Each entry gives the study, how it was done, what it found, what it supports and, as plainly, what it does not. Grades describe what kind of evidence a source is; how to read them is on the main page. Every entry has a permanent link and a copyable citation.
Practitioner method#
Eberle, R. F. (1971). Scamper: Games for Imagination Development
D.O.K. Publishers. The 1971 first edition could not be opened: the Internet Archive scan of this title is the 1996 Prufrock Press edition, and the resolvable object below is the 2023 Routledge combined edition, DOI 10.4324/9781003423560. Both checked at source 10 September 2026
- Method
- A classroom activity book. Eberle assembles a seven-part mnemonic, Substitute, Combine, Adapt, Modify or Magnify, Put to other uses, Eliminate, Rearrange or Reverse, from the idea-spurring checklist in Alex Osborn's Applied Imagination of 1953. No study design.
- Finding
- SCAMPER is a rewriting of an older checklist into a mnemonic for children, and its content belongs to Osborn before Eberle.
- What it supports
- That SCAMPER is an established teaching device with a traceable lineage, and that neither the letters nor the method are anybody's recent invention.
- What it does not support
- 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 book.
Practitioner framework#
Wack, P. (1985). Scenarios: uncharted waters ahead
Harvard Business Review, September 1985
- Method
- An account by the head of Royal Dutch Shell's scenario planning group of how the method was developed and used, written after the events it describes. A participant account, not an evaluation.
- Finding
- Describes scenario planning as a way of changing managers' mental models rather than of predicting outcomes, and reports that Shell's scenarios had rehearsed an oil supply disruption before the 1973 shock arrived.
- What it supports
- Where scenario planning as a corporate practice came from, what it was intended to do, and that its stated purpose was rehearsal of mental models rather than forecast accuracy.
- What it does not support
- That Shell outperformed its competitors because of it. The claim comes from accounts written by the people who ran the scenarios and has never been benchmarked against any other firm, which is the usual form of this story and the reason it should not be cited as evidence of effect.
Practitioner method#
Ohno, T. (1988). Toyota Production System: Beyond Large-Scale Production
Productivity Press. First English printing 1988; the Japanese original, Toyota seisan hoshiki, is 1978. Publisher record and DOI confirmed at Taylor and Francis 10 September 2026, where the current reissue is dated 2019
- Method
- A plant director's account of a production system he built, setting out the five-times-why procedure with a worked example of a machine that stopped. No experiment, no sample, no control.
- Finding
- 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 the first answer.
- What it supports
- That the five-times-why has a dated, named industrial origin and belongs to Toyota rather than to whoever most recently put it on a slide.
- What it does not support
- 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 sixty years of adoption and no controlled test behind it.
Practitioner framework#
Heifetz, R. A. (1994). Leadership Without Easy Answers
Harvard University Press, 1994; the practice is developed further in Heifetz, Grashow and Linsky, The Practice of Adaptive Leadership, 2009
- Method
- A theory of adaptive leadership argued from case analysis and clinical teaching practice. No controlled study.
- Finding
- Proposes the balcony and the dance floor: a leader has to step off the floor to see the pattern in what is happening and then return to act in it, and the discipline is the movement between the two rather than a preference for either.
- What it supports
- That the move between immersion and perspective is a named and long-established leadership practice, predating any AI argument.
- What it does not support
- That leaders who make the move perform better. The account is argued from cases rather than measured.
Peer-reviewed#
Swan, G. E. and Carmelli, D. (1996). Curiosity and mortality in aging adults: A 5-year follow-up of the Western Collaborative Group Study
Psychology and Aging, 11(3)
- Method
- Prospective cohort. 1,118 men, mean age 70.6 at baseline, with an ancillary sample of 1,035 women.
- Finding
- Higher curiosity at baseline was associated with survival at five-year follow-up, and state curiosity remained significant after adjustment for other risk factors.
- What it supports
- That the association survives controlling for medical risk factors in one large cohort.
- What it does not support
- That curiosity extends life. Residual confounding, in particular underlying health driving both curiosity and survival, cannot be excluded.
Peer-reviewed#
Edmondson, A. (1999). Psychological Safety and Learning Behavior in Work Teams
Administrative Science Quarterly, 44(2)
- Method
- Multi-method field study of 51 work teams in a single manufacturing company.
- Finding
- Team psychological safety predicted learning behaviour, which in turn mediated the relationship with team performance.
- What it supports
- That the route from safety to performance runs through learning behaviour rather than directly.
