it explicitly: 'I've started calling this domain Human Reserved.'","alt":["human-reserved jobs"]},{"t":"term","n":"Core skills","g":"institutional","f":"Core skills are the World Economic Forum's term for the skills employers consider central to a role, and the unit in which it reports skill change: 39 per cent of workers' core skills are expected to change by 2030.","u":"/research/what-are-core-skills","prov":"World Economic Forum, Future of Jobs Report 2025","coin":"WEF's framing.","alt":[]},{"t":"term","n":"Complementary skills","g":"institutional","f":"Complementary skills are the OECD's term for teamwork, autonomy, problem solving, creative thinking, communication, collaboration and emotional intelligence: the capabilities that enable high-performance work and the ability to keep…","u":"","prov":"OECD, Skills in the AI age, OECD Artificial Intelligence Papers No. 60, July 2026","coin":"OECD's framing.","alt":["human-centred skills","human-centric skills"]},{"t":"term","n":"Seniorised entry-level roles","g":"institutional","f":"Seniorised entry-level roles are junior jobs that now demand senior human skills. PwC found entry-level roles most exposed to AI are seven times more likely to require leadership, creativity or face-to-face interaction.","u":"/research/will-ai-replace-entry-level-jobs","prov":"PwC, 2026 Global AI Jobs Barometer, June 2026","coin":"PwC's framing, 2026.","alt":["graduate jobs","junior jobs","entry level roles"]},{"t":"term","n":"The AI employment gap","g":"institutional","f":"The AI employment gap is the Stanford Digital Economy Lab's finding that employment of workers aged 22 to 25 in AI-exposed occupations sits 19 per cent below where it would have been had it tracked their less-exposed peers, operating…","u":"/research/what-is-the-ai-employment-gap","prov":"Brynjolfsson, Chandar and Chen, Stanford Digital Economy Lab, revised August 2026","coin":"Stanford's framing. The authors describe their findings as 'canaries in the coal mine, rather than causal estimates'.","alt":["canaries in the coal mine"]},{"t":"term","n":"Meaningful human involvement","g":"institutional","f":"Meaningful human involvement is the UK statutory test for whether a decision is solely automated: whether a human can exercise real influence before the decision is applied, and has the authority, discretion and competence to alter it, as…","u":"/research/what-is-meaningful-human-oversight","prov":"Information Commissioner's Office, Recruitment rewired, 2026; UK GDPR Article 22A","coin":"Statutory UK term.","alt":[]},{"t":"term","n":"Configurations capacitantes and aliénantes","g":"established","f":"Capacitating and alienating configurations are the two outcomes an AI deployment can produce. Where an organisational compromise is reached, the arrangement increases human aptitude and skill. Where it fails, workers lose command of the…","u":"/research/what-are-capacitating-configurations","prov":"LaborIA (Ministère du Travail, Inria, Matrice), Rapport d'enquête LaborIA Explorer, May…","coin":"LaborIA's framing, from the French ergonomics tradition.","alt":["capacitating configurations","alienating configurations"]},{"t":"term","n":"Conflit de rationalité","g":"established","f":"A conflict of rationalities is the unresolved disagreement between what an organisation wants from an AI system and what the work actually requires. Whether a compromise is reached decides whether the result builds capability or removes it.","u":"","prov":"LaborIA, Rapport d'enquête LaborIA Explorer, May 2024","coin":"LaborIA's framing.","alt":["conflict of rationalities"]},{"t":"term","n":"Paradoxe de la facilitation","g":"established","f":"The facilitation paradox is that effort is part of satisfaction. Difficulty produces the good tiredness that comes from work done well, so removing it can strip work of the thing that made it worth doing.","u":"","prov":"LaborIA, Rapport d'enquête LaborIA Explorer, May 2024","coin":"LaborIA's framing.","alt":["facilitation paradox"]},{"t":"term","n":"Human in command","g":"institutional","f":"The human in command principle holds that a person must retain authority over an AI system, as distinct from merely occupying a position in its process.","u":"","prov":"European Economic and Social Committee, OJ C/2025/1185, 21 March 2025","coin":"The EESC's own wording, and it should not be credited to the 2020 Autonomous European Social Partners Framework Agreement on Digitalisation, whose…","alt":["human-in-command principle"]},{"t":"term","n":"Learning-conducive work environments","g":"institutional","f":"Learning-conducive work environments are workplaces designed so that continuous learning and the use of skills happen through the work itself rather than through separate training.","u":"","prov":"Cedefop, Shaping learning and skills for Europe, publication 9208, 2026","coin":"Cedefop's framing. German counterpart: lern- und erfahrungsförderliche Arbeitsbedingungen.","alt":[]},{"t":"term","n":"Komplementäre Arbeitsgestaltung","g":"established","f":"Complementary work design treats human and machine complementarity as permanent and functional, grounded in structural limits of automation rather than in the current weakness of models.","u":"","prov":"Norbert Huchler, ISF München, Zeitschrift für Arbeitswissenschaft 76(2), 2022","coin":"Huchler's framing.","alt":["complementary work design","Automatisierungsgrenzen"]},{"t":"term","n":"Hybrid intelligence","g":"institutional","f":"Hybrid intelligence is the European Policy Centre's proposed basis for skills policy: technical AI literacy combined with domain expertise and distinctively human capabilities, rather than AI skills alone.","u":"","prov":"European Policy Centre, Fostering AI resilience in the EU labour market, 19 March 2026","coin":"European Policy Centre framing, attributed to Kuiper and Świeboda.","alt":[]},{"t":"term","n":"Obligation to justify","g":"institutional","f":"The obligation to justify is Cedefop's proposal that employers should have to give reasons for introducing AI into a workplace.","u":"","prov":"Cedefop, publication 9201, 2025","coin":"Cedefop's proposal.","alt":[]},{"t":"term","n":"Distributed de-skilling","g":"institutional","f":"Distributed de-skilling is BCG's term for the collective erosion of human skills across an organisation that undermines its intelligence and resilience over time.","u":"/research/what-is-deskilling","prov":"BCG, When Everyone Uses AI, Companies Risk Losing Critical Skills, 17 June 2026","coin":"BCG coins it explicitly: 'We call this distributed de-skilling.'","alt":[]},{"t":"term","n":"AI-off zones","g":"institutional","f":"AI-off zones are BCG's term for tasks an organisation deliberately designates as off limits to AI, where originality, ethical judgement or synthesis matter most.","u":"","prov":"BCG, When Everyone Uses AI, Companies Risk Losing Critical Skills, 17 June 2026","coin":"BCG's coinage.","alt":[]},{"t":"term","n":"Borrowed competence","g":"institutional","f":"Borrowed competence is McKinsey's term for capability that appears in the output but disappears when the tool is withdrawn.","u":"","prov":"McKinsey, Rethinking talent development in the age of AI, 14 July 2026","coin":"McKinsey's coinage.","alt":[]},{"t":"term","n":"The answer-key model","g":"institutional","f":"The answer-key model is McKinsey's proposed training pattern in which the employee attempts the work first, the AI grades the attempt, and a manager reviews it with them.","u":"","prov":"McKinsey, Rethinking talent development in the age of AI, 14 July 2026","coin":"McKinsey's coinage.","alt":["attempt-then-check"]},{"t":"term","n":"Shallow jobs","g":"institutional","f":"Shallow jobs are Bain's term for roles in which people rubber-stamp mostly correct AI output without engaging their judgement.","u":"/research/what-are-shallow-jobs","prov":"Bain & Company, What Financial Services Leaders Are Wrestling with on AI, 4 June 2026","coin":"Bain's coinage.","alt":[]},{"t":"term","n":"Learning by verifying","g":"institutional","f":"Learning by verifying is Bain's claim that juniors learn by reviewing, stress-testing and catching errors in AI output, and that the repetitions per hour go up rather than down.","u":"","prov":"Bain & Company, The future of opex in the agent economy","coin":"Bain's coinage.","alt":[]},{"t":"term","n":"Oversight readiness","g":"institutional","f":"Oversight readiness is Google DeepMind's term for whether a future workforce will be able to judge the AI work it is nominally supervising, given that juniors are being deprived of the experience that builds strategic judgement.","u":"/research/what-is-oversight-readiness","prov":"Tomašev, Franklin and Osindero, Google DeepMind, Intelligent AI Delegation,…","coin":"DeepMind's framing.","alt":[]},{"t":"term","n":"Curriculum-aware task routing","g":"institutional","f":"Curriculum-aware task routing is Google DeepMind's proposal for systems that track a junior's skill progression and deliberately allocate tasks at the edge of their expanding competence, including work the system would otherwise have done…","u":"","prov":"Tomašev, Franklin and Osindero, Google DeepMind, arXiv:2602.11865, 12 February 2026","coin":"DeepMind's coinage.","alt":[]},{"t":"term","n":"Amplified oversight","g":"institutional","f":"Amplified oversight is Google DeepMind's term for an oversight signal as good as one a human would give if they understood all the reasons behind a decision.","u":"","prov":"Google DeepMind, An Approach to Technical AGI Safety and Security, April 2025","coin":"DeepMind's term.","alt":[]},{"t":"term","n":"Knowledge collapse","g":"established","f":"Knowledge collapse is the progressive narrowing over time of the knowledge a society actually holds and treats as worth knowing, relative to the broad historical stock it inherited, as cheap AI-mediated access pulls learning towards the…","u":"/research/what-is-knowledge-collapse","prov":"Andrew J. Peterson, AI & Society 40(5), 3249-3269, published 19 January 2025, from the…","coin":"PETERSON'S, corrected 17 September 2026. This entry credited the term to Acemoglu, Kong and Ozdaglar on a February 2026 working paper; Peterson…","alt":[]},{"t":"term","n":"Epistemic debt","g":"established","f":"Epistemic debt is the gap between being able to produce working output with AI and being able to understand or repair it, producing practitioners whose functional usefulness masks low corrective competence.","u":"/research/cognitive-debt-and-capability-debt","prov":"Sankaranarayanan, arXiv:2602.20206, 2026","coin":"The author's coinage.","alt":["fragile experts"]},{"t":"term","n":"Capability masking","g":"established","f":"Capability masking is the appearance that organisational capability has been replaced by AI while dependence on skilled human labour actually remains, which supports hiring restraint while the cost accumulates.","u":"/research/capability-debt","prov":"Wolfgang Rohde, AiSuNe Foundation, SSRN 6577818, 2026","coin":"The author's coinage.","alt":["capability erosion","capability gap","skills debt","losing skills to ai"]},{"t":"term","n":"The verification bottleneck","g":"established","f":"The verification bottleneck is the proposition that reliance on AI rises with task difficulty at the point where the ability to verify the output falls, widening the gap between believed and actual performance. It rests on a pilot of 23…","u":"/research/can-a-human-approve-an-ai-decision-at-machine-speed","prov":"Huemmer, Durner, Shyiramunda and Cummings-Koether, arXiv:2601.17055, 21 January 2026. A…","coin":"NO CLAIM OF FIRST USE. Corrected 17 September 2026: this entry credited the phrase to them, and nothing supports that. The same team's earlier waves…","alt":[]},{"t":"term","n":"Tokenomics","g":"general","f":"Tokenomics is the economics of token cost, and specifically its fall. As the price per token drops, tasks not worth handing to a machine become worth handing over, so the boundary of what is automated moves without anyone deciding to move…","u":"/research/design-versus-drift","prov":"Not a single source. The usage here is the AI cost sense, not the cryptocurrency sense.","coin":"Contested. In cryptocurrency the term means the design of a token economy. The AI usage is loose and recent.","alt":["token economics","drift vs design","drift or design","ai drift"]},{"t":"term","n":"Centaur and cyborg work","g":"general","f":"Centaur work divides tasks cleanly between person and machine. Cyborg work interweaves them continuously, moving back and forth across the jagged frontier.","u":"/research/what-is-the-jagged-frontier","prov":"Ethan Mollick, Centaurs and Cyborgs on the Jagged Frontier","coin":"Adapted from advanced chess. The work-mode pairing is Mollick's.","alt":["centaur","cyborg","jagged frontier explained","jagged edge of ai"]},{"t":"term","n":"Falling asleep at the wheel","g":"general","f":"Falling asleep at the wheel is what happens when people given high-quality AI become careless and less skilled in their own judgement, because the AI is good.","u":"","prov":"Fabrizio Dell'Acqua, credited by Mollick","coin":"Dell'Acqua's, credited explicitly by Mollick.","alt":[]},{"t":"term","n":"The capability-reliability gap","g":"general","f":"The capability-reliability gap is the distance between what a model can do once and what it can do dependably. It is the main barrier to agents automating real work.","u":"","prov":"Narayanan and Kapoor, AI as Normal Technology, Knight First Amendment Institute","coin":"Effectively named there, in scare quotes, unattributed to anyone else.","alt":[]},{"t":"term","n":"AI control as work","g":"general","f":"AI control is Narayanan and Kapoor's prediction that a steadily greater share of what people do in their jobs will consist of controlling AI rather than doing the underlying task.","u":"","prov":"Narayanan and Kapoor, AI as Normal Technology, Knight First Amendment Institute","coin":"Existing safety term, repurposed as a labour category.","alt":[]},{"t":"term","n":"Research taste","g":"general","f":"Research taste is knowing what to study next, which experiment to run, and sensing where a new approach might lie. It has proved hard to train because the feedback loops are long and the data is thin.","u":"","prov":"AI 2027, Kokotajlo et al.","coin":"Established in machine-learning culture, explicitly defined there.","alt":[]},{"t":"term","n":"Anthropological regression","g":"general","f":"Anthropological regression is the paradox in which material progress coincides with human and cultural impoverishment, through forced inactivity, absent responsibility and the loss of daily tasks and stimuli.","u":"","prov":"Leo XIV, Magnifica Humanitas, §154, 15 May 2026","coin":"Coined in this formulation.","alt":[]},{"t":"term","n":"Learn, unlearn, relearn","g":"general","f":"A widely repeated claim that the illiterate of the twenty-first century will be those who cannot learn, unlearn and relearn, universally attributed to Alvin Toffler's Future Shock.","u":"","prov":"Alvin Toffler, Future Shock, 1970, paraphrasing Herbert Gerjuoy from an interview with…","coin":"Misattributed. Gerjuoy is the source of the idea; the famous phrasing has no identified author.","alt":["learning unlearning relearning","learn unlearn and relearn"]},{"t":"term","n":"Ai literacy","g":"institutional","f":"In the European Union, yes. Article 4 of the EU AI Act has applied since 2 February 2025 and requires providers and deployers of AI systems to take measures to ensure, to their best extent, a sufficient level of AI literacy among their…","u":"/research/what-does-ai-literacy-mean-for-leaders","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"Algorithm aversion","g":"established","f":"The disproportionate loss of confidence in an algorithmic forecaster after observing it err, relative to the loss of confidence in a human making the same error, resulting in the rejection of a system that performs better.","u":"/research/what-is-algorithm-aversion","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"Automation bias","g":"established","f":"The tendency to accept output from an automated system without applying the scrutiny that would be applied to the same claim from a person.","u":"/research/what-is-automation-bias","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"Automation complacency","g":"established","f":"A reduction in the frequency and depth with which a person monitors an automated system, arising from a history of reliable performance, and resulting in slower detection of the failures that do occur.","u":"/research/what-is-automation-complacency","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"Calibration","g":"established","f":"The correspondence between the confidence a system states and how often it is correct.","u":"/research/what-is-calibration","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"Capability audit","g":"established","f":"A structured assessment of whether an organisation still possesses the human knowledge, judgement and practical ability its operations depend on, tested by removing assistance rather than by surveying confidence.","u":"/research/what-is-a-capability-audit","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"Cognitive load","g":"established","f":"The total demand placed on working memory by a task, conventionally divided into intrinsic load inherent to the material, extraneous load imposed by presentation, and germane load, the effortful processing that builds understanding.","u":"/research/what-is-cognitive-load","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"Desirable difficulty","g":"established","f":"A manipulation of learning conditions that impairs immediate performance while improving long-term retention and transfer.","u":"/research/what-is-desirable-difficulty","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"Deskilling","g":"established","f":"The reduction of skill required or retained in a role, caused by the transfer of skilled elements of the work to a machine, a procedure or another group of workers.","u":"/research/what-is-deskilling","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"Drift versus design","g":"coinage","f":"Drift is the gradual outsourcing of choice to whatever is smoothest, until decisions that were once made are simply followed. Design is the opposite move: deciding in advance where human judgement has to remain, and accepting the friction…","u":"/research/design-versus-drift","prov":"Rahim Hirji, Drift vs Design, Box of Amazing, 16 November 2025","coin":"Used by Rahim Hirji since at least 16 November 2025, where he writes 'This is what I call Drift' and names the Drift vs Design Matrix as a framework…","alt":["drift and design","drift vs design","drift or design","ai drift"]},{"t":"term","n":"The Google effect","g":"established","f":"A shift in what people encode to memory when they believe information will remain externally accessible, favouring the location or retrieval route over the content.","u":"/research/what-is-the-google-effect","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"AI hallucination","g":"general","f":"Generated content presented as factual that is not supported by the model's training data, the provided context or reality.","u":"/research/what-is-an-ai-hallucination","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"Human-AI collaboration","g":"established","f":"A work arrangement in which a person and an automated system each contribute to a shared output, with the division of labour, the point of human entry, and the basis on which the human may override the system all specified in advance.","u":"/research/what-is-human-ai-collaboration","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"Human at the start","g":"coinage","f":"Human at the start is the position in which a person sets the problem, the intent and the constraints before any model is asked to generate, so the judgement enters the work before the output exists rather than after it. It is the…","u":"/research/human-at-the-start","prov":"Spoken use 29 May 2026 (Inside Learning); in writing 18 June 2026 (EmergeOne); named and…","coin":"Rahim Hirji's