- What it does not support
- Causation, or generalisation beyond one manufacturing firm. It is a correlational field study.
Practitioner framework#
Meadows, D. H. (1999). Leverage points: places to intervene in a system
The Sustainability Institute, 1999, developing an essay first published in Whole Earth Review, 1997
- Method
- An argued ranking of twelve kinds of intervention in a system, from parameters at the weakest to paradigms and the ability to transcend paradigms at the strongest. Developed from systems modelling practice; no trial and no dataset.
- Finding
- Argues that interventions people reach for first, adjusting numbers and parameters, are the least powerful, and that the most powerful are changes to goals, to the paradigm from which the system arises, and to the capacity to hold paradigms lightly.
- What it supports
- That there is a published, widely used ordering of where in a system an intervention does most work, and that it is the canonical answer to the question of which level to act at.
- What it does not support
- The ordering itself, which is argued rather than measured. Meadows states in the essay that the list is not to be taken as precise and that leverage points are frequently counter-intuitive and often pushed in the wrong direction.
Peer-reviewed#
Pulakos, E. D., Arad, S., Donovan, M. A. and Plamondon, K. E. (2000). Adaptability in the Workplace: Development of a Taxonomy of Adaptive Performance
Journal of Applied Psychology, 85(4)
- Method
- Content analysis of over 1,000 critical incidents drawn from 21 jobs, followed by scale validation.
- Finding
- Eight dimensions of adaptive performance, including handling emergencies, managing stress, solving problems creatively and dealing with uncertain situations.
- What it supports
- That adaptability is decomposable into observable behaviours rather than being a single trait.
- What it does not support
- That employers reward these dimensions, or that they transfer across every occupation. The taxonomy was derived, not tested against outcomes.
Compiled review#
Simons, T. (2002). The High Cost of Lost Trust
Harvard Business Review, September 2002
- Method
- Survey of more than 6,500 employees at 76 Holiday Inn hotels in the United States and Canada, matched to hotel financial records.
- Finding
- 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 as large an effect on profits.
- What it supports
- That whether managers are seen to mean what they say tracks measurable financial outcomes.
- What it does not support
- Causation. It is a correlational study within one hotel chain, published in a practitioner magazine rather than a peer-reviewed journal.
Peer-reviewed#
Orlitzky, M., Schmidt, F. L. and Rynes, S. L. (2003). Corporate Social and Financial Performance: A Meta-Analysis
Organization Studies, 24(3)
- Method
- Meta-analysis of 52 studies, 33,878 observations.
- Finding
- A positive association between corporate social performance and financial performance, which the authors describe as bidirectional.
- What it supports
- That principled conduct and financial results are not in general tension.
- What it does not support
- Causation in either direction, and the relationship is not uniform across how social performance is operationalised.
Practitioner framework#
Snowden, D. J. and Boone, M. E. (2007). A leader's framework for decision making
Harvard Business Review, November 2007
- Method
- A sensemaking framework argued in an edited venue, sorting situations into simple, complicated, complex and chaotic domains, with a different mode of action appropriate to each. No dataset.
- Finding
- Argues that the error leaders make is applying the practice of one domain in another, most often treating a complex situation as a complicated one and demanding an analysis that the situation cannot support.
- What it supports
- That there is an established framework whose whole purpose is telling a decision maker which kind of situation they are in before they choose how to act, which is the nearest existing relative of any altitude-selection device.
- What it does not support
- That sorting a situation correctly improves outcomes. Cynefin is a sensemaking aid, has been revised repeatedly by its author, and reports no measurement.
Peer-reviewed#
Maddux, W. W. and Galinsky, A. D. (2009). Cultural Borders and Mental Barriers: The Relationship Between Living Abroad and Creativity
Journal of Personality and Social Psychology, 96(5)
- Method
- Five studies with MBA and undergraduate participants, combining correlational designs with causal priming experiments.
- Finding
- Time spent living abroad predicted success on creative-insight tasks and creative negotiation outcomes. Time spent travelling abroad did not.
- What it supports
- That adaptation to living in another culture, rather than exposure to it, is what relates to creativity.
- What it does not support
- A specific effect size for the general population. The samples are business and undergraduate students.
Peer-reviewed#
Harrison, S. H., Sluss, D. M. and Ashforth, B. E. (2011). Curiosity adapted the cat: The role of trait curiosity in newcomer adaptation
Journal of Applied Psychology, 96(1)
- Method
- Longitudinal field study of 123 newcomers across 12 call-centre organisations.
- Finding
- Specific curiosity predicted information-seeking from colleagues, which in turn was associated with more creative handling of customer problems.
- What it supports
- That curiosity operates through a behavioural mechanism, asking, rather than as a disposition on its own.
- What it does not support
- That curiosity can be trained into people, or that the effect holds outside newcomer adaptation.
Practitioner framework#
Kanter, R. M. (2011). Managing yourself: zoom in, zoom out
Harvard Business Review, March 2011
- Method
- An argued framework in an edited management venue, drawing on the author's observation of leaders. No new data.
- Finding
- Proposes that leaders operate along a continuum between a close-in view, where relationships and particulars dominate, and a distant view, where patterns and principles dominate, and that the failure is getting stuck at either end rather than preferring one.
- What it supports
- That the zoom metaphor for levels of attention is long established in the management literature, and that the failure it names is fixity rather than choice of level.
- What it does not support
- That moving between the two improves decisions. No measurement is reported and no instrument is offered.
Peer-reviewed#
Konrath, S. H., O'Brien, E. H. and Hsing, C. (2011). Changes in Dispositional Empathy in American College Students Over Time: A Meta-Analysis
Personality and Social Psychology Review, 15(2)
- Method
- Cross-temporal meta-analysis of 72 samples of American college students, total N 13,737, from 1979 to 2009.
- Finding
- Empathic Concern fell by 48 per cent and Perspective Taking by 34 per cent across the period, with most of the decline after 2000.
- What it supports
- That self-reported dispositional empathy declined measurably in this population over three decades.
- What it does not support
- A cause. The authors speculate about individualism and media but test no mechanism. It is also American college students only.
Peer-reviewed#
Sadri, G., Weber, T. J. and Gentry, W. A. (2011). Empathic emotion and leadership performance: An empirical analysis across 38 countries
The Leadership Quarterly, 22(5)
- Method
- 360-degree ratings of 6,731 mid to upper-level managers across 38 countries. Subordinates rated empathy, superiors rated performance.
- Finding
- Managers rated as more empathic by subordinates received higher performance ratings from their own superiors, with the effect moderated by national power distance.
- What it supports
- That the association holds at scale and across many national contexts.
- What it does not support
- That empathy training improves performance. The design is cross-sectional and correlational.
Peer-reviewed#
von Stumm, S., Hell, B. and Chamorro-Premuzic, T. (2011). The Hungry Mind: Intellectual Curiosity Is the Third Pillar of Academic Performance
Perspectives on Psychological Science, 6(6)
- Method
- Path-model synthesis of prior meta-analytic correlation matrices. Component samples range from 608 to 28,471.
- Finding
- Intellectual curiosity predicts academic performance independently of intelligence and effort, which the authors describe as a third pillar.
- What it supports
- That curiosity carries predictive weight that conscientiousness and ability do not account for.
- What it does not support
- 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.
Peer-reviewed#
Huang, J. L., Ryan, A. M., Zabel, K. L. and Palmer, A. (2014). Personality and Adaptive Performance at Work: A Meta-Analytic Investigation
Journal of Applied Psychology, 99(1)
- Method
- Meta-analysis of 71 independent samples, total N 7,535.
- Finding
- Emotional stability and ambition predict adaptive performance, and the pattern differs from predictors of routine task performance.
- What it supports
- That individual differences carry predictive weight for adaptive performance specifically.
- What it does not support
- That adaptive performance is a distinct construct. This paper assumes the construct from earlier taxonomy work rather than establishing it.
Peer-reviewed#
Mehta, R., Zhu, R. and Meyers-Levy, J. (2014). When Does a Higher Construal Level Increase or Decrease Indulgence? Resolving the Myopia versus Hyperopia Puzzle
Journal of Consumer Research, 41(2)
- Method
- Multi-study laboratory experiments.
- Finding
- Where the self is focal, a higher construal level increases indulgence rather than reducing it, reversing the effect the earlier literature predicted.
- What it supports
- That the benefit of stepping back is conditional, and the condition is identifiable.
- What it does not support
- That distant-future thinking is generally counterproductive. The effect is moderated, not reversed outright.
Institutional survey#
United States Environmental Protection Agency (2015). Notice of Violation, Volkswagen Group
US EPA enforcement record
- Method
- Regulatory enforcement notice.
- Finding
- Affected 2.0-litre vehicles emitted nitrogen oxides at up to 40 times the standard in normal driving while appearing compliant in laboratory testing.