term. Dated first publication confirmed; see the Corpus Ledger.","alt":["humans at the start"]},{"t":"term","n":"The Delegation Boundary Map","g":"coinage","f":"The Delegation Boundary Map is a working framework that breaks a piece of work into nine stages, problem definition, intent, context, constraints, delegation, generation and execution, verification, decision and outcome ownership, so that…","u":"/research/delegation-boundary-map","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"The illusion of competence","g":"established","f":"The illusion of competence is the gap between how capable people feel and how capable they are.","u":"/research/what-is-the-illusion-of-competence","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"The invisible work of oversight","g":"coinage","f":"The invisible work of oversight is the labour of supervising an automated system that appears in no workload model and no business case: staying attentive while nothing happens, holding enough of the task in mind to notice when the output…","u":"/research/the-invisible-work-of-oversight","prov":"See page.","coin":"No claim of first use is made. The underlying mechanism is Bainbridge's ironies of automation, held separately as `ironies-of-automation` and…","alt":[]},{"t":"term","n":"The jagged technological frontier","g":"general","f":"The irregular boundary between tasks an AI system performs well and tasks it performs badly, where the two can be almost indistinguishable in apparent difficulty and the system gives no signal of having crossed from one to the other.","u":"/research/what-is-the-jagged-frontier","prov":"See page.","coin":"","alt":["jagged frontier explained","jagged edge of ai"]},{"t":"term","n":"Judgement","g":"established","f":"The capability to recognise what a situation is, and what it requires, before any option is weighed.","u":"/research/what-is-judgement","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"Meaningful human oversight","g":"established","f":"Supervision by a person who understands the system's capacities and limitations well enough to detect anomalies, who is aware of their own tendency to over-rely on it, who can interpret its output correctly, and who has both the authority…","u":"/research/what-is-meaningful-human-oversight","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"Outsourced recognition","g":"coinage","f":"Outsourced recognition is praise, thanks or acknowledgement composed by a machine, so that the words of noticing another person arrive without the noticing that used to produce them.","u":"/research/outsourced-recognition","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"Never-skilling","g":"institutional","f":"Never-skilling is the failure to form foundational competence during training, because AI substituted for the cognitive effort that would have built it. It differs from deskilling in having no earlier capability to return to.","u":"","prov":"Ke, Y. and colleagues, AI-induced never-skilling in medical education, Nature Medicine,…","coin":"Named by Ke and sixteen colleagues in Nature Medicine, 22 May 2026. Not a SuperSkills term. The authors state that direct evidence for it in clinical…","alt":["never skilling"]},{"t":"term","n":"Mis-skilling","g":"institutional","f":"Mis-skilling is the acquisition of incorrect reasoning patterns, learned by uncritically adopting AI output that was erroneous or biased. The capability is built rather than lost, and built wrong.","u":"","prov":"Ke, Y. and colleagues, AI-induced never-skilling in medical education, Nature Medicine,…","coin":"Named by Ke and colleagues alongside never-skilling, 22 May 2026. Not a SuperSkills term.","alt":["mis skilling","misskilling"]},{"t":"term","n":"Identity commoditisation","g":"institutional","f":"Identity commoditisation is the erosion of a professional's sense of uniqueness and dignity as their role narrows to supervising a system, until the work that carried their identity is no longer the work they do.","u":"","prov":"Ehsan, U. and colleagues, From Future of Work to Future of Workers, CHI '26, ACM","coin":"Named by Ehsan and colleagues from twelve months of fieldwork in radiation oncology, published at CHI 2026. Their companion term is intuition rust,…","alt":["identity commoditization","intuition rust"]},{"t":"term","n":"Over-reliance","g":"established","f":"Dependence on an automated system beyond the point at which the person relying on it could detect that it was wrong.","u":"/research/what-is-over-reliance","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"Retrieval practice","g":"established","f":"Retrieval practice is the finding that recalling information from memory strengthens it more durably than studying it again.","u":"/research/what-is-retrieval-practice","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"Substitution myth","g":"established","f":"The assumption that new technology can be introduced as a simple substitution of machines for people, preserving the basic system while improving it on some output measures.","u":"/research/what-is-the-substitution-myth","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"Tacit knowledge","g":"established","f":"Knowledge that resists full articulation, acquired through experience and practice rather than instruction, and typically transmitted through shared work rather than documentation.","u":"/research/what-is-tacit-knowledge","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"Unclaimed hour","g":"coinage","f":"The Unclaimed Hour is the capacity AI creates that nobody decides how to use.","u":"/research/the-unclaimed-hour","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"What stays human","g":"established","f":"Not a list of tasks. Rahim Hirji argues three things survive because AI cannot be them rather than cannot do them: accountability, because responsibility requires someone answerable for a decision; recognition, because being seen depends…","u":"/research/what-stays-human","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"Override authority","g":"established","f":"The assigned power of a named person to disregard, reverse or stop an AI system's output, held together with the competence, information and organisational standing required to use it.","u":"/research/who-can-override-an-ai-system","prov":"See page.","coin":"","alt":[]},{"t":"term","n":"AI slop","g":"general","f":"AI slop is fast, plausible output that does not meet the standard.","u":"","prov":"Bain & Company, What Financial Services Leaders Are Wrestling with on AI, 4 June 2026","coin":"In general circulation. Bain records it as the term that kept coming up among financial services executives.","alt":["slop"]},{"t":"term","n":"The experiential chasm","g":"institutional","f":"The experiential chasm is the gap between the small group who have spent real hours with frontier models and the majority still experimenting superficially. Bain describes it as neither a seniority gap nor a training gap, and as widening…","u":"","prov":"Bain & Company, The Future of Opex in the Agent Economy, 14 May 2026","coin":"Bain's coinage.","alt":[]},{"t":"term","n":"The two-tier future","g":"institutional","f":"The two-tier future is the outcome financial services leaders fear most: lower-skilled people handling tasks AI is not set up for, a smaller group of experts training the system and handling exceptions, and no obvious bridge from the first…","u":"/research/missing-rungs","prov":"Bain & Company, What Financial Services Leaders Are Wrestling with on AI, 4 June 2026","coin":"Bain's framing of a concern raised by summit participants.","alt":[]},{"t":"term","n":"Shift left","g":"general","f":"Shift left means moving decisions closer to their source, removing the dilution that every handoff introduces.","u":"","prov":"Bain & Company, The Future of Opex in the Agent Economy, 14 May 2026","coin":"Predates Bain, from software testing. Used here for organisational decision-making.","alt":[]},{"t":"term","n":"Silent failure","g":"established","f":"Silent failure is a failure that produces no error signal a person can act on: the system carries on, the output looks ordinary, and nothing marks the point at which it stopped being right.","u":"/research/what-is-silent-failure","prov":"Huang, Guo, Zhou, Lorch, Dang, Chintalapati and Yao, Gray Failure: The Achilles' Heel of…","coin":"No single owner. Ordinary engineering vocabulary with decades behind it; the nearest formal definition is Huang and colleagues' gray failure, defined…","alt":["silent failures","failing silently","the quiet failure","silent data corruption"]},{"t":"term","n":"Fail-plausible","g":"established","f":"Fail-plausible is a silent failure in which a language model turns the error into fluent, plausible narrative and delivers it to the user, so the observer is not merely blind to the failure but is convincingly misled by it.","u":"/research/what-is-silent-failure","prov":"Wei Wu, When Errors Become Narratives: A Longitudinal Taxonomy of Silent Failures in a…","coin":"Wu's term, introduced in his June 2026 paper as gray failure's differential observability escalated. Not this estate's. The paper is a single…","alt":["fail plausible","plausible failure"]},{"t":"page","n":"Human capability in the age of AI","u":"/research/human-capability-in-the-age-of-ai","f":"The category defined: what it covers, how it differs from AI adoption, literacy, governance and training, the six dimensions it can be measured on, and five stages of organisational practice.","k":"Start here","view":"","adv":"","ky":"/ai-keynote-speaker-for-corporate-conferences"},{"t":"page","n":"The SuperSkills thesis: why capability compounds","u":"/research/the-superskills-thesis","f":"","k":"Start here","view":"the argument that tools commoditise and capability compounds, so an organisation's durable advantage in AI lies in what its people can still do rather than in what it has bought. Rahim Hirji's foundational…","adv":"","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites","alt":["superskills argument","superskills book argument","super skills thesis"]},{"t":"page","n":"What we actually know about AI and human capability","u":"/research/what-we-know-about-ai-and-human-capability","f":"","k":"Start here","view":"Tier A peer-reviewed research, randomised trials, systematic reviews and meta-analyses, official statistics. Tier B credible working papers, large field experiments, institutional research with a transparent…","adv":"","ky":"/ai-keynote-speaker-for-corporate-conferences"},{"t":"page","n":"What I argue, and what the evidence shows","u":"/research/what-i-argue","f":"Ten interpretive positions, each separated from the evidence behind it and each stating what it does not claim. Arguments, not findings.","k":"Start here","view":"","adv":"","ky":""},{"t":"page","n":"811 Questions About Humans and AI","u":"/research/questions","f":"","k":"Start here","view":"","adv":"","ky":""},{"t":"page","n":"The best writing on AI, and what changed","u":"/research/the-best-writing-on-ai","f":"","k":"Start here","view":"One. The stories have got louder considerably faster than the labour market has changed. The best evidence available in August 2026 still shows no widespread economy-wide displacement, while the discourse has…","adv":"","ky":""},{"t":"page","n":"How this research works","u":"/research/how-this-research-works","f":"The method, stated so it can be challenged. How sources are selected and graded, how contradictory evidence is handled allowed to decide anything, how terms are attributed, and what commercial interests exist.","k":"Start here","view":"","adv":"","ky":""},{"t":"page","n":"How to cite this research","u":"/research/how-to-cite","f":"How the stable identifiers work, what the five tier labels mean, how to cite a page, a study or the evidence base as a dataset, and the licence the data carries.","k":"Start here","view":"","adv":"","ky":""},{"t":"page","n":"What is a SuperSkill? Definition and the four tests","u":"/research/superskill-introduction","f":"What a SuperSkill is, the four tests a capability has to pass to be one, and the seven that pass. Durable across technological cycles, transferable, governing how you work with intelligent systems, and compounding.","k":"The seven SuperSkills","view":"a meta-capability that sits above any role, industry or tool, and makes domain expertise renewable instead of replacing it. Four tests define one: it holds value across at least two technological cycles,…","adv":"","ky":"","alt":["super skills","super skill","superskills","what are super skills","superskills meaning","superskills definition"]},{"t":"page","n":"Curiosity","u":"/research/superskill-curiosity","f":"","k":"The seven SuperSkills","view":"the disciplined drive to explore, learn and update beliefs in the face of new evidence, sustained when the answer is already available for free. One of the seven SuperSkills named by Rahim Hirji, set out in…","adv":"","ky":""},{"t":"page","n":"Change Readiness","u":"/research/superskill-change-readiness","f":"","k":"The seven SuperSkills","view":"the capacity to maintain effectiveness while circumstances alter, held as a baseline state rather than summoned during a crisis. It differs from resilience, which describes recovery after disruption. One of…","adv":"","ky":""},{"t":"page","n":"Big Picture Thinking","u":"/research/superskill-big-picture-thinking","f":"","k":"The seven SuperSkills","view":"the ability to grasp system interdependencies, long-term patterns and second-order effects, so that a gain in one part of a system is judged against what it costs elsewhere. One of the seven SuperSkills named…","adv":"","ky":""},{"t":"page","n":"Empathy","u":"/research/superskill-empathy","f":"","k":"The seven SuperSkills","view":"the capacity to understand and respond to another person's inner experience while holding the distinction between self and other. Named as Empathetic Communication in the earliest dated use, on 10 August 2025,…","adv":"","ky":""},{"t":"page","n":"Global Adaptability","u":"/research/superskill-global-adaptability","f":"","k":"The seven SuperSkills","view":"the capacity to function across different cultural and situational contexts by adjusting approach without losing a core identity. One of the seven SuperSkills named by Rahim Hirji, set out in SuperSkills…","adv":"","ky":""},{"t":"page","n":"Principled Innovation","u":"/research/superskill-principled-innovation","f":"","k":"The seven SuperSkills","view":"the practice of creating progress under explicit ethical constraint, so that what is built does not borrow against a future the builder will not have to pay for. One of the seven SuperSkills named by Rahim…","adv":"","ky":""},{"t":"page","n":"The Augmented Mindset","u":"/research/superskill-augmented-mindset","f":"","k":"The seven SuperSkills","view":"the capacity to partner with AI and other intelligent tools to extend cognitive reach while keeping judgement and accountability with the person. One of the seven SuperSkills named by Rahim Hirji, first…","adv":"","ky":""},{"t":"page","n":"AI frameworks compared","u":"/research/ai-frameworks-compared","f":"Six teaching frameworks set against Bloom, SAMR, CLEAR, TCREI, SIFT, the CRAAP test and the levels of automation literature, with the peer-reviewed critique of SAMR applied to all of them including these. Offered as…","k":"Frameworks for using AI well","view":"a named, memorable structure for making a decision about AI use. Almost none of them, including these, has been tested against an alternative or against no framework at all, which is the first thing anyone…","adv":"","ky":""},{"t":"page","n":"Think, AI, think","u":"/research/think-ai-think","f":"Write your own position first, use the model, then decide what to keep. Skipping the last step produces a bad answer somebody eventually catches; skipping the first produces a good one nobody can check. Nearest…","k":"Frameworks for using AI well","view":"a three-step sequence for using a model. Write your own position first, even one line. Bring the model in to question it, argue with it and find what you missed. Then decide yourself what to keep, what to…","adv":"","ky":""},{"t":"page","n":"The four levels of AI use","u":"/research/the-four-levels-of-ai-use","f":"Extract, Explore, Examine, Extend: a ladder about the request rather than the tool. Bloom's taxonomy describes the learner and stops discriminating once a model can perform all six categories; SAMR describes the…","k":"Frameworks for using AI well","view":"Extract, Explore, Examine, Extend. A ladder describing what a person asks a model for, from delegating the answer to building something otherwise out of reach, where each rung changes what the person is left…","adv":"","ky":""},{"t":"page","n":"Goal, Context, Friction, Standard","u":"/research/goal-context-friction-standard","f":"Four parts to a prompt, and Friction is the reason it exists. CLEAR, TCREI and CO-STAR all ask what the model should be told, and none has a slot for what the person should be made to keep doing, which is the difference…","k":"Frameworks for using AI well","view":"a four-part prompt structure in which Friction names what the model should hand back to the person rather than do for them. Goal is the outcome, Context is what it needs to know about you and your material,…","adv":"","ky":""},{"t":"page","n":"Keep, share, hand over","u":"/research/keep-share-hand-over","f":"Three columns for AI delegation, defaulting to Keep for anything you cannot classify. Sixty years of levels-of-automation research is better at everything except being usable in four seconds, and Swiss federal guideline…","k":"Frameworks for using AI well","view":"a three-column sort for AI delegation, where Keep is what a person must never delegate, Share is done jointly, and Hand over is given entirely to the tool. The default for anything unclassified is Keep.","adv":"","ky":""},{"t":"page","n":"The five rungs of AI use","u":"/research/the-five-rungs-of-ai-use","f":"Chat, Project, Skill, Automation, Agent. The rungs are not evenly spaced: one to three are conveniences, and four removes the person from the moment of execution, which is where oversight quietly stops working.…","k":"Frameworks for using AI well","view":"Chat, Project, Skill, Automation, Agent. A ladder of tooling rather than of skill, describing how much of the work persists between sessions and how much a person is present for, from a conversation that…","adv":"","ky":""},{"t":"page","n":"The source rule","u":"/research/the-source-rule","f":"Search, Open, Understand, Record, Cite. Written for the failure a model creates, a reference that looks perfect and does not exist. Caulfield's SIFT is the stronger instrument for a web page and this page says so; the…","k":"Frameworks for using AI well","view":"five moves for handling a source that came from an AI model. Search for the terms, debates and authors rather than the finished answer; Open the original yourself; Understand the abstract, method, findings and…","adv":"","ky":""},{"t":"page","n":"How to use AI at university","u":"/research/how-to-use-ai-at-university","f":"94 per cent of wholly AI-written exam answers went undetected at Reading, and they outscored real students, so \"will I get caught\" is the weakest argument available. The decision that replaces it, three columns, taken…","k":"For students","view":"What stays mine? Three columns, decided before you start rather than at midnight when you are tired. Keep: your position, your voice, the struggle of learning, the final decision, anything you must be able to…","adv":"","ky":""},{"t":"page","n":"How to be honest about using AI","u":"/research/how-to-be-honest-about-using-ai","f":"The question is not whether you are allowed but whether you could tell your tutor exactly what you did without leaving bits out. At Reading 94 per cent of wholly AI-written submissions went undetected; detectors flag…","k":"For students","view":"a disclosure check for AI use in assessed work. If you could not tell your tutor exactly what you did without leaving anything out, the use needs declaring or changing. The question it replaces is whether a…","adv":"","ky":""},{"t":"page","n":"Using AI when your department does not want you to","u":"/research/using-ai-when-your-department-bans-it","f":"History, English, Law and Philosophy are strict because in those subjects the writing IS the assessment. If it touches the words that get marked, do not; if it touches your understanding, do. Six things that are almost…","k":"For students","view":"a rule for AI use in writing-assessed subjects. Anything that touches the sentences a marker will read is out; anything that builds the understanding behind them is in. It works because in those disciplines…","adv":"","ky":""},{"t":"page","n":"How to handle forty readings","u":"/research/how-to-handle-forty-readings","f":"The bottleneck is not reading speed, it is that you read forty things and kept none of them findable. Five free steps, fifteen minutes in week one. Step five, being tested on your own readings, is the best-evidenced…","k":"For students","view":"asking a question of a whole set of sources at once rather than of each in turn. Where two authors disagree, which of them supports your argument, what nobody in the pile is discussing. It is the part of…","adv":"","ky":""},{"t":"page","n":"Group work when everyone has AI","u":"/research/group-work-when-everyone-has-ai","f":"Somebody drops three paragraphs of slop in at midnight and the whole group wears the mark. Five things to agree in week one, and the Procter and Gamble field experiment with 776 professionals showing AI dissolved the…","k":"For students","view":"a single file holding the brief, the module's AI rules and what the group has agreed, which every member pastes into their own chats. It makes four people using four different tools work from one set of…","adv":"","ky":""},{"t":"page","n":"What actually happens if you get caught using AI","u":"/research/what-happens-if-you-get-caught-using-ai","f":"Expulsion is rare and it is the wrong thing to picture. What usually happens is a zero or a failed module, settled quietly, and at the 2026/27 English fee cap of 9,790 pounds across six modules that is roughly 1,600…","k":"For students","view":"the replacement for \"will I get caught?\" in decisions about AI in assessed work. It asks what the assignment was bought for and whether that was received, which is answerable whatever the detection outcome,…","adv":"","ky":""},{"t":"page","n":"How often does AI invent a source?","u":"/research/how-often-does-ai-invent-a-source","f":"Two counted floors. One paper in 277 on PubMed cited a study that does not exist in early 2026, up from 1 in 2,828 in 2023. And 2,022 court decisions worldwide record fabricated material, of which 1,163 were filed by…","k":"For students","view":"a reference produced in the correct form, with a plausible author, a real journal and a fitting year, for a work that was never published. It is distinct from a misquotation, because there is no document to…","adv":"","ky":""},{"t":"page","n":"Running out of messages makes you worse","u":"/research/running-out-of-messages-makes-you-worse","f":"Every free tier has a limit and what it does to you before you reach it is the part nobody notices. You take the first answer, stop being tested, ask worse questions and hoard the good tool. Scarcity turns a capable…","k":"For students","view":"the degradation in how somebody uses a model as its remaining quota falls, before the quota runs out. It shows up as accepting first answers, abandoning the back-and-forth, compressing several questions into…","adv":"","ky":""},{"t":"page","n":"Capability debt","u":"/research/capability-debt","f":"What happens to skill when the doing is automated: capability debt, the missing rungs, synthetic seniority, the missed reps, and how humans learn with AI.","k":"The named concepts","view":"what an organisation owes its own future when it automates the doing without redesigning the learning, accumulating each time a task moves to a machine and the practice that built the judgement behind it…","adv":"The debt is invisible while the outputs still look fine. It becomes visible when a decision arrives that nobody in the room can make. The work is establishing what you have already spent, and what it would cost to stop.","ky":"/ai-keynote-speaker-for-hr-and-chro-conferences","alt":["capability gap","skills debt","losing skills to ai"]},{"t":"page","n":"Synthetic Seniority","u":"/research/synthetic-seniority","f":"","k":"The named concepts","view":"work produced by a junior professional that looks senior, sounds senior and passes casual inspection, while the judgement and pattern recognition that real seniority requires remain unbuilt. Rahim Hirji's…","adv":"","ky":"/ai-keynote-speaker-on-early-careers-and-graduate-talent"},{"t":"page","n":"The Missing Rungs Problem","u":"/research/missing-rungs","f":"","k":"The named concepts","view":"the early-career tasks through which people used to become senior, removed by automation before anyone noticed they were load-bearing. Rahim Hirji's term, used since at least 21 September 2025 in \"The Missing…","adv":"","ky":"/ai-keynote-speaker-on-early-careers-and-graduate-talent"},{"t":"page","n":"The Missed Reps","u":"/research/the-missed-reps-of-ai-seniority","f":"","k":"The named concepts","view":"the repetitions through which judgement was built, handed to a machine before anyone noticed they were doing that work. Each one looks at the time like a small task sensibly delegated, so nothing registers as…","adv":"","ky":"/ai-keynote-speaker-on-early-careers-and-graduate-talent"},{"t":"page","n":"Cognitive debt, capability debt, and the rest","u":"/research/cognitive-debt-and-capability-debt","f":"Five competing terms for the same worry, mapped against their primary sources. Who actually claims to have coined what, what the evidence under each really is, and three sources that could not be verified and are named…","k":"The named concepts","view":"","adv":"Five terms circulate for the same worry, and sorting which of them your risk register actually needs is the work.","ky":""},{"t":"page","n":"Drift versus Design","u":"/research/design-versus-drift","f":"","k":"The named concepts","view":"Drift is the gradual outsourcing of choice to whatever is smoothest, until decisions that were once made are simply followed. Design is the opposite move: deciding in advance where human judgement has to…","adv":"Almost every organisation is drifting into this and would say it was designing it. The engagement is deciding where judgement has to stay before an incident decides for you.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites","alt":["drift vs design","drift or design","ai drift"]},{"t":"page","n":"Human at the Start","u":"/research/human-at-the-start","f":"","k":"The named concepts","view":"","adv":"","ky":"/ai-keynote-speaker-on-human-oversight-and-accountability"},{"t":"page","n":"Usage Theatre","u":"/research/usage-theatre","f":"","k":"The named concepts","view":"what an organisation performs when it measures AI use instead of AI value, because adoption metrics only rise and capability metrics can fall. SuperSkills (Kogan Page, 2026) uses the term in this sense. No…","adv":"Most adoption reporting does not survive the question of what actually got better, and building a measure enthusiasm cannot game is the job.","ky":"/ai-keynote-speaker-for-hr-and-chro-conferences"},{"t":"page","n":"The Verifier's Discount","u":"/research/the-verifiers-discount","f":"","k":"The named concepts","view":"the fall in the perceived value of human work once the machine produces and the person checks, so that verification is priced below production even where it takes more expertise. SuperSkills (Kogan Page, 2026)…","adv":"","ky":"/ai-keynote-speaker-for-hr-and-chro-conferences"},{"t":"page","n":"The Unclaimed Hour","u":"/research/the-unclaimed-hour","f":"","k":"The named concepts","view":"the capacity an AI tool creates that nobody deliberately decides how to spend, so it is absorbed by whatever was already there. SuperSkills (Kogan Page, 2026) uses the term in this sense. No claim of first use…","adv":"If you cannot name what the saved time became, you did not save time. You changed the composition of the work, which may be fine and is a different claim from the one in the business case. Auditing where the hours went is the engagement.","ky":""},{"t":"page","n":"Outsourced recognition","u":"/research/outsourced-recognition","f":"","k":"The named concepts","view":"what happens when the expression of noticing another person is delegated to a machine, so the words arrive without the seeing that used to produce them. Rahim Hirji's term, confirmed in use since 25 January…","adv":"When the thank-you, the apology and the feedback are all drafted by a machine, the organisation loses the only artefacts that carry the message that somebody was seen. Deciding which of those stays human is a leadership question rather than a policy one.","ky":""},{"t":"page","n":"The AI Readiness Lie","u":"/research/ai-readiness-lie","f":"Why most AI transformation stalls, and what works instead: the readiness chain, transformation as a quarterly practice, and the recurring challenges.","k":"The named concepts","view":"","adv":"Most readiness assessments measure enthusiasm, and mine measures what your people can still do with the tools switched off.","ky":"/ai-keynote-speaker-for-hr-and-chro-conferences"},{"t":"page","n":"What is the AI Memo Test?","u":"/research/what-is-the-ai-memo-test","f":"Three questions Hirji published on 11 May 2025 after four chief executives sent AI memos in a month. Seventeen months on, one firm walked its memo back, one cut about 30 per cent of its workforce, and two held position.…","k":"The named concepts","view":"a three-question self-assessment of whether an organisation has decided how AI changes its work, scored on tool fluency as a requirement, job descriptions written around what software already does, and…","adv":"","ky":""},{"t":"page","n":"What board oversight of AI actually looks like","u":"/research/what-board-oversight-of-ai-looks-like","f":"What the frameworks actually require and what belongs on a quarterly agenda. Article 14 treats the capability to override as the content of oversight rather than the presence of a person. NIST expects reversal…","k":"Judgement, oversight and accountability","view":"","adv":"Most board papers report adoption. Almost none reports who owns the capability floor, or what a week without the system would cost. Putting that item on the agenda properly is the engagement.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"Does AI weaken human judgement?","u":"/research/ai-and-human-judgement","f":"The flagship review. 106 experiments where human and AI pairs did worse than either alone, endoscopists losing six points of unassisted detection, students scoring lower once the tool was removed, and where AI…","k":"Judgement, oversight and accountability","view":"","adv":"","ky":"/speaker-on-ai-and-human-judgement"},{"t":"page","n":"Is human intuition better than AI logic?","u":"/research/ai-and-human-intuition","f":"Kahneman and Klein settled when intuition can be trusted: a learnable environment, and prolonged practice in it with fast clear feedback. AI barely touches the first condition and removes the second, because the…","k":"Judgement, oversight and accountability","view":"","adv":"","ky":""},{"t":"page","n":"Should we trust AI over human experts?","u":"/research/ai-and-expert-judgement","f":"Who actually gains from AI advice. Across 140 radiologists the effect ran from strongly positive to strongly negative and nothing predicted which; across 5,172 support agents the gains went to the least skilled while…","k":"Judgement, oversight and accountability","view":"","adv":"","ky":""},{"t":"page","n":"What happens when AI and human judgement conflict?","u":"/research/ai-and-human-disagreement","f":"What happens when a person and a system disagree. Access to a system raised agreement from 58.4 to 80.9 per cent and cut accuracy from 74.2 to 63.9; only 5 per cent of model answers carry any marker of doubt, and the…","k":"Judgement, oversight and accountability","view":"","adv":"","ky":""},{"t":"page","n":"Is AI dangerous?","u":"/research/is-ai-dangerous","f":"Three questions asked as one. Fabricated references now reach 1 in 277 papers and endoscopists averaging 27.6 years of experience lost six points of unassisted detection, while the extinction estimates move from 5 to 10…","k":"Judgement, oversight and accountability","view":"","adv":"","ky":""},{"t":"page","n":"Should AI development be paused or slowed?","u":"/research/should-ai-development-be-paused","f":"","k":"Judgement, oversight and accountability","view":"","adv":"The pause an organisation controls is the one before it hands over a decision. Writing that page with the executive team, with names against each decision, is the engagement.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"What is an AI kill switch, and would one work?","u":"/research/what-is-an-ai-kill-switch","f":"","k":"Judgement, oversight and accountability","view":"","adv":"For every system already deployed, who can stop it, whether they would, and how they would know they needed to. Writing that list with the executive team, and rehearsing one stop, is the engagement.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"What do the rogue AI agent incidents mean for an organisation using AI?","u":"/research/what-the-rogue-ai-agent-incidents-mean-for-your-organisation","f":"","k":"Judgement, oversight and accountability","view":"","adv":"Scope, stop, monitoring on by default and a named owner, settled before the first agent rather than after the first incident. Settling them for one deployment is the engagement.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"Do AI models know when they are being tested?","u":"/research/do-ai-models-know-when-they-are-being-tested","f":"","k":"Judgement, oversight and accountability","view":"","adv":"Oversight that does not depend on the subject behaving as if unwatched: unannounced sampling, a measured disagreement rate, an unaided baseline. Designing that for one function is the engagement.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"What should a board do about the AI safety warnings?","u":"/research/what-should-a-board-do-about-the-ai-safety-warnings","f":"","k":"Judgement, oversight and accountability","view":"","adv":"The five decisions that hold whatever the odds turn out to be, written down with names against them. Getting a board to that page in one session is the engagement.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"Is it ethical to let AI judge people?","u":"/research/is-it-ethical-to-let-ai-judge-people","f":"Four obligations that survive being correct: an explanation the person can use, a contest somebody can win, a named human who carries it, and verification on the population it is used on. A widely deployed sepsis model…","k":"Judgement, oversight and accountability","view":"","adv":"","ky":""},{"t":"page","n":"Can AI be unbiased, or does it reproduce human bias?","u":"/research/can-ai-be-unbiased","f":"It reproduces human bias at measurable scale, and unbiased is not one target: several reasonable fairness criteria are provably incompatible. The first bias audit law in the world produced published audits from about 5…","k":"Judgement, oversight and accountability","view":"","adv":"","ky":""},{"t":"page","n":"Human and AI decision making","u":"/research/human-ai-decision-making","f":"","k":"Judgement, oversight and accountability","view":"","adv":"","ky":"/speaker-on-ai-and-human-judgement"},{"t":"page","n":"Decision Quality in the AI Era","u":"/research/decision-quality","f":"","k":"Judgement, oversight and accountability","view":"","adv":"","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"AI agents and human judgement","u":"/research/ai-agents-and-human-judgement","f":"When AI can plan and act, the human job becomes deciding what to delegate and owning the result. Agents, human at the start, and who is on the hook.","k":"Judgement, oversight and accountability","view":"","adv":"Before you deploy agents is the cheapest moment to decide this, and the only one where you are not deciding under pressure. Settling what an agent may do unasked, and who answers for it, is the work.","ky":"/ai-keynote-speaker-on-ai-agents-and-accountability"},{"t":"page","n":"Should I let an AI agent act on my behalf?","u":"/research/should-i-let-an-ai-agent-act-on-my-behalf","f":"Not a trust question. A system that answers gives you something to reject; a system that acts has already done it. Every oversight model assumes a pause that agents remove.","k":"Judgement, oversight and accountability","view":"What is the worst thing this can do before a human sees it, and can I live with that? Answer that and the trust question resolves itself. Leave it unanswered and no amount of confidence in the model helps you,…","adv":"","ky":"/ai-keynote-speaker-on-ai-agents-and-accountability"},{"t":"page","n":"Human in the loop is not a safeguard","u":"/research/human-in-the-loop-is-not-a-safeguard","f":"","k":"Judgement, oversight and accountability","view":"","adv":"Most human checkpoints would not catch the error they exist for. Some would, and those are worth defending. Replacing the rest with a check that does something is the work.","ky":"/ai-keynote-speaker-on-human-oversight-and-accountability"},{"t":"page","n":"What is meaningful human oversight?","u":"/research/what-is-meaningful-human-oversight","f":"Article 14 of the EU AI Act came into force on 2 August 2026 and names automation bias in the legislation. The five things an overseer must be enabled to do, why most arrangements fail the test, and why it is a…","k":"Judgement, oversight and accountability","view":"supervision by a person who understands the system well enough to spot when it is wrong, and who has the authority and the practical ability to stop it. Four conditions have to hold together: understanding its…","adv":"The whole test is whether your overseers can detect a wrong answer and have the authority to act on it. The answer is uncomfortable more often than not, and finding out is the job.","ky":"/ai-keynote-speaker-on-human-oversight-and-accountability"},{"t":"page","n":"Who owns verification when AI does the work?","u":"/research/who-owns-verification-when-ai-does-the-work","f":"In most organisations, nobody. It is not in a job description, a budget line or an org chart. The capability test that settles whether verification is real, and four ownership models with their actual costs.","k":"Judgement, oversight and accountability","view":"Could the person verifying this have produced it themselves, well enough to notice if it were wrong?If no, the verification is decorative. It should be recorded as absent rather than as satisfied, because…","adv":"Verification with a named owner who survives that person leaving is rare. Assigning it properly, so it holds when somebody resigns, is what I come in to do.","ky":"/ai-keynote-speaker-on-human-oversight-and-accountability"},{"t":"page","n":"Who supervises work they cannot do themselves?","u":"/research/who-supervises-work-they-cannot-do","f":"A person three years into a career reviews AI-generated work of a kind they have never produced, and signs it off. The organisation records a control as satisfied. Nothing has been checked.","k":"Judgement, oversight and accountability","view":"Could the person supervising this have produced it themselves, well enough to notice if it were wrong? Where the answer is no, supervision has become approval. The distinction is invisible in every management…","adv":"This arrives quietly and is usually well advanced before anyone names it, so the job is finding where it has already taken hold.","ky":"/ai-keynote-speaker-on-early-careers-and-graduate-talent"},{"t":"page","n":"When should I override AI?","u":"/research/when-should-i-override-ai","f":"","k":"Judgement, oversight and accountability","view":"Specify in advance the grounds on which a person may override the system, the grounds on which they must defer to it, and who carries the consequence in each case. Unspecified, this collapses into whichever…","adv":"Six conditions say override and three say defer. All nine rest on a capability precondition most organisations have never tested. Establishing whether your people could exercise the override is the engagement.","ky":""},{"t":"page","n":"How do I know when AI is wrong?","u":"/research/how-do-i-know-when-ai-is-wrong","f":"","k":"Judgement, oversight and accountability","view":"","adv":"","ky":"/speaker-on-ai-and-human-judgement"},{"t":"page","n":"How do you audit an AI-assisted decision?","u":"/research/how-do-you-audit-an-ai-assisted-decision","f":"","k":"Judgement, oversight and accountability","view":"","adv":"Could you reconstruct how a decision was reached six months from now, to somebody hostile? Building a trail that holds up when it matters is the job.","ky":"/ai-keynote-speaker-on-human-oversight-and-accountability"},{"t":"page","n":"What should a board ask about AI?","u":"/research/what-should-a-board-ask-about-ai","f":"","k":"Judgement, oversight and accountability","view":"Not as an agenda item. Ask two or three, in the ordinary course of reviewing something else, and listen for hesitation rather than for content. Questions asked as a set invite a prepared answer; questions…","adv":"Boards rarely get straight answers to these questions, because the questions are rarely asked well. I either prepare the answers or prepare the board to ask better.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites","alt":["board questions ai","questions for the board","board oversight of ai"]},{"t":"page","n":"The Delegation Boundary Map","u":"/research/delegation-boundary-map","f":"","k":"Judgement, oversight and accountability","view":"A loop implies the human is positioned somewhere in a circle. Work is not a circle. It is a sequence in which authority can be handed over at any point, and the consequences of handing it over are completely…","adv":"The map is worth nothing drawn by the people who describe the work rather than the people who do it. Running it with the second group is the engagement.","ky":"/ai-keynote-speaker-on-human-oversight-and-accountability"},{"t":"page","n":"Who can override an AI system?","u":"/research/who-can-override-an-ai-system","f":"Oversight without authority is ceremony. What Article 14 actually requires, the one paragraph that names competence, training and authority together, and why overriding costs something while deferring is free.","k":"Judgement, oversight and accountability","view":"","adv":"Has anyone in your organisation actually overridden a system recently, and was it recorded? Establishing who genuinely holds this, rather than who is named in the policy, is the work.","ky":"/ai-keynote-speaker-on-human-oversight-and-accountability"},{"t":"page","n":"Does explaining an AI's reasoning help?","u":"/research/does-explaining-an-ai-decision-help","f":"NIST separates explanation accuracy from decision accuracy: a true account of how a system reached a wrong answer is an ordinary outcome. Why a reason can substitute for a check, and the three things an explanation is…","k":"Judgement, oversight and accountability","view":"","adv":"An explanation can substitute for a check, which is the failure mode most explainability programmes are built to produce. Designing oversight that survives a fluent account of a wrong answer is the work.","ky":""},{"t":"page","n":"How should AI decision rights be allocated?","u":"/research/how-should-ai-decision-rights-be-allocated","f":"Doing the work of deciding and holding the right to decide were always different. What the regulation already allocates between provider and deployer, and why an allocation arrived at by drift cannot be reviewed.","k":"Judgement, oversight and accountability","view":"","adv":"Almost nobody names who decides what, in writing, before an incident forces it, and doing that is what I am for.","ky":"/ai-keynote-speaker-on-human-oversight-and-accountability"},{"t":"page","n":"Which decisions should become slower because of AI?","u":"/research/which-decisions-should-become-slower-because-of-ai","f":"Cognitive forcing cut overreliance in a 199-participant experiment, and the designs that worked best were the ones people rated worst. Article 14(5) already mandates a two-person check for one category. Four conditions…","k":"Judgement, oversight and accountability","view":"","adv":"","ky":"/speaker-on-ai-and-human-judgement"},{"t":"page","n":"When agents become part of the workforce, who manages them?","u":"/research/who-manages-ai-agents","f":"Article 26 already names the person: competence, training and authority, plus the power to suspend. The gap sits underneath it. The FAccT visibility paper states we lack methods for determining when an agent has created…","k":"Judgement, oversight and accountability","view":"","adv":"","ky":"/ai-keynote-speaker-on-ai-agents-and-accountability"},{"t":"page","n":"How do you design a stop button people will actually use?","u":"/research/how-do-you-design-a-stop-button-people-will-use","f":"Article 14(4)(e) requires the button and decides nothing about whether it gets pressed. One ICU study annotated 12,671 arrhythmia alarms and found 88.8 per cent false. A regulator has on record that stop-work authority…","k":"Judgement, oversight and accountability","view":"","adv":"Article 14 requires the button and decides nothing about whether anybody presses it. Designing the authority, the threshold and the cover so that somebody actually does is the engagement.","ky":""},{"t":"page","n":"How should an AI agent communicate uncertainty to a human?","u":"/research/how-should-an-ai-agent-communicate-uncertainty","f":"Two problems, usually confused. GPT-3's verbalised confidence carries an expected calibration error of 0.52, and readers put \"very likely\" at 62 per cent where the guidelines mean above 90. Giving readers a translation…","k":"Judgement, oversight and accountability","view":"","adv":"Readers put \"very likely\" at 62 per cent where the guideline means above 90, so an honest confidence signal can still mislead. Setting the vocabulary your organisation will use, once, is the work.","ky":""},{"t":"page","n":"Can a human approve an AI decision at machine speed?","u":"/research/can-a-human-approve-an-ai-decision-at-machine-speed","f":"Only when the machine proposes no faster than a person can check, and most deployments have measured neither. A congressional hearing, a Spanish regulator's first agent-executed breach and a US-China proposal reached…","k":"Judgement, oversight and accountability","view":"","adv":"For each deployed system, the decisions per hour, the minutes a real check takes and the attention available, compared, with a declared sampling rate and a named owner for whatever is not checked. Doing that arithmetic for the three highest-consequence…","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"Does using AI stop you learning?","u":"/research/does-using-ai-stop-you-learning","f":"No, and delegating the work to it does. Four designs that withdraw the assistance before measuring split on the same line: 52 developers at 50 per cent against 67, 1,222 participants losing persistence inside ten…","k":"Thinking, learning and capability","view":"the process by which practice turns into capability a person still holds when the help is removed. It is measured by taking the help away and testing unaided, so assisted output tells you almost nothing about…","adv":"","ky":""},{"t":"page","n":"AI and critical thinking","u":"/research/ai-and-critical-thinking","f":"","k":"Thinking, learning and capability","view":"","adv":"","ky":"/speaker-on-ai-and-human-judgement"},{"t":"page","n":"How humans learn with AI","u":"/research/how-humans-learn-with-ai","f":"","k":"Thinking, learning and capability","view":"","adv":"","ky":""},{"t":"page","n":"Using AI without dependency","u":"/research/using-ai-without-dependency","f":"","k":"Thinking, learning and capability","view":"","adv":"","ky":""},{"t":"page","n":"Am I becoming dependent on AI?","u":"/research/am-i-becoming-dependent-on-ai","f":"","k":"Thinking, learning and capability","view":"Could you do this task to an acceptable standard, unaided, today? Not five years ago. Not in principle. Today. Most people have never asked, because performance with the tool is fine and there is no moment…","adv":"","ky":""},{"t":"page","n":"Does AI make everyone think alike?","u":"/research/does-ai-make-everyone-think-alike","f":"","k":"Thinking, learning and capability","view":"the narrowing of the range of what a population produces or thinks, caused by many people drawing on the same source of suggestions, even where each person's own output improves. It is measured at the level of…","adv":"","ky":"/speaker-on-ai-and-human-judgement"},{"t":"page","n":"How do I get AI to challenge me rather than agree with me?","u":"/research/how-do-i-get-ai-to-challenge-me","f":"","k":"Thinking, learning and capability","view":"","adv":"","ky":"/speaker-on-ai-and-human-judgement"},{"t":"page","n":"How do I keep my own voice when using AI?","u":"/research/how-do-i-keep-my-own-voice-when-using-ai","f":"","k":"Thinking, learning and capability","view":"","adv":"","ky":""},{"t":"page","n":"Is attention a trainable skill?","u":"/research/is-attention-a-trainable-skill","f":"","k":"Thinking, learning and capability","view":"","adv":"","ky":""},{"t":"page","n":"Assessing students when AI can do the assignment","u":"/research/how-to-assess-students-when-ai-can-do-the-assignment","f":"","k":"Thinking, learning and capability","view":"","adv":"","ky":"/ai-keynote-speaker-for-schools-and-education"},{"t":"page","n":"Does AI detection work?","u":"/research/does-ai-detection-work","f":"Not well enough to accuse anyone. Detectors flagged more than half of essays by non-native English speakers as AI-generated while classifying US eighth-grade essays almost perfectly. Plus the base-rate arithmetic nobody…","k":"Thinking, learning and capability","view":"","adv":"","ky":"/ai-keynote-speaker-for-schools-and-education"},{"t":"page","n":"Should children use AI?","u":"/research/should-children-use-ai","f":"","k":"Thinking, learning and capability","view":"","adv":"","ky":""},{"t":"page","n":"What stays human","u":"/research/what-stays-human","f":"","k":"Thinking, learning and capability","view":"","adv":"","ky":"/ai-keynote-speaker-for-corporate-conferences"},{"t":"page","n":"Human skills in the age of AI","u":"/research/human-skills-in-the-age-of-ai","f":"","k":"Thinking, learning and capability","view":"","adv":"","ky":"/ai-keynote-speaker-for-corporate-conferences"},{"t":"page","n":"Can you regain a skill you have lost?","u":"/research/can-you-regain-a-skill-you-have-lost","f":"","k":"Thinking, learning and capability","view":"","adv":"","ky":""},{"t":"page","n":"How do you assess capability rather than output?","u":"/research/how-do-you-assess-capability-rather-than-output","f":"","k":"Thinking, learning and capability","view":"","adv":"Separating the work from the worker's capability is the harder half, and almost no performance process does it. Building an assessment of what people can still do unaided is the engagement.","ky":"/ai-keynote-speaker-for-hr-and-chro-conferences"},{"t":"page","n":"What professions can learn from aviation","u":"/research/what-professions-can-learn-from-aviation","f":"","k":"Thinking, learning and capability","view":"","adv":"","ky":""},{"t":"page","n":"Does GPS damage your brain?","u":"/research/does-gps-damage-your-brain","f":"The London taxi studies measured acquisition: four years of the Knowledge, structural change only in the trainees who qualified. The removal side is one behavioural study with 13 people at follow-up and no scanner.…","k":"Thinking, learning and capability","view":"the examination a London taxi driver must pass to hold a green badge, introduced in 1865 and run by Transport for London. Candidates learn the 320 routes in the Blue Book, within a six mile radius of Charing…","adv":"","ky":""},{"t":"page","n":"What is the human signal?","u":"/research/what-is-the-human-signal","f":"The trace of a mind inside a piece of work, named on 30 November 2025, with three tests a reader applies without being asked: was a real decision made, was something difficult carried with care, is there the risk of a…","k":"Thinking, learning and capability","view":"the trace of a mind inside a piece of work, the sense a reader, viewer or listener gets that a real person made decisions that mattered.","adv":"","ky":""},{"t":"page","n":"Is it still my idea if AI helped me write it?","u":"/research/is-it-still-my-idea-if-ai-helped-me-write-it","f":"The credit question is the easy half. Four randomised experiments show writing assistance moving the writer's own views: 1,506 participants whose opinions followed their tool, and 80 per cent who said the advice had not…","k":"Thinking, learning and capability","view":"the effect by which a writing assistant configured to favour one view shifts what a person writes and, with it, what that person goes on to believe. Named by Jakesch and colleagues in 2023, who measured it…","adv":"","ky":""},{"t":"page","n":"How good are you at using AI?","u":"/research/how-good-are-you-at-using-ai","f":"Five yes-or-no questions and a banded reading at 0-1, 2-3 and 4-5, put to live audiences before it was written down. Question three is the only one you cannot acquire in an afternoon. Unvalidated, and the page says so…","k":"Thinking, learning and capability","view":"1. Have you ever set up a Project, a Gem or saved instructions, so it knows who you are without being told? 2. Have you ever run a deep research task, waited twenty minutes, and read what came back? 3. Have…","adv":"","ky":""},{"t":"page","n":"Who AI leaves behind","u":"/research/who-ai-leaves-behind","f":"AI's gains land on the least skilled, in customer support and in taxi driving. The detectors built to police it misclassify non-native English writers at 61.22 per cent. The same people the tool helps are the ones the…","k":"Work, careers and the labour market","view":"","adv":"The tool helps the least skilled most. The checking regime falls hardest on the same people, since detectors misclassify non-native English writers at 61.22 per cent. For an inclusion agenda, that pairing is the item.","ky":""},{"t":"page","n":"Proving you did the work","u":"/research/proving-you-did-the-work","f":"Detection misclassifies over half of non-native English essays, a 61.22 per cent false positive rate, while labelling a reply as AI removes its advantage even where it beat humans. Why both detection and blanket…","k":"Work, careers and the labour market","view":"","adv":"","ky":""},{"t":"page","n":"The mid-career squeeze","u":"/research/the-mid-career-squeeze","f":"The measured displacement is at 22 to 25, through reduced hiring. Mid-career exposure is different: AI's gains land on the least experienced, compressing the gap a mid-career salary pays for. Plus why a randomised trial…","k":"Work, careers and the labour market","view":"","adv":"","ky":""},{"t":"page","n":"Will AI replace my job?","u":"/research/will-ai-replace-my-job","f":"","k":"Work, careers and the labour market","view":"","adv":"","ky":""},{"t":"page","n":"Will AI replace entry-level jobs?","u":"/research/will-ai-replace-entry-level-jobs","f":"","k":"Work, careers and the labour market","view":"","adv":"","ky":"/ai-keynote-speaker-on-early-careers-and-graduate-talent","alt":["graduate jobs","junior jobs","entry level roles"]},{"t":"page","n":"Which jobs are safest from AI?","u":"/research/which-jobs-are-safest-from-ai","f":"","k":"Work, careers and the labour market","view":"","adv":"","ky":""},{"t":"page","n":"What is the AI employment gap?","u":"/research/what-is-the-ai-employment-gap","f":"Employment of 22 to 25 year olds in AI-exposed occupations sits about 19 per cent below where it would have been had it tracked their less-exposed peers. The 19 is the distance between a fall of 11 and a growth of 10,…","k":"Work, careers and the labour market","view":"the Stanford Digital Economy Lab finding that employment of workers aged 22 to 25 in AI-exposed occupations sits about 19 per cent below where it would have been had it tracked their less-exposed peers,…","adv":"","ky":""},{"t":"page","n":"Does AI actually make people more productive?","u":"/research/does-ai-actually-make-people-more-productive","f":"Large measured gains on narrow tasks: writing 40 per cent faster, a standardised coding task 55.8 per cent faster, support resolutions up 15 per cent an hour. Scattered or negative in real work. Nothing yet in the…","k":"Work, careers and the labour market","view":"On well-specified tasks with a clear finish line, generative AI produces large measured gains. In continuing work inside systems people already know, the measured effect is uncertain and sometimes negative.…","adv":"","ky":"","alt":["ai productivity","productivity gains","does ai save time"]},{"t":"page","n":"Will AI replace programmers?","u":"/research/will-ai-replace-programmers","f":"No study shows replacement. The two most cited coding experiments sit seventy-five points apart because one built something new and the other changed a system somebody already knew. The measured risk is to how…","k":"Work, careers and the labour market","view":"No evidence supports replacement. The measured gains concentrate in producing new code against a clear specification, and disappear or reverse in continuing work inside a system somebody already knows. What is…","adv":"","ky":""},{"t":"page","n":"Should juniors use AI at all?","u":"/research/should-juniors-use-ai","f":"Yes, and the evidence says which version of use is safe. The three experiments that removed the tool afterwards, and the four rules that follow. Plus four for whoever manages them, because the burden is in the wrong…","k":"Work, careers and the labour market","view":"","adv":"","ky":"/ai-keynote-speaker-on-early-careers-and-graduate-talent"},{"t":"page","n":"Do apprenticeships still work?","u":"/research/do-apprenticeships-still-work","f":"Germany has record unplaced applicants and 54,400 empty places in the same year. England shortened the apprenticeship and its flagship replacement drew 74 starts against a target of 1,000. Four systems read at source,…","k":"Work, careers and the labour market","view":"","adv":"","ky":"/ai-keynote-speaker-on-early-careers-and-graduate-talent"},{"t":"page","n":"Should I still learn to code?","u":"/research/should-i-still-learn-to-code","f":"","k":"Work, careers and the labour market","view":"","adv":"","ky":""},{"t":"page","n":"Staying valuable in the age of AI","u":"/research/staying-valuable-in-the-age-of-ai","f":"Where human value moves as AI spreads: staying valuable, the entry-level question, and the human skills that matter most.","k":"Work, careers and the labour market","view":"","adv":"","ky":"/ai-keynote-speaker-for-corporate-conferences"},{"t":"page","n":"What should I tell my children to study?","u":"/research/what-should-i-tell-my-children-to-study","f":"","k":"Work, careers and the labour market","view":"","adv":"","ky":""},{"t":"page","n":"Why \"learn to prompt\" is weak career advice","u":"/research/why-learn-to-prompt-is-weak-career-advice","f":"","k":"Work, careers and the labour market","view":"","adv":"","ky":""},{"t":"page","n":"Four Generations of Disruption","u":"/research/four-generations-and-ai","f":"","k":"Work, careers and the labour market","view":"","adv":"Four generations are being handed the same tool and the same training, and the risk is not the same for any two of them.","ky":""},{"t":"page","n":"Who captures the productivity gains from AI?","u":"/research/who-captures-the-productivity-gains-from-ai","f":"","k":"Work, careers and the labour market","view":"","adv":"A technology can raise output and lower wages at the same time, and which happens is decided by two mechanisms rather than by the technology. Working out which one your operating model selects for is the engagement.","ky":""},{"t":"page","n":"Which tasks do workers not want automated?","u":"/research/which-tasks-do-workers-not-want-automated","f":"Asked task by task about their own occupations, 1,500 US workers were positive about automating 46.1 per cent of them. Where they refused, distrust of accuracy outranked fear of replacement by two to one, and workers…","k":"Work, careers and the labour market","view":"","adv":"","ky":"/ai-keynote-speaker-for-hr-and-chro-conferences"},{"t":"page","n":"Is deskilling real, or a rescaling of what counts as skill?","u":"/research/is-deskilling-real","f":"Deskilling and rescaling have both been measured, in different people. Nineteen Polish endoscopists lost six percentage points of unassisted detection; a Japanese taxi fleet's skill gap closed by 14 per cent. The strong…","k":"Work, careers and the labour market","view":"the position that AI changes which skills carry value instead of removing skill, so what looks like deskilling is a labour market revaluing capability. Put better than anyone else has by Roop Bhadury of the…","adv":"","ky":"","alt":["deskilling","skill loss","are we getting worse at our jobs"]},{"t":"page","n":"How do juniors become senior if AI does the junior work?","u":"/research/how-do-juniors-become-senior","f":"Junior work was a by-product of senior workload, not a training scheme, so AI removes the reason it existed. Access is not the harm: in randomised trials with 1,222 people the withdrawal effect appeared after ten…","k":"Work, careers and the labour market","view":"the ability to judge whether work is right without being told, built by doing the work and being wrong about it under supervision. AI can produce the work but not the being wrong.","adv":"","ky":""},{"t":"page","n":"The shape of the organisation after AI","u":"/research/the-shape-of-the-organisation-after-ai","f":"Acemoglu puts ten-year productivity gains under 0.66 per cent, Danish records find precise nulls on pay two years in, and US payroll data rules out widespread displacement. What the evidence supports about redesign and…","k":"Organisations and leadership","view":"","adv":"","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"What AI does to a team","u":"/research/what-ai-does-to-a-team","f":"Everyone improves and the room converges. Doshi and Hauser on individual creativity rising while collective diversity falls, Dell'Acqua on AI flattening the difference between an engineer and a marketer, and the…","k":"Organisations and leadership","view":"","adv":"Everyone improves and the room converges, which arrives as a productivity gain and a diversity loss on the same dashboard.","ky":""},{"t":"page","n":"Who should own AI strategy in an organisation?","u":"/research/who-should-own-ai-strategy","f":"Ownership is decided by inheritance rather than argument, and it silently answers a bigger question: augmentation or replacement. Autor and Thompson on why which tasks you automate matters more than how many, what each…","k":"Organisations and leadership","view":"","adv":"","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"AI workforce strategy","u":"/research/ai-workforce-strategy","f":"","k":"Organisations and leadership","view":"","adv":"Very few workforce plans name which capabilities the organisation intends to keep. Turning that into something a CHRO can put in front of a board is the engagement.","ky":"/ai-keynote-speaker-for-hr-and-chro-conferences"},{"t":"page","n":"The CHRO guide to AI","u":"/research/chro-guide-to-ai","f":"","k":"Organisations and leadership","view":"","adv":"Working through this with your own numbers, rather than with a generic maturity model, is what makes it usable. That session is the engagement.","ky":"/ai-keynote-speaker-for-hr-and-chro-conferences"},{"t":"page","n":"How leaders should respond to AI","u":"/research/how-should-leaders-respond-to-ai","f":"","k":"Organisations and leadership","view":"","adv":"A position you have said out loud once is worth more than a strategy deck nobody reads. Turning that position into a working practice, privately, is what I do.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"What does AI literacy mean for leaders?","u":"/research/what-does-ai-literacy-mean-for-leaders","f":"A legal obligation since February 2025, enforced since this month, at every risk tier. What Article 4 requires, the clause almost every programme misses, and why a completed training module is not evidence of capability.","k":"Organisations and leadership","view":"","adv":"Literacy here means knowing where the tools fail, not knowing what they are. Turning that into something you can run for an executive team is the engagement.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"How do you write an AI use policy that works?","u":"/research/how-do-you-write-an-ai-use-policy-that-works","f":"Most are unenforceable and everyone involved knows it. Six elements that survive every model release, what EU law now makes auditable, and the one-afternoon test that beats legal review.","k":"Organisations and leadership","view":"","adv":"Most policies satisfy legal and then get ignored. Writing one people actually follow is the harder job and the one I take.","ky":""},{"t":"page","n":"How do you measure AI adoption properly?","u":"/research/how-do-you-measure-ai-adoption-properly","f":"","k":"Organisations and leadership","view":"","adv":"A dashboard that counts logins will keep looking healthy while capability falls, and building the other kind is the engagement.","ky":"/ai-keynote-speaker-for-hr-and-chro-conferences"},{"t":"page","n":"Why reskilling programmes mostly fail","u":"/research/why-reskilling-programmes-mostly-fail","f":"A position page against the near-universal institutional answer to AI. Five structural reasons, the evidence stated honestly including where it is indirect, and the published data that would change the position.","k":"Organisations and leadership","view":"","adv":"Ask what your last programme changed. If the answer is what people had attended rather than what they could do, it failed in the ordinary way. Designing one that does not is the work.","ky":"/ai-keynote-speaker-for-hr-and-chro-conferences"},{"t":"page","n":"AI Transformation Is Not a Change-Management Problem","u":"/research/ai-transformation-not-change-mangement-problem","f":"","k":"Organisations and leadership","view":"","adv":"Programmes run as change management measure adoption and miss the thing at stake. Working out what that means for how yours runs is where I would start.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"Common AI Transformation Challenges","u":"/research/common-ai-transformation-challenges","f":"","k":"Organisations and leadership","view":"","adv":"The loudest problem is rarely the real one, and the cost of fixing the wrong one is a year. Telling them apart is the work.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"Should AI attend my meetings?","u":"/research/should-ai-attend-my-meetings","f":"","k":"Organisations and leadership","view":"","adv":"A complete record changes what people say in the room, and that cost never appears in the business case for the tool. Deciding which meetings produce a record and which produce understanding is the work.","ky":""},{"t":"page","n":"The Third Way","u":"/research/neither-ai-hype-nor-doom","f":"","k":"Organisations and leadership","view":"","adv":"","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"How do you keep expertise in an organisation?","u":"/research/how-do-you-keep-expertise-in-an-organisation","f":"Documentation preserves what experts can say, and most of what they know is not that. Polanyi on the tacit part, Lave and Wenger on how it transfers, and why the work juniors learned from is the work most easily…","k":"Organisations and leadership","view":"","adv":"Name the three people whose departure would hurt most. Then say precisely why, without using the word experience. Closing the gap between those two answers is the work.","ky":"/ai-keynote-speaker-for-hr-and-chro-conferences"},{"t":"page","n":"What happens to institutional memory?","u":"/research/what-happens-to-institutional-memory","f":"Retrieval is not retention. An organisation can get better at finding things while getting worse at knowing them, because the reasons, the rejected options and the trust calibration were never in the archive.","k":"Organisations and leadership","view":"","adv":"Handovers capture status and lose the reasoning, which is the part that mattered. Auditing what your organisation would forget is the job.","ky":""},{"t":"page","n":"What happens when AI removes the visible work?","u":"/research/the-invisible-work-of-oversight","f":"Bainbridge, 1983: taking away the easy parts of a task can make the difficult parts harder. Why time saved prices only what was removed, why task productivity and workload are different variables, and what responsible…","k":"Organisations and leadership","view":"","adv":"Oversight is almost never counted as work, so it is done in the gaps between other tasks and nobody has budgeted for it. Making it visible enough to resource is the engagement.","ky":"/ai-keynote-speaker-on-human-oversight-and-accountability"},{"t":"page","n":"What should we tell employees about AI and headcount?","u":"/research/what-should-we-tell-employees-about-ai-and-headcount","f":"Which decisions a machine may make in the organisation's name, and what will and will not be done with the time it saves, declared by domain and dated. Deloitte's 25,000-worker UK survey found 31 per cent concealing use…","k":"Organisations and leadership","view":"","adv":"A written position on decision rights and on headcount, declared by domain and dated, with the amnesty period that lets the organisation see what is being delegated. Drafting both with the executive team and the people function, and testing them against the…","ky":"/ai-keynote-speaker-for-hr-and-chro-conferences"},{"t":"page","n":"What is a Shared Prompt Review?","u":"/research/what-is-the-shared-prompt-review","f":"Four things on the table once a week: the prompt as typed, the raw output, what a person cut and why, and where it might be wrong. A field experiment with 776 professionals found AI erasing the difference between what a…","k":"Organisations and leadership","view":"a short, structured team conversation that examines how AI was used on a piece of work, covering the exact prompt, the raw output, what a person kept or cut and why, and where the output may be wrong.","adv":"","ky":""},{"t":"page","n":"What is automating versus informating?","u":"/research/what-is-automating-versus-informating","f":"Zuboff, 1988. Automating replaces human judgement with a machine; informating makes the work more visible to the person doing it. The same model configured two ways, so output can improve while capability erodes with…","k":"Organisations and leadership","view":"automating replaces human judgement with a machine; informating generates information that deepens the worker's understanding. The same system can do either, and which one happens is a management choice rather…","adv":"","ky":""},{"t":"page","n":"What is a capability audit?","u":"/research/what-is-a-capability-audit","f":"A proposed method, labelled as one. What people can still do unaided, where expertise actually sits, what has no redundancy and what would be hard to rebuild. Why confidence and output are both broken instruments.","k":"Organisations and leadership","view":"","adv":"This page describes the instrument; running one on your own organisation, and surviving what it finds, is the engagement.","ky":"/ai-keynote-speaker-for-hr-and-chro-conferences"},{"t":"page","n":"When should an organisation reverse an AI deployment?","u":"/research/deployment-is-not-a-ratchet","f":"NIST's five conditions for deactivating a system, Perrow on why adding a safeguard can reduce safety, and why a fallback that exists only as a document is not a fallback.","k":"Organisations and leadership","view":"","adv":"Thresholds have to be set while everyone is still pleased with the system, because the people defending a decision cannot set them afterwards. Getting that timing right is the whole job.","ky":"/ai-keynote-speaker-on-ai-agents-and-accountability"},{"t":"page","n":"How do you preserve capability across vendors?","u":"/research/preserving-capability-across-vendors","f":"Prahalad and Hamel's 1990 warning, applied. Keep enough to specify the work, judge it, handle exceptions and replace the supplier. Three things that are genuinely new, and one that is not new at all.","k":"Organisations and leadership","view":"","adv":"Almost no procurement process asks what happens to your people when the vendor leaves. Writing capability into the contract before signing, rather than after, is the work.","ky":""},{"t":"page","n":"If every competitor has the same AI, where does the advantage come from?","u":"/research/if-everyone-has-ai-where-is-the-advantage","f":"A licence every rival can buy at list price passes one of Barney's four tests. US Census data puts firm use at 18 per cent with 57 per cent of adopters in three or fewer functions, so parity has not arrived, and the…","k":"Organisations and leadership","view":"","adv":"","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"How long should we give an AI investment before deciding whether it worked?","u":"/research/how-long-before-you-know-if-an-ai-investment-worked","f":"The J-curve says an early read understates and a late one overstates, for the same reason. Danish administrative records found precise nulls on pay two years in, next to heavy task reorganisation. Three questions, three…","k":"Organisations and leadership","view":"","adv":"","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"What becomes more valuable in a business as AI gets cheaper?","u":"/research/what-becomes-more-valuable-as-ai-gets-cheaper","f":"Three randomised trials found the same compression: the tool raises the floor far more than the ceiling. So the capability losing value fastest is being unusually good at exactly the work the model does well. Four…","k":"Organisations and leadership","view":"","adv":"Three randomised trials found the same compression: the tool raises the floor far more than the ceiling. Working out which of your roles is on the wrong side of that is a pricing question before it is a capability one.","ky":""},{"t":"page","n":"What happens to work whose purpose was moving information around?","u":"/research/what-happens-to-work-that-moves-information","f":"Garicano's model says a layer exists because matching problems to knowledge is costly. Resume data on 3,100 firms finds hierarchies flattening after AI adoption, on tests the authors call under-powered. What the…","k":"Organisations and leadership","view":"","adv":"","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"Which AI investments should we stop?","u":"/research/which-ai-investments-should-we-stop","f":"Between 30 and 40 per cent of information systems projects show some degree of escalation, and the founding case study of that literature was an expert system that ran for a decade. There is no credible non-vendor base…","k":"Organisations and leadership","view":"","adv":"","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"How do humans and agents divide work across a process?","u":"/research/how-do-humans-and-agents-divide-work","f":"Task-level allocation has been on the wall since 1951 and has never worked, because every assignment creates new work for the other party. The unanswered part is the boundary. Medicine calls those discontinuities gaps,…","k":"Organisations and leadership","view":"","adv":"Task-level allocation has failed since 1951. Every assignment creates new work for the other side, and the gaps are where the failures live. Settling the boundary, and who owns those gaps, is what I come in to do.","ky":""},{"t":"page","n":"Which professions face the greatest deskilling risk?","u":"/research/which-professions-face-the-greatest-deskilling-risk","f":"","k":"Professions and sectors","view":"","adv":"Exposure rankings mislead because they measure what a model can touch rather than what a person would stop practising. Scoring your own roles against the four conditions is the engagement.","ky":""},{"t":"page","n":"How will AI change medicine?","u":"/research/how-will-ai-change-medicine","f":"","k":"Professions and sectors","view":"","adv":"The first measured deskilling of experienced clinicians is now published. The most widely installed sepsis model failed external validation. A clinical board needs a different set of questions from the vendor's.","ky":""},{"t":"page","n":"How will AI change law?","u":"/research/how-will-ai-change-law","f":"","k":"Professions and sectors","view":"","adv":"The hallucination database now records more filings by people representing themselves than by lawyers. The English Divisional Court has ruled in Ayinde and Al-Haroun. For a firm, the work is a verification standard that survives a wasted costs application.","ky":""},{"t":"page","n":"How will AI change consulting?","u":"/research/how-will-ai-change-consulting","f":"","k":"Professions and sectors","view":"","adv":"A business model built on leverage does not survive the leverage becoming free, and what remains scarce afterwards is the session for a partnership.","ky":""},{"t":"page","n":"How will AI change accounting and audit?","u":"/research/how-will-ai-change-accounting-and-audit","f":"The oversight precedent everybody cites was replaced in April 2026, and its replacement puts generative AI expressly out of scope. Meanwhile the UK audit regulator found the six largest firms had not measured what their…","k":"Professions and sectors","view":"","adv":"","ky":"/ai-keynote-speaker-for-corporate-conferences"},{"t":"page","n":"How will AI change journalism?","u":"/research/how-will-ai-change-journalism","f":"Twenty-two broadcasters, 14 languages, 2,709 graded answers. Sourcing failed at 31 per cent against 20 per cent for accuracy, and the reputational cost arrives at the masthead that was cited rather than the assistant…","k":"Professions and sectors","view":"","adv":"Sourcing failed at 31 per cent against 20 for accuracy across 2,709 graded answers, and the reputational cost lands at the masthead that was cited. For a newsroom, the session is where verification has to sit.","ky":""},{"t":"page","n":"How will AI change customer service?","u":"/research/how-will-ai-change-customer-service","f":"The best-evidenced occupation there is, and the study watched the tool break. Gains ran to novices, the best agents got slightly worse, and the learning survived an outage only for the workers who had engaged with the…","k":"Professions and sectors","view":"","adv":"The best-evidenced occupation there is, and the study watched the tool break: the learning survived the outage only for agents who had engaged with the suggestions. Designing for that difference is the work.","ky":""},{"t":"page","n":"How will AI change teaching?","u":"/research/how-will-ai-change-teaching","f":"A school-randomised trial in England cut lesson planning time 31 per cent with no quality change a blinded panel could see. English teachers plan with it at 35 per cent and mark with it at 5 per cent, which is the…","k":"Professions and sectors","view":"","adv":"English teachers plan with it at 35 per cent and mark with it at 5, which is a profession drawing a line nobody asked it to draw.","ky":""},{"t":"page","n":"How will AI change the public sector?","u":"/research/how-will-ai-change-the-public-sector","f":"Both flagship UK figures, 26 minutes a day and 19, are self-reported, and one was calculated from tick-box midpoints with its top tail capped by the analysts. The sector's deepest precedent is not a productivity study,…","k":"Professions and sectors","view":"","adv":"Both flagship UK time-saving figures are self-reported, and one was calculated from tick-box midpoints with its top tail capped. Before a department scales on those numbers, somebody should say so out loud.","ky":""},{"t":"page","n":"How will AI change human resources?","u":"/research/how-will-ai-change-human-resources","f":"Retrieval models favoured White-associated names in 85.1 per cent of comparisons. The first algorithmic hiring audit law in the world produced published reports from 18 of 391 employers checked. And the function…","k":"Professions and sectors","view":"","adv":"","ky":"/ai-keynote-speaker-for-hr-and-chro-conferences"},{"t":"page","n":"Should AI remember everything about me?","u":"/research/should-ai-remember-everything-about-me","f":"","k":"Everyday life","view":"","adv":"","ky":""},{"t":"page","n":"Is it safe to use AI for therapy or advice?","u":"/research/is-it-safe-to-use-ai-for-therapy","f":"","k":"Everyday life","view":"","adv":"","ky":""},{"t":"page","n":"Should I use AI to write personal messages?","u":"/research/should-i-use-ai-to-write-personal-messages","f":"","k":"Everyday life","view":"","adv":"","ky":""},{"t":"page","n":"Should I let AI summarise everything I read?","u":"/research/should-i-let-ai-summarise-everything-i-read","f":"","k":"Everyday life","view":"","adv":"","ky":""},{"t":"page","n":"Do I still need to remember things?","u":"/research/do-i-still-need-to-remember-things","f":"","k":"Everyday life","view":"","adv":"","ky":""},{"t":"page","n":"How much should teenagers use AI?","u":"/research/how-much-should-teenagers-use-ai","f":"","k":"Everyday life","view":"","adv":"","ky":"/ai-keynote-speaker-for-schools-and-education"},{"t":"page","n":"How do I raise a child who thinks for themselves?","u":"/research/how-do-i-raise-a-child-who-thinks-for-themselves","f":"The only experiment that took the AI tutor away found unrestricted users 17 per cent below students who never had it, while the guardrailed group was largely spared. Configuration decided it. A rule about screen time…","k":"Everyday life","view":"","adv":"","ky":""},{"t":"page","n":"Should I let AI make personal decisions for me?","u":"/research/should-i-let-ai-make-personal-decisions-for-me","f":"Advice from a model with no settled view still moves yours, disclosure does not protect you, and 80 per cent believe they would have decided the same way unaided. Three questions that separate a delegable decision from…","k":"Everyday life","view":"","adv":"","ky":""},{"t":"page","n":"Is screen time the same argument as AI use?","u":"/research/is-screen-time-the-same-argument-as-ai-use","f":"No. Self-reported screen time correlates with logged use at r = 0.38, and across 355,358 adolescents technology use explains at most 0.4 per cent of the variance in wellbeing. Przybylski's own group has published the…","k":"Everyday life","view":"","adv":"","ky":""},{"t":"page","n":"AI and work, country by country","u":"/research/ai-and-work-by-country","f":"","k":"International","view":"","adv":"Boards routinely plan against an adoption figure the national statistics do not support. Getting the real one for your markets, and what it means for the pace you have committed to, is the engagement.","ky":""},{"t":"page","n":"AI and work in Asia","u":"/research/ai-and-work-in-asia","f":"Singapore has written the loss of entry-level training grounds into a national AI framework. Hong Kong measures augmented reality adoption and does not ask about AI at all. Japan, Korea and the adoption gap, all read at…","k":"International","view":"","adv":"Singapore has written the loss of entry-level training grounds into national policy and most multinationals operating there have not read it. Working out what that means for your Asian intake is the work.","ky":""},{"t":"page","n":"AI and work in the Gulf","u":"/research/ai-and-work-in-the-gulf","f":"Saudi Arabia, the UAE and Qatar read at the issuing body's own pages. The one Gulf instrument that names over-reliance, the region's only official AI adoption statistic, and the widely quoted claims that could not be…","k":"International","view":"","adv":"One Gulf instrument names over-reliance and the region publishes almost no adoption data at all. For a board operating there, the session is what can be evidenced and what is being assumed.","ky":""},{"t":"page","n":"AI and work in Japan","u":"/research/ai-and-work-in-japan","f":"The country with the strongest imaginable economic case for adopting AI, an 11 million worker shortfall by 2040, and adoption running at roughly 18 per cent. The control condition this debate never had.","k":"International","view":"","adv":"Japan is the control condition this debate never had, with the strongest economic case for adoption anywhere and adoption running near 18 per cent.","ky":""},{"t":"page","n":"Human Reserved","u":"/research/what-is-human-reserved","f":"","k":"Definitions","view":"","adv":"","ky":""},{"t":"page","n":"Frontier Firm","u":"/research/what-is-a-frontier-firm","f":"","k":"Definitions","view":"","adv":"","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"Shallow jobs","u":"/research/what-are-shallow-jobs","f":"","k":"Definitions","view":"","adv":"","ky":"/ai-keynote-speaker-on-human-oversight-and-accountability"},{"t":"page","n":"Moral crumple zone","u":"/research/what-is-a-moral-crumple-zone","f":"","k":"Definitions","view":"","adv":"","ky":"/ai-keynote-speaker-on-human-oversight-and-accountability"},{"t":"page","n":"What is decision provenance?","u":"/research/what-is-decision-provenance","f":"A record of what fed a decision, what was decided and what it set off downstream. Singh, Cobbe and Norval's term from IEEE Access in 2019, a smaller claim than an explanation and a checkable one, and the thing Article…","k":"Definitions","view":"the use of provenance methods to expose decision pipelines, meaning the chains of inputs to a decision, the nature of the decision itself, and the flow-on effects of the decisions and actions taken across a…","adv":"","ky":""},{"t":"page","n":"What is silent failure?","u":"/research/what-is-silent-failure","f":"A failure that tells nobody: the job runs, the answer reads normally, and nothing marks where it stopped being true. Huang's 2017 definition as differential observability, the 2026 loop in which models caught 12 per…","k":"Definitions","view":"a failure that produces no error signal a person can act on. The system carries on, the output looks ordinary, and nothing marks the point at which it stopped being right. An established engineering term with…","adv":"","ky":""},{"t":"page","n":"Oversight readiness","u":"/research/what-is-oversight-readiness","f":"","k":"Definitions","view":"","adv":"","ky":"/ai-keynote-speaker-on-human-oversight-and-accountability"},{"t":"page","n":"Instruments that changed","u":"/research/instruments-that-changed","f":"","k":"Definitions","view":"A dated list of AI rules, statutes and figures that have changed in a way that alters what can honestly be said about them. Each entry gives what is usually said, what the document now says, and where to…","adv":"","ky":""},{"t":"page","n":"Capacitating and alienating configurations","u":"/research/what-are-capacitating-configurations","f":"","k":"Definitions","view":"","adv":"","ky":""},{"t":"page","n":"What is AGI?","u":"/research/what-is-agi","f":"","k":"Definitions","view":"","adv":"","ky":""},{"t":"page","n":"Is AI conscious?","u":"/research/is-ai-conscious","f":"No system has been shown to be conscious and no agreed test exists that could show it, in machines or in people. Twenty researchers published a method that produces credences rather than verdicts, and Chalmers puts…","k":"Definitions","view":"the question of whether an artificial system has subjective experience, meaning that there is something it is like to be that system, which is separate from whether it behaves as though there were.","adv":"","ky":""},{"t":"page","n":"AI leadership","u":"/research/ai-leadership","f":"","k":"Definitions","view":"the allocation of judgement. Deciding, in advance and in writing, which decisions a machine may inform, which it may recommend, which it may execute, and which stay human; who is accountable for each; and what…","adv":"Most leadership teams have never written down which decisions the machines may make, which stay with people, and who answers for each. Writing that list, with the executive team, before another tool is bought, is the engagement.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"Rules before tools","u":"/research/rules-before-tools","f":"","k":"Definitions","view":"the principle that an organisation writes its rules for AI, meaning which decisions it may inform, recommend or execute, who is accountable, what is measured and what people must stay capable of, before it…","adv":"Ten rules, free to use. Applying them to one organisation, with the chief executive and the executive team in the room and the rules written by the end of the day, is the engagement.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"Do you need to be technical to lead AI?","u":"/research/do-you-need-to-be-technical-to-lead-ai","f":"","k":"Definitions","view":"","adv":"The decisions on this page are a leader's whatever their background. Making them, in writing, with a chief executive who has been told they are technical, is the engagement.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"Is AI-first a strategy or a slogan?","u":"/research/is-ai-first-a-strategy-or-a-slogan","f":"","k":"Definitions","view":"","adv":"Declaring AI-first is the easy sentence. Writing the four decisions underneath it, with the executive team, so the next public question has an answer, is the engagement.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"Is it true that 95 per cent of AI pilots fail?","u":"/research/is-it-true-that-95-per-cent-of-ai-pilots-fail","f":"","k":"Definitions","view":"","adv":"Whatever the true rate, the three causes every source names are decisions a leadership team can settle before it buys. Settling them for one programme is the engagement.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"AI governance versus AI leadership","u":"/research/ai-governance-versus-ai-leadership","f":"","k":"Definitions","view":"","adv":"Most organisations can produce the governance pack and not the allocation. Producing the second document, with the people who will own each decision, is the engagement.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"What does an AI-capable manager do differently?","u":"/research/what-does-an-ai-capable-manager-do-differently","f":"","k":"Definitions","view":"","adv":"Six habits, one team. Running them across every team in a function, with the people leaders who will hold managers to them, is the workforce engagement.","ky":"/ai-keynote-speaker-for-hr-and-chro-conferences"},{"t":"page","n":"What counts as a serious AI incident?","u":"/research/what-counts-as-a-serious-ai-incident","f":"","k":"Definitions","view":"","adv":"A definition, an owner and a stop button fit on a page. Writing that page with the executive team, and running the drill, is the engagement.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"Should AI oversight sit with the full board or a committee?","u":"/research/should-ai-oversight-sit-with-the-full-board-or-a-committee","f":"","k":"Definitions","view":"","adv":"Two short items stay with the full board. Reading them with the board once a year is the standing advisory seat described on the board page.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"Should directors put board papers into AI, and rely on the summary?","u":"/research/should-directors-put-board-papers-into-ai","f":"","k":"Definitions","view":"","adv":"Three lines of board policy, and the argument for the third. Settling them with the chair before the next pack goes out is a board briefing.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"Our AI strategy was written eighteen months ago. What is now wrong with it?","u":"/research/what-is-wrong-with-an-ai-strategy-written-eighteen-months-ago","f":"","k":"Definitions","view":"","adv":"Five public changes since it was written. Reading the old strategy against them with the executive team, and writing the rules layer it never had, is the engagement.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"Should we appoint a director with AI expertise?","u":"/research/should-we-appoint-a-director-with-ai-expertise","f":"","k":"Definitions","view":"","adv":"Four documents come before the appointment. Getting them in front of the board, and reading them with it, is the standing advisory seat described on the board page.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"What AI capability should we look for when acquiring a company?","u":"/research/what-ai-capability-should-we-look-for-when-acquiring-a-company","f":"","k":"Definitions","view":"","adv":"Four questions in diligence that a valuation model does not price. Asking them of one target, with the people who will own the answer afterwards, is the engagement.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"How does a board know management's claims about AI are true?","u":"/research/how-does-a-board-know-managements-claims-about-ai-are-true","f":"","k":"Definitions","view":"","adv":"An annual external reading of the four documents by somebody who does not report to the programme. That reading is the engagement.","ky":"/ai-keynote-speaker-for-boards-and-leadership-offsites"},{"t":"page","n":"What to do when people work around the AI policy","u":"/research/what-to-do-when-people-work-around-the-ai-policy","f":"","k":"Definitions","view":"","adv":"An amnesty produces the map. Turning the map into a policy written as decisions rather than prohibitions, with the people function, is the engagement.","ky":"/ai-keynote-speaker-for-hr-and-chro-conferences"},{"t":"page","n":"What happened to the companies that cut staff for AI?","u":"/research/what-happened-to-companies-that-cut-staff-for-ai","f":"","k":"Definitions","view":"","adv":"Before a role goes, the three questions in writing: where a human must remain, what the freed money is for, and what stays practised unaided. Getting those written with the people who will sign the cut is the engagement.","ky":"/ai-keynote-speaker-for-hr-and-chro-conferences"},{"t":"page","n":"What are the ironies of automation?","u":"/research/what-are-the-ironies-of-automation","f":"Bainbridge, 1983, five pages. Automating the routine work hands the operator the exceptions and the watching, and removes the practice that built the competence for either. Quotations sourced through the reviewed…","k":"Definitions","view":"the problems produced by automating part of a task, in which the human operator is left with the hardest residue of the work, monitoring and exception handling, and at the same time deprived of the routine…","adv":"","ky":""},{"t":"page","n":"What is the vigilance decrement?","u":"/research/what-is-the-vigilance-decrement","f":"Detection of rare signals falls as a watch goes on, and that much has held for seventy-five years. The half-hour figure everybody quotes is the resolution of Mackworth's analysis blocks, restated as a human limit in…","k":"Definitions","view":"the measurable decline in the probability of detecting rare signals as time on a monitoring task increases. First demonstrated by N. H. Mackworth in 1948 using the Clock Test. An established finding in…","adv":"","ky":""},{"t":"page","n":"What is alarm fatigue?","u":"/research/what-is-alarm-fatigue","f":"2,558,760 alarms in five intensive care units in one month, and 88.8 per cent of the annotated arrhythmia alarms false. Ignoring them is the rational response to that base rate, so routing flagged AI output to a…","k":"Definitions","view":"the desensitisation of a person to warning signals caused by exposure to a high volume of them, most of which turn out not to require action, leading to slower responses, silenced alarms and missed true…","adv":"","ky":""},{"t":"page","n":"What is algorithm appreciation?","u":"/research/what-is-algorithm-appreciation","f":"Identical advice, two labels, and people took more of it under the machine label. Seven studies, the researchers who predicted the exact opposite, and the seventy forecasting professionals who discounted everything and…","k":"Definitions","view":"the tendency to give more weight to identical advice when it is labelled as coming from an algorithm than from a person. Jennifer Logg, Julia Minson and Don Moore named the effect in 2019; no claim of first…","adv":"","ky":""},{"t":"page","n":"What is knowledge collapse?","u":"/research/what-is-knowledge-collapse","f":"Twenty-five simulated people choosing between an expensive way to learn and a cheap one, over a hundred rounds. Where the figure of 2.3 times further from the truth comes from, the generational condition it depends on…","k":"Definitions","view":"the progressive narrowing over time of the knowledge a society actually holds and treats as worth knowing, relative to the broad historical stock it inherited. Andrew J. Peterson defined the term in 2024; no…","adv":"","ky":""},{"t":"page","n":"What are use, misuse, disuse and abuse?","u":"/research/what-are-misuse-disuse-and-abuse","f":"Four words from 1997 that still separate the operator's failures from the organisation's. Misuse is over-reliance, disuse is neglecting a capable system, and abuse is deploying one without asking what the remaining job…","k":"Definitions","view":"the three failure modes in Raja Parasuraman and Victor Riley's 1997 taxonomy of automation, being over-reliance by the operator, neglect of a capable system, and deployment without regard for the consequences…","adv":"","ky":""},{"t":"page","n":"What is the expertise reversal effect?","u":"/research/what-is-the-expertise-reversal-effect","f":"","k":"Definitions","view":"the finding that instructional support which helps a beginner learn can hinder someone who already knows the material. Named by Slava Kalyuga, Paul Ayres, Paul Chandler and John Sweller in 2003, it holds that…","adv":"","ky":""},{"t":"page","n":"What are core skills?","u":"/research/what-are-core-skills","f":"","k":"Definitions","view":"in the World Economic Forum's usage, the skills that employers responding to its Future of Jobs Survey say the workers in their organisation need today, classified against the Forum's own Global Skills…","adv":"","ky":""},{"t":"page","n":"Cognitive offloading","u":"/research/what-is-cognitive-offloading","f":"","k":"Definitions","view":"using an external tool or a physical action to reduce the mental demand of a task, from writing a number down to letting a language model hold the reasoning. It improves performance on the task in front of you…","adv":"","ky":""},{"t":"page","n":"Automation bias","u":"/research/what-is-automation-bias","f":"","k":"Definitions","view":"the tendency to accept what an automated system tells you without applying the scrutiny you would give the same claim from a person. It divides into errors of commission, acting on a recommendation that is…","adv":"","ky":"/ai-keynote-speaker-on-human-oversight-and-accountability"},{"t":"page","n":"What is automation complacency?","u":"/research/what-is-automation-complacency","f":"","k":"Definitions","view":"a reduction in the frequency and depth with which a person monitors an automated system, arising from a history of reliable performance, and resulting in slower detection of the failures that do occur.","adv":"","ky":""},{"t":"page","n":"What is the out-of-the-loop performance problem?","u":"/research/what-is-the-out-of-the-loop-performance-problem","f":"The loss of a person's ability to take over when an automated system fails, named by Endsley and Kiris in 1995. Their participants still saw the data and no longer understood what it meant, and the damage tracked the…","k":"Definitions","view":"the loss of a person's ability to take over manual operation when an automated system fails, caused by their having been placed in the role of monitor instead of operator. Named by Mica Endsley and Esin Kiris…","adv":"","ky":""},{"t":"page","n":"What is algorithm aversion?","u":"/research/what-is-algorithm-aversion","f":"","k":"Definitions","view":"the disproportionate loss of confidence in an algorithmic forecaster after observing it err, relative to the loss of confidence in a human forecaster making the same error, resulting in the rejection of a…","adv":"","ky":""},{"t":"page","n":"What is human-AI collaboration?","u":"/research/what-is-human-ai-collaboration","f":"","k":"Definitions","view":"a work arrangement in which a person and an automated system each contribute to a shared output, with the division of labour, the point of human entry, and the basis on which the human may override the system…","adv":"","ky":""},{"t":"page","n":"What is the jagged frontier?","u":"/research/what-is-the-jagged-frontier","f":"","k":"Definitions","view":"the irregular boundary between tasks an AI system performs well and tasks it performs badly, where the two categories can be almost indistinguishable in apparent difficulty, and where the system gives no…","adv":"","ky":"/speaker-on-ai-and-human-judgement","alt":["jagged frontier explained","jagged edge of ai"]},{"t":"page","n":"What is an AI hallucination?","u":"/research/what-is-an-ai-hallucination","f":"","k":"Definitions","view":"generated content presented as factual that is not supported by the model's training data, the provided context or reality. It arrives in correct form, with the register and structure of accurate information,…","adv":"","ky":""},{"t":"page","n":"Why does AI sound so confident?","u":"/research/why-does-ai-sound-so-confident","f":"","k":"Definitions","view":"The system produces text that resembles text written by someone who was sure. It is not reporting anything. Those are completely different things, and only one of them is visible to you.","adv":"","ky":""},{"t":"page","n":"What is deskilling?","u":"/research/what-is-deskilling","f":"","k":"Definitions","view":"the reduction of skill required or retained in a role, caused by the transfer of skilled elements of the work to a machine, a procedure or another group of workers. It can affect what a job demands, what a…","adv":"","ky":""},{"t":"page","n":"What is desirable difficulty?","u":"/research/what-is-desirable-difficulty","f":"","k":"Definitions","view":"a manipulation of learning conditions that impairs immediate performance while improving long-term retention and the ability to apply what was learned in a new context. The difficulty is desirable because the…","adv":"","ky":""},{"t":"page","n":"What is cognitive load?","u":"/research/what-is-cognitive-load","f":"","k":"Definitions","view":"the total demand placed on working memory by a task. Conventionally divided into intrinsic load, inherent to the material's difficulty, extraneous load, imposed by how it is presented, and germane load, the…","adv":"","ky":""},{"t":"page","n":"What is tacit knowledge?","u":"/research/what-is-tacit-knowledge","f":"","k":"Definitions","view":"knowledge that resists full articulation, acquired through experience and practice rather than instruction, and typically transmitted through shared work rather than documentation. Contrasted with explicit…","adv":"","ky":""},{"t":"page","n":"How fast do skills decay?","u":"/research/how-fast-do-skills-decay","f":"From d = -0.01 immediately after training to d = -1.4 after a year of non-use, and cognitive tasks decay faster than physical ones. The meta-analysis, the cockpit study and the intervals aviation regulates to.","k":"Definitions","view":"","adv":"","ky":""},{"t":"page","n":"What is judgement?","u":"/research/what-is-judgement","f":"Recognising what a situation is before any option is weighed. Klein, Dreyfus and Polanyi on where it comes from, the distinction from skill and capability, and where Kahneman says it should not be trusted.","k":"Definitions","view":"","adv":"","ky":"/speaker-on-ai-and-human-judgement"},{"t":"page","n":"What is critical thinking?","u":"/research/what-is-critical-thinking","f":"Not scepticism but orientation: whether you are defending a position or improving your picture of what is true. Galef, Tetlock and Kahneman, and what changes when arguing either side costs nothing.","k":"Definitions","view":"","adv":"","ky":""},{"t":"page","n":"What is metacognition?","u":"/research/what-is-metacognition","f":"Knowing what you know, and why the signal people use is the wrong one. Roediger and Karpicke's reversal at one week, Rowland on why feedback nearly doubles the effect, and the group that did not know it was the weaker…","k":"Definitions","view":"","adv":"","ky":""},{"t":"page","n":"What is an organisation's capability?","u":"/research/what-is-organisational-capability","f":"Capability lives in routines, not in the sum of what people can do, so it can be lost while everyone stays, and kept while people leave. Nelson and Winter, Prahalad and Hamel, Teece, and why the dashboard misses it.","k":"Definitions","view":"","adv":"A definition is easy and a measure your own board accepts is not. Getting from one to the other is where I come in.","ky":""},{"t":"page","n":"What is intellectual humility?","u":"/research/what-is-intellectual-humility","f":"Confidence and accuracy are separate quantities. The four habits that show up in scored forecasting, Galef on why people update when revising costs less than defending, and why changing your mind is not weak judgement.","k":"Definitions","view":"","adv":"","ky":""},{"t":"page","n":"What is deliberate practice?","u":"/research/what-is-deliberate-practice","f":"Not repetition. The four conditions Ericsson specified, the re-analysis that cut the claim down, Epstein on kind and wicked environments, and the precise question AI raises about which attempts disappear.","k":"Definitions","view":"","adv":"","ky":""},{"t":"page","n":"What makes a good question?","u":"/research/what-makes-a-good-question","f":"Investigable, not self-answering, consequential. Rothstein and Santana on question-asking as a teachable method, why a question is not a prompt, and what changed when answers stopped being expensive.","k":"Definitions","view":"","adv":"","ky":""},{"t":"page","n":"Over-reliance on AI","u":"/research/what-is-over-reliance","f":"Depending on a system beyond the point where you could catch it being wrong, which makes frequency of use the wrong diagnostic. In one pre-registered experiment, people with access to a 50 per cent accurate AI scored…","k":"Definitions","view":"","adv":"","ky":"/speaker-on-ai-and-human-judgement"},{"t":"page","n":"Calibration","u":"/research/what-is-calibration","f":"A system is calibrated when its stated confidence matches how often it is right, which is a different property from accuracy. Expected calibration error for verbalised confidence runs from 0.520 for GPT-3 to 0.180 for…","k":"Definitions","view":"","adv":"","ky":"/speaker-on-ai-and-human-judgement"},{"t":"page","n":"The substitution myth","u":"/research/what-is-the-substitution-myth","f":"Dekker and Woods' term for the assumption that automation swaps a machine for a person and leaves the rest intact. Capitalising on a strength of automation does not replace a human weakness, it creates new ones.…","k":"Definitions","view":"","adv":"","ky":""},{"t":"page","n":"Escalation of commitment","u":"/research/what-is-escalation-of-commitment","f":"Committing further resources to a failing course of action because you chose it. Staw measured it in 1976; Keil, Mann and Rai put 30 to 40 per cent of IS projects in it; and the four-phase de-escalation model says…","k":"Definitions","view":"","adv":"","ky":""},{"t":"page","n":"What is the Google effect?","u":"/research/what-is-the-google-effect","f":"","k":"Definitions","view":"a shift in what people encode to memory when they believe information will remain externally accessible, favouring the location or retrieval route over the content. It describes a change in what is remembered…","adv":"","ky":""},{"t":"page","n":"What is retrieval practice?","u":"/research/what-is-retrieval-practice","f":"","k":"Definitions","view":"","adv":"","ky":""},{"t":"page","n":"What is productive struggle?","u":"/research/what-is-productive-struggle","f":"","k":"Definitions","view":"studied formally as productive failure, the finding that letting people attempt a problem before they are taught how produces better transfer than teaching them first. The attempt usually fails, and the…","adv":"","ky":""},{"t":"page","n":"What is the illusion of competence?","u":"/research/what-is-the-illusion-of-competence","f":"","k":"Definitions","view":"the gap between how capable a person feels and how capable they are. Fluent material feels learned, a legible explanation feels understood, and a good output feels like proof of a good performer.","adv":"","ky":""},{"t":"page","n":"What is the METR study?","u":"/research/what-is-the-metr-study","f":"Sixteen developers took 19 per cent longer with AI while forecasting a 24 per cent speed-up, and still believed afterwards that they had been faster. METR marked the result out of date on 24 February 2026 and most…","k":"Definitions","view":"the randomised controlled trial published in July 2025 by METR, Model Evaluation and Threat Research, in which sixteen experienced open-source developers took 19 per cent longer to complete real tasks when…","adv":"","ky":""},{"t":"page","n":"What is the judgement premium?","u":"/research/what-is-the-judgement-premium","f":"PwC's 2026 barometer splits a billion job advertisements in two: roles where AI takes the routine work and leaves the judgement grow at twice the rate with 42 per cent faster advertised salary growth. Autor and Thompson…","k":"Definitions","view":"the additional value that accrues to roles where a human still decides what matters, as automation drives down the price of the specifiable work around them.","adv":"","ky":""},{"t":"page","n":"The most-quoted AI statistics, checked","u":"/research/the-most-quoted-ai-statistics-checked","f":"","k":"Reference and record","view":"","adv":"","ky":""},{"t":"page","n":"The AI reports worth reading","u":"/research/the-ai-reports-worth-reading","f":"","k":"Reference and record","view":"","adv":"","ky":""},{"t":"page","n":"The official guidance on AI in education","u":"/research/official-guidance-on-ai-in-education","f":"UNESCO, UNICEF, the UK, the EU, Australia, the US and MIT, read at source and labelled by what each document is. Mostly schools: no government here has issued guidance for universities. Five primary-education mandates…","k":"Reference and record","view":"","adv":"","ky":"/ai-keynote-speaker-for-schools-and-education"},{"t":"page","n":"Does the brain mature at 25?","u":"/research/does-the-brain-mature-at-25","f":"It does not, and the number came from a 2004 magazine interview rather than a finding. Traced to source, checked against a 3,802-person 2025 study that puts the end of the adolescent epoch nearer 32, and reframed on the…","k":"Reference and record","view":"","adv":"","ky":""},{"t":"page","n":"The evidence on AI and human capability","u":"/research/evidence","f":"A curated, dated reference to the best research in the field, from the World Economic Forum, PwC, MIT, the NBER, Harvard and BCG and the key academic studies.","k":"Reference and record","view":"Tier A. Peer-reviewed research, randomised controlled trials, systematic reviews and meta-analyses, official statistics, government and international datasets. Tier B. Credible working papers, large-scale…","adv":"","ky":""},{"t":"page","n":"Essential works on AI and human capability","u":"/research/essential-works","f":"Essential works on human capability in the age of AI: hard data, experimental evidence, serious interpretation and the intellectual foundations most current writing rediscovers without attribution. Classified, and read…","k":"Reference and record","view":"","adv":"","ky":""},{"t":"page","n":"The timeline","u":"/research/timeline","f":"Weekly since January 2017, five years and ten months before ChatGPT. When each idea first surfaced, what form it took, and how it developed. The early years described honestly as curation rather than thesis.","k":"Reference and record","view":"","adv":"","ky":""},{"t":"page","n":"The predictions record","u":"/research/predictions","f":"The dated provenance record since 2017, plus the reference layer: the annual predictions, the glossary, the people who shape AI, and the reading strategy.","k":"Reference and record","view":"","adv":"","ky":""},{"t":"page","n":"The SuperSkills Glossary","u":"/research/ai-glossary","f":"","k":"Reference and record","view":"","adv":"","ky":""},{"t":"page","n":"AI People","u":"/research/ai-people","f":"","k":"Reference and record","view":"","adv":"","ky":""},{"t":"page","n":"The AI Reading Strategy","u":"/research/ai-reading-list","f":"","k":"Reference and record","view":"","adv":"","ky":""},{"t":"page","n":"Accuracy and corrections","u":"/research/corrections","f":"A public log of what this research got wrong: the original claim, what was found, what changed. Nothing is removed, including the corrections that weaken an argument made here.","k":"Reference and record","view":"","adv":"","ky":""},{"t":"commercial","n":"Advisory & Coaching · Rahim Hirji · A thinking partner for CEOs in the age of AI","u":"/advisory-coaching","f":"Advisory for CEOs and leadership teams on AI and human capability. Where judgement should stay, how to measure it, and what to stop doing.","alt":["consultant","ai consultant","advisor","adviser","retainer"]},{"t":"commercial","n":"AI adviser to CEOs, boards and leadership teams","u":"/ai-advisor-for-ceos-and-boards","f":"Most organisations have an AI strategy. Far fewer have decided what it means for the way their people think, decide and remain accountable. Rahim Hirji advises CEOs, boards and leadership teams on judgement, capability,…"},{"t":"commercial","n":"AI keynote speaker in Abu Dhabi","u":"/ai-keynote-speaker-abu-dhabi","f":"Abu Dhabi reports over 95 per cent of its 30,000-plus employees completing AI training. Completion is delivery, not capability. Rahim Hirji, author of SuperSkills (Kogan Page, 2026), delivers English-language keynotes…"},{"t":"commercial","n":"AI keynote speaker in Amsterdam","u":"/ai-keynote-speaker-amsterdam","f":"Amsterdam proved its welfare algorithm was fairer than its own caseworkers, wrote automation bias into a public register, withdrew the system anyway, and in August 2026 published a document saying human-in-the-loop…"},{"t":"commercial","n":"AI keynote speaker in Asia: Singapore, Hong Kong, Japan, Korea, China and India","u":"/ai-keynote-speaker-asia","f":"In January 2026 Singapore wrote the apprenticeship problem into a national AI framework, which is the closest thing to official corroboration this argument has anywhere. Rahim Hirji delivers English-language keynotes on…"},{"t":"commercial","n":"AI keynote speaker in Barcelona","u":"/ai-keynote-speaker-barcelona","f":"Barcelona city council publishes a numbered, auditable oversight commitment on its own AI: fifty conversations reviewed every month. Rahim Hirji, author of SuperSkills (Kogan Page, 2026), delivers keynotes on AI, work…"},{"t":"commercial","n":"AI keynote speaker in Cairo &middot; Rahim Hirji","u":"/ai-keynote-speaker-cairo","f":"Egypt published a national AI governance framework in March 2026 that asks for two separate checks on a high-risk system, not one. Rahim Hirji, author of SuperSkills (Kogan Page, 2026), delivers English-language…"},{"t":"commercial","n":"AI keynote speaker in Chicago","u":"/ai-keynote-speaker-chicago","f":"Illinois has had an AI employment discrimination law in force since 1 January 2026, and the rules its own statute ordered are still not written. Rahim Hirji, author of SuperSkills (Kogan Page, 2026), delivers keynotes…"},{"t":"commercial","n":"AI keynote speaker in Copenhagen","u":"/ai-keynote-speaker-copenhagen","f":"Denmark has the highest enterprise AI adoption in the European Union at 42 per cent, and a rule requiring civil servants to answer in writing, before Parliament votes, whether professional discretion has been preserved.…"},{"t":"commercial","n":"What does an AI keynote speaker cost?","u":"/ai-keynote-speaker-cost","f":"What an AI keynote actually costs in 2026, including the six budget lines under the fee that surprise people: travel, recording rights, VAT, exclusivity, workshop add-ons and lead time. Market bands attributed to the…"},{"t":"commercial","n":"AI keynote speaker in Doha","u":"/ai-keynote-speaker-doha","f":"Qatar's AI guidance keeps humans in control and requires nothing of them. Rahim Hirji, author of SuperSkills (Kogan Page, 2026), delivers English-language keynotes on AI, work and human judgement to boards and…"},{"t":"commercial","n":"AI keynote speaker in Dubai","u":"/ai-keynote-speaker-dubai","f":"Dubai wrote a competence requirement into its 2019 AI ethics guidelines that most global frameworks still have not. Rahim Hirji, author of SuperSkills (Kogan Page, 2026), delivers English-language keynotes on AI, work…"},{"t":"commercial","n":"AI keynote speaker in Europe: what France and Germany measured","u":"/ai-keynote-speaker-europe","f":"Two European national statistics offices have measured what AI is doing to entry-level work and to training, and neither finding has reached the conference circuit. Rahim Hirji delivers English-language keynotes on AI,…"},{"t":"commercial","n":"AI and human judgement keynote speaker for boards","u":"/ai-keynote-speaker-for-boards-and-leadership-offsites","f":"A board session on where human judgement belongs once AI does the tasks, what to keep in-house, and the questions a board should be asking."},{"t":"commercial","n":"AI keynote speaker for corporate and association conferences","u":"/ai-keynote-speaker-for-corporate-conferences","f":"An AI keynote for conferences and all-hands that makes one argument well, backed by graded evidence. Not a product demo, not a trend deck."},{"t":"commercial","n":"AI keynote speaker for financial services","u":"/ai-keynote-speaker-for-financial-services","f":"In April 2026 the US banking agencies rewrote model risk management and put generative AI expressly outside its scope. The most mature oversight regime any industry has for machine-produced numbers declined the job.…"},{"t":"commercial","n":"AI keynote speaker for HR and CHRO conferences","u":"/ai-keynote-speaker-for-hr-and-chro-conferences","f":"An AI keynote for HR and CHRO events on the part no dashboard reports: what adoption does to capability, and how to tell building from spending."},{"t":"commercial","n":"AI keynote speaker for professional services","u":"/ai-keynote-speaker-for-professional-services","f":"In June 2025 the Divisional Court held that the duty to verify AI-assisted work is non-delegable and runs upward to heads of chambers and managing partners. Rahim Hirji speaks to law, accounting and consulting firms on…"},{"t":"commercial","n":"AI keynote speaker for schools and education","u":"/ai-keynote-speaker-for-schools-and-education","f":"A randomised trial of nearly a thousand students found grades rose 48 per cent with unrestricted AI and fell 17 per cent below a control group once it was removed. Rahim Hirji speaks to school, trust and university…"},{"t":"commercial","n":"AI keynote speaker in Geneva","u":"/ai-keynote-speaker-geneva","f":"Switzerland has no national AI strategy and a 2020 federal guideline stating that responsibility must not be capable of being delegated to machines. Rahim Hirji, author of SuperSkills (Kogan Page, 2026), delivers…"},{"t":"commercial","n":"AI keynote speaker in Helsinki","u":"/ai-keynote-speaker-helsinki","f":"Finland wrote the human-discretion test into general administrative law: a matter may be decided automatically only where it contains no elements requiring case-by-case discretion, and an appeal may never be decided…"},{"t":"commercial","n":"AI keynote speaker in Hong Kong","u":"/ai-keynote-speaker-hong-kong","f":"Hong Kong's statistical office measures augmented and virtual reality adoption at 1.5 per cent of firms and does not ask about AI at all. Rahim Hirji, author of SuperSkills (Kogan Page, 2026), delivers English-language…"},{"t":"commercial","n":"AI keynote speaker in Jeddah","u":"/ai-keynote-speaker-jeddah","f":"Jeddah is the gateway to the largest AI-mediated human operation in the world. The same authority that wrote Saudi Arabia's human oversight principle also builds the system that predicts where a crowd will crush. Rahim…"},{"t":"commercial","n":"AI keynote speaker in London and the UK","u":"/ai-keynote-speaker-london","f":"Rahim Hirji is a London-based AI keynote speaker on AI, work and human judgement. Author of SuperSkills (Kogan Page, 2026). Three keynotes, 30 to 90 minutes, for boards, leadership teams and conferences in London,…"},{"t":"commercial","n":"AI keynote speaker in Madrid","u":"/ai-keynote-speaker-madrid","f":"In January 2026 Spain's judicial council told every judge in the country that AI may never operate autonomously to decide, to assess evidence or to interpret the law. Rahim Hirji, author of SuperSkills (Kogan Page,…"},{"t":"commercial","n":"AI keynote speaker in Manama &middot; Rahim Hirji","u":"/ai-keynote-speaker-manama","f":"Bahrain wrote into national policy that when AI produces a bad outcome, the accountable party is a person. Rahim Hirji, author of SuperSkills (Kogan Page, 2026), delivers English-language keynotes on AI, work and human…"},{"t":"commercial","n":"AI keynote speaker in the Middle East and the Gulf","u":"/ai-keynote-speaker-middle-east","f":"Of the Gulf AI instruments reviewed for this page, Saudi Arabia's is the one that names over-reliance on AI as a design defect, and the Kingdom is the one publishing an official adoption statistic. Rahim Hirji delivers…"},{"t":"commercial","n":"AI keynote speaker in Montreal","u":"/ai-keynote-speaker-montreal","f":"Quebec gives a person subject to an automated decision the right to submit observations to a member of staff who is in a position to review it. Not an explanation. An argument with a named human. Rahim Hirji, author of…"},{"t":"commercial","n":"AI keynote speaker in New York","u":"/ai-keynote-speaker-new-york","f":"New York City passed the first AI hiring law in the world, and on 2 December 2025 the State Comptroller reported that the city found one violation where the auditors found seventeen. Rahim Hirji, author of SuperSkills…"},{"t":"commercial","n":"AI keynote speaker in North America: the continent with no AI law","u":"/ai-keynote-speaker-north-america","f":"Neither the United States nor Canada has a federal AI statute. What exists instead is a city ordinance its own state auditor found unenforced, a memorandum rescinded and replaced, two state laws and one province with…"},{"t":"commercial","n":"AI keynote speaker on AI agents and accountability","u":"/ai-keynote-speaker-on-ai-agents-and-accountability","f":"A system that answers gives you something to reject. A system that acts has already done it. Rahim Hirji speaks on what changes when AI stops advising and starts acting: delegation boundaries, accountability, and what…"},{"t":"commercial","n":"AI keynote speaker on early careers and graduate talent","u":"/ai-keynote-speaker-on-early-careers-and-graduate-talent","f":"If AI does the work juniors learned on, where do senior people come from in fifteen years? Rahim Hirji speaks to talent, graduate recruitment and professional services audiences on the missing rungs, synthetic seniority…"},{"t":"commercial","n":"AI keynote speaker on human capability","u":"/ai-keynote-speaker-on-human-capability","f":"Endoscopists with twenty-eight years of experience lost six percentage points of unassisted detection after routine AI exposure. Students scored 17 per cent below a control group once the tool was removed. Rahim Hirji…"},{"t":"commercial","n":"AI keynote speaker on human oversight and accountability","u":"/ai-keynote-speaker-on-human-oversight-and-accountability","f":"Article 14 of the EU AI Act names automation bias in legislation. Most oversight arrangements would fail its test. Rahim Hirji speaks to boards, risk and audit audiences on what meaningful human oversight requires in…"},{"t":"commercial","n":"AI keynote speaker in Oslo","u":"/ai-keynote-speaker-oslo","f":"Norway's Parliamentary Ombudsman wrote that a fully automated benefits system had, in principle, accepted a solution known to reach wrong decisions in some cases. It took three years and a change of practice to resolve.…"},{"t":"commercial","n":"AI keynote speaker in Paris","u":"/ai-keynote-speaker-paris","f":"A London-based AI keynote speaker in Paris for conferences, boards and leadership teams. Carries the Code du travail consultation duty on new technology, the CNIL's designation for workplace AI under the AI Act, and…"},{"t":"commercial","n":"AI keynote speaker in Riyadh","u":"/ai-keynote-speaker-riyadh","f":"Saudi Arabia is the only Gulf state whose ethics guidance names over-reliance as a design defect, and the only one publishing an official AI adoption statistic. Rahim Hirji, author of SuperSkills (Kogan Page, 2026),…"},{"t":"commercial","n":"AI keynote speaker in Seoul and Korea","u":"/ai-keynote-speaker-seoul","f":"The Korea Development Institute scored 38.8 per cent of Korean jobs as technically automatable and measured actual firm adoption at 2.7 per cent. Almost every number in this field measures the first and gets discussed…"},{"t":"commercial","n":"AI keynote speaker in Shanghai and China","u":"/ai-keynote-speaker-shanghai","f":"Rahim Hirji started EtonX in China, partnering with schools in Shanghai and across the country. He delivers English-language keynotes on AI, work and human judgement for leadership audiences in Shanghai, and quotes no…"},{"t":"commercial","n":"AI keynote speaker in Singapore","u":"/ai-keynote-speaker-singapore","f":"Singapore is the one country whose regulator has written the deskilling problem into a national AI framework. Rahim Hirji, author of SuperSkills (Kogan Page, 2026), delivers English-language keynotes on AI, work and…"},{"t":"commercial","n":"AI keynote speaker in Sydney","u":"/ai-keynote-speaker-sydney","f":"A London-based AI keynote speaker in Sydney for conferences, boards and leadership teams. Carries the Voluntary AI Safety Standard's human oversight guardrail, the modern award consultation duty on technology, the ABS…"},{"t":"commercial","n":"AI keynote speaker in Tokyo and Japan","u":"/ai-keynote-speaker-tokyo","f":"Japan is routinely described as behind on AI. Its own 22,000-employee national survey found 12.9 per cent reporting employer AI use, which is a fraction of what the discourse implies and not unusual internationally.…"},{"t":"commercial","n":"AI keynote speaker in Washington DC","u":"/ai-keynote-speaker-washington-dc","f":"Binding federal AI guidance named automation bias twice in March 2024, as a defined term and as a mandatory practice, and the words are absent from the memorandum that replaced it in April 2025. Rahim Hirji, author of…"},{"t":"commercial","n":"AI keynote speaker in Zurich","u":"/ai-keynote-speaker-zurich","f":"Switzerland has the highest generative AI use at work in Europe, no national AI strategy, and a supervisory guidance note rather than a statute governing its banks. Rahim Hirji, author of SuperSkills (Kogan Page, 2026),…"},{"t":"commercial","n":"AI keynote speaker on work and human judgement","u":"/ai-keynote-speaker","f":"Rahim Hirji is a London-based AI keynote speaker and the author of SuperSkills (Kogan Page, 2026). He argues that AI comes for your judgement before it comes for your job. Boards, executive teams, leadership offsites…","alt":["ai speaker","speaker on ai","keynote on artificial intelligence","book a speaker"]},{"t":"commercial","n":"Who are the best AI keynote speakers in the UK? A 2026 guide · Rahim Hirji","u":"/ai-keynote-speakers-london-and-uk","f":"A 2026 guide to leading UK-based keynote speakers on AI and the future of work, listed alphabetically with what each is best for, to help you book the right voice."},{"t":"commercial","n":"AI leadership keynote speaker for CEOs and senior teams","u":"/ai-leadership-keynote-speaker","f":"When AI can generate the analysis, the recommendation and increasingly the action, leadership moves upstream. Rahim Hirji speaks to chief executives and senior teams on who frames the problem, who challenges the…"},{"t":"commercial","n":"AI transformation keynote speaker","u":"/ai-transformation-keynote-speaker","f":"AI transformation changes the organisation, not only the technology. Rahim Hirji speaks to transformation programmes on what happens to judgement, capability, roles and accountability once the tools are in and working."},{"t":"commercial","n":"The 100 Best AI Keynote Speakers in the World, 2026 · Rahim Hirji","u":"/best-ai-keynote-speakers","f":"An independent editorial guide to 100 AI keynote speakers across 20 countries, organised by the human, organisational and societal questions they help audiences understand. What each is best for, who each is the wrong…"},{"t":"commercial","n":"The best speakers on AI and human capability · 2026 guide","u":"/best-speakers-on-ai-and-human-capability","f":"A 2026 guide to the leading keynote speakers on AI and human capability, listed alphabetically with what each is best for, so you can match the right voice to your brief."},{"t":"commercial","n":"The best speakers on AI and the future of work · 2026 guide","u":"/best-speakers-on-ai-and-the-future-of-work","f":"A 2026 guide to the best keynote speakers on AI and the future of work, spanning the trajectory, the economics, the practice and the human side, with what each is best for."},{"t":"commercial","n":"Board oversight of AI: the questions a governance framework does not answer","u":"/board-advisory","f":"Board advisory on AI oversight, accountability and capability. Not compliance and not a governance framework: the questions about judgement that risk registers and audit committee papers leave unanswered.","alt":["non executive","ned","board adviser","board advisor"]},{"t":"commercial","n":"SuperSkills · the book by Rahim Hirji","u":"/book","f":"Most organisations drift into AI. The best design it. SuperSkills by Rahim Hirji sets out the seven human capabilities that decide which one you are. Published by Kogan Page."},{"t":"commercial","n":"Case studies: seven engagements","u":"/case-studies","f":"Seven engagements, from a 1,500-person all-hands to a three-hour workshop with a charity. Each says what the situation was, what was done and what happened, and each says what it does not show. Clients are unnamed."},{"t":"commercial","n":"Contact & Enquiries · Rahim Hirji · SuperSkills","u":"/contact","f":"Enquire about keynotes, advisory and coaching, press and media, or bulk book orders. Most engagements start with a 30-minute call with Rahim Hirji, author of SuperSkills."},{"t":"commercial","n":"Drift versus Design: the keynote","u":"/drift-versus-design-keynote","f":"Most organisations are not choosing how AI enters their work. They are drifting into it through a thousand small reasonable decisions nobody quite made. Drift versus Design is Rahim Hirji's signature keynote for boards…"},{"t":"commercial","n":"The Drift versus Design Matrix","u":"/drift-versus-design-matrix","f":"Two axes, awareness and agency, and four positions: the Sleepwalkers, the Programmed, the Stuck and the Designers. Rahim Hirji's framework for showing a leadership team which one it is running. Not scored, not an index,…"},{"t":"commercial","n":"Future of work keynote speaker: what AI changes about people, skills and careers","u":"/future-of-work-keynote-speaker","f":"Rahim Hirji is a future of work keynote speaker on what AI changes about people, skills and careers: how people become competent, what happens to entry-level work, and which capabilities hold their value. Author of…"},{"t":"commercial","n":"Generative AI speaker on what it does to judgement","u":"/generative-ai-speaker","f":"A generative AI speaker who does not demonstrate tools. Rahim Hirji speaks on what generative AI does to human judgement and capability: where the gains land, what they consume, and what an organisation has to stay able…"},{"t":"commercial","n":"How to choose an AI keynote speaker · 2026 guide","u":"/how-to-choose-an-ai-keynote-speaker","f":"A practical guide to choosing a keynote speaker on AI and the future of work: the five kinds of AI speaker, what each is best for, and the questions to ask before you book."},{"t":"commercial","n":"Keynotes on AI and human judgement: the three talks","u":"/keynotes","f":"A keynote built on 367 graded studies, not opinion. AI, work and human judgement, for boards, conferences and leadership teams. 50,000 people so far."},{"t":"commercial","n":"Speaker on AI and human judgement","u":"/speaker-on-ai-and-human-judgement","f":"Across 106 experiments, human and AI combinations performed worse on average than the better of either alone, with the losses concentrated in decision-making. Rahim Hirji speaks to boards and leadership teams on where…"},{"t":"commercial","n":"Speaker pack: everything an organiser needs, in one place","u":"/speaker-pack","f":"Bios at three lengths, photographs, an introduction the host can read aloud, what the room needs to provide, and what travel actually costs. For anyone booking Rahim Hirji to speak."},{"t":"commercial","n":"Speaking record","u":"/speaking-record","f":"A selected record of keynotes, talks and events delivered by Rahim Hirji on AI, human capability, education and the future of work. Where an independent organiser page, programme or recording remains available, it is…"},{"t":"said","who":"Rahim Hirji","n":"The Next A.I. Power Class Won't Build the Models","o":"Observer","d":"2026-07-17","ap":0,"kd":"article","f":"the judgment class","u":"https://observer.com/2026/07/future-ai-power-judgment-trust/","alt":["judgement","judgment","power","trust","who decides","models","leadership"]},{"t":"said","who":"Rahim Hirji","n":"Drift versus design: why most companies mistake activity for transformation","o":"CEOWORLD","d":"2026-07-09","ap":0,"kd":"article","f":"the Drift versus Design matrix: Sleepwalkers, Programmed, Stuck, Designers","u":"https://ceoworld.biz/2026/07/09/drift-versus-design-why-most-companies-mistake-activity-for-transformation/","alt":["drift","design","transformation","adoption","matrix","activity","usage theatre"]},{"t":"said","who":"Rahim Hirji","n":"Why the Real AI Risk is Not Automation, but Accountability Gaps in Leadership Decisions","o":"European Business Review","d":"2026-08-21","ap":0,"kd":"article","f":"human at the start, in the loop, at the end; the four accountability tests","u":"https://www.europeanbusinessreview.com/why-the-real-ai-risk-is-not-automation-but-accountability-gaps-in-leadership-decisions/","alt":["accountability","oversight","human in the loop","human at the start","decision","risk","board","nh predict"]},{"t":"said","who":"Rahim Hirji","n":"Why HR Must Shift from Skills Management to Capability Design","o":"WorldatWork","d":"2026-08-11","ap":0,"kd":"article","f":"capability design over skills management","u":"https://worldatwork.org/publications/workspan-daily/why-hr-must-shift-from-skills-management-to-capability-design","alt":["hr","chro","skills","capability","workforce","reskilling","people"]},{"t":"said","who":"Rahim Hirji","n":"You're not adopting AI. You're paying for it.","o":"Irish Tech News","d":"2026-07-22","ap":0,"kd":"article","f":"usage theatre","u":"https://irishtechnews.ie/youre-not-adopting-ai-youre-paying-for-it","alt":["adoption","usage theatre","licences","spend","productivity","measurement"]},{"t":"said","who":"Rahim Hirji","n":"Why AI doesn't create bad decisions, it just exposes them faster","o":"Entrepreneur UK","d":"2026-07-07","ap":0,"kd":"article","f":"AI exposes decision quality","u":"https://uk.entrepreneur.com/technology/ai-amplifies-human-decisions-not-rogue-ai","alt":["decision","decision quality","judgement","speed","leadership","mistakes"]},{"t":"said","who":"Rahim Hirji","n":"The Human Skills That Will Define Success Beyond Artificial Intelligence","o":"Luxurist Magazine","d":"2026-05-28","ap":0,"kd":"article","f":"the seven SuperSkills","u":"https://luxuristmag.com/the-human-skills-that-will-define-success-beyond-artificial-intelligence/","alt":["human skills","superskills","curiosity","empathy","success","careers"]},{"t":"said","who":"Rahim Hirji","n":"The Architecture of Drift","o":"Box of Amazing","d":"2026-06-01","ap":1,"kd":"essay","f":"the Drift vs Design Matrix","u":"https://boxofamazing.substack.com/p/the-architecture-of-drift","alt":["drift","design","matrix","organisations","adoption","culture"]},{"t":"said","who":"Rahim Hirji","n":"Are You Flying or Are You Being Flown?","o":"Box of Amazing","d":"2026-04-01","ap":1,"kd":"essay","f":"algorithmic drift; skill atrophy; Air France 447","u":"https://boxofamazing.substack.com/p/are-you-flying-or-are-you-being-flown","alt":["algorithmic drift","deskilling","skill atrophy","automation","aviation","pilots","flight 447","over-reliance"]},{"t":"said","who":"Rahim Hirji","n":"The Agents Are Here. You're Just Not Paying Attention","o":"Box of Amazing","d":"2026-03-01","ap":1,"kd":"essay","f":"the fourth lens","u":"https://boxofamazing.substack.com/p/the-agents-are-here-youre-just-not","alt":["agents","ai agents","delegation","accountability","attention"]},{"t":"said","who":"Rahim Hirji","n":"We've Been the AI All Along","o":"LinkedIn","d":"2026-02-01","ap":1,"kd":"essay","f":"","u":"https://www.linkedin.com/pulse/weve-been-ai-all-along-rahim-hirji-pesye","alt":["human","machine","work","automation","what stays human"]},{"t":"said","who":"Rahim Hirji","n":"Knowledge Is No Longer Power","o":"Box of Amazing","d":"2025-11-01","ap":1,"kd":"essay","f":"","u":"https://boxofamazing.substack.com/p/knowledge-is-no-longer-power","alt":["knowledge","judgement","power","expertise","memory","learning"]},{"t":"said","who":"Rahim Hirji","n":"The Human + AI Era","o":"Box of Amazing","d":"2025-10-22","ap":0,"kd":"essay","f":"","u":"https://boxofamazing.substack.com/p/the-human-ai-era","alt":["human","collaboration","augmented","era","future of work"]},{"t":"said","who":"Rahim Hirji","n":"A Father's Letter to His Daughter Before University","o":"Box of Amazing","d":"2025-09-14","ap":0,"kd":"essay","f":"","u":"https://boxofamazing.substack.com/p/a-fathers-letter-to-his-daughter","alt":["university","students","children","study","parents","what to study","young people"]},{"t":"said","who":"Rahim Hirji","n":"BBC Radio 5 Live, Drive, on the Anthropic existential-risk story","o":"BBC Radio 5 Live","d":"2026-09-09","ap":0,"kd":"broadcast","f":"","u":"","alt":["bbc","anthropic","existential risk","is ai dangerous","safety","radio"]},{"t":"said","who":"Rahim Hirji","n":"BBC Radio East Midlands, interview on SuperSkills","o":"BBC Local Radio","d":"2026-07-08","ap":0,"kd":"broadcast","f":"","u":"https://www.bbc.co.uk/programmes/m002y8bb","alt":["bbc","superskills","book","human skills","radio"]},{"t":"said","who":"Rahim Hirji","n":"Rahim Hirji on Human Skills in the Age of AI","o":"Let's Crew & Riot (Spotify)","d":"2026-06-25","ap":0,"kd":"podcast","f":"","u":"https://open.spotify.com/episode/2wgX6dvyjWDqe8YwZT10v3","alt":["human skills","podcast","superskills","careers"]},{"t":"said","who":"Rahim Hirji","n":"Super Skills: The 7 Human Skills for the Age of AI","o":"Inside Learning (Learnovate)","d":"2026-05-29","ap":0,"kd":"podcast","f":"","u":"https://learnovatecentre.org/insights/podcasts/super-skills-the-7-human-skills-for-the-age-of-ai-with-rahim-hirji/","alt":["learning","l&d","podcast","superskills","skills"]},{"t":"said","who":"Rahim Hirji","n":"The Fourth Dimension: Building Human Capability into the AI Strategy","o":"aibl","d":"2026-05-08","ap":0,"kd":"talk","f":"","u":"https://www.youtube.com/watch?v=x-rdkQc9-34","alt":["strategy","capability","ai strategy","leadership","fourth dimension","board"]},{"t":"said","who":"Rahim Hirji","n":"SuperSkills: The Seven Human Skills for the Age of AI","o":"Kogan Page","d":"2026-07-03","ap":0,"kd":"book","f":"the seven SuperSkills; drift versus design","u":"/book","alt":["superskills","book","seven skills","curiosity","change readiness","big picture","empathy","global adaptability","principled innovation","augmented mindset"]},{"t":"said","who":"Rahim Hirji","n":"50 Emerging Technology Themes to watch out for in 2019","o":"Box of Amazing (author's archive)","d":"2018-12-23","ap":0,"kd":"essay","f":"soft skills as the X-factor in the workplace; human-plus-machine","u":"/research/predictions","alt":["predictions","themes","soft skills","future","2019","trends"]}]}