- What it supports
- That the defeat device produced a measured gap between test and road conditions of that magnitude.
- What it does not support
- That the same multiple applies across the range. The separate 3.0-litre notice cited up to nine times.
Institutional modelling#
Barton, D., Manyika, J., Koller, T., Palter, R., Godsall, J. and Zoffer, J. (2017). Measuring the Economic Impact of Short-Termism
McKinsey Global Institute
- Method
- Corporate Horizon Index applied to 615 large and mid-cap United States public companies, 2001 to 2014.
- Finding
- 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.
- What it supports
- That a measurable long-term orientation tracks with stronger cumulative growth in this sample.
- What it does not support
- Causation, or application outside large United States listed companies. The index is a constructed measure, not an observed policy.
Practitioner framework#
Woodward, I. C., with Charan, R. (2017). The three altitudes of leadership
INSEAD Knowledge, 27 October 2017. Read at source 26 September 2026
- Method
- A framework developed from decades of observing chief executives and leaders, set out in an institutional knowledge publication. No study design, no sample and no dataset are reported.
- Finding
- Proposes that leaders think at three altitudes, 50,000 feet for the strategic panorama, 50 feet for planning and execution, and 5 feet for self-awareness, and that effective leaders move between all three. Names a pathology, altitude sickness, in which a leader is trapped at one altitude, and states that around 70 per cent of senior executives display it, with the largest group stuck at 50 feet.
- What it supports
- That the altitude metaphor for levels of leadership thinking is established, named and institutionally published, and that the failure it describes is an inability to move between levels.
- What it does not support
- The 70 per cent figure. No study, sample or measurement is cited for it anywhere in the article, and it rests on the authors' observation. Nothing here establishes that moving between altitudes improves any outcome.
Peer-reviewed#
Bratsberg, B. and Rogeberg, O. (2018). Flynn effect and its reversal are both environmentally caused
Proceedings of the National Academy of Sciences, 115(26), 6674-6678. Published online 11 June 2018, in issue 26 June 2018. Edited by Richard E. Nisbett and approved 14 May 2018. PMID 29891660. Full text read at source on 23 September 2026.
- Method
- Analysis of Norwegian administrative registers linked to military conscription cognitive testing at age 18 to 19, birth cohorts 1962 to 1991, restricted to native-born men with two native-born parents present in Norway on their eighteenth birthday. Overall sample 817,611, of whom 736,808 (90.1 per cent) had a valid ability score. Family fixed-effects models on scored brothers (n=355,438 families with at least two scored brothers), plus a Bayesian model correcting for selection into scoring, since conscription test coverage fell from 93 per cent for the 1980 cohort to 83 per cent for 1991 and the missing were disproportionately the lower-scoring.
- Finding
- Mean IQ rose from 99.5 for the 1962 birth cohort to 102.3 for 1975, then fell to 99.4 by 1989 and 99.7 by 1991. The rise, the turning point and the fall are all recoverable from WITHIN-family variation, comparing brothers. Selection-corrected estimates give 0.20 IQ points per year of increase within families for 1962 to 1975 against 0.18 across families, and 0.33 points per year of decline within families for 1975 to 1991 against 0.34 across families.
- What it supports
- That both the rise and the reversal of the Flynn effect in this population are environmental and operate within families, which rules out dysgenic fertility and immigration composition, the two explanations most often offered for falling scores. It also fixes the date: the turning point is the 1975 birth cohort, tested around 1993, and the last falling cohort was tested around 2009.
- What it does not support
- Which environmental factor is responsible. The authors state they cannot identify the causal structure, and list changing media exposure, educational change, and nutrition or health among the hypotheses their design leaves standing. It is one country, men only, one conscription test, and cohorts born no later than 1991, so it says nothing whatever about any technology introduced since. IT IS NOT EVIDENCE THAT AI LOWERS OR RAISES INTELLIGENCE, in either direction; the whole series predates generative AI. Its use in this research is to date the decline people attribute to AI, not to explain it.
Compiled review#
Gino, F. (2018). The Business Case for Curiosity
Harvard Business Review, September to October 2018
- Method
- Survey of more than 3,000 employees across a range of firms, reported in a practitioner magazine.
- Finding
- 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.
- What it supports
- That the stated value of curiosity and the felt experience of it diverge sharply inside organisations.
- What it does not support
- Any causal link between curiosity and a business outcome. Self-report, not peer-reviewed, and the sample is not nationally representative.
Peer-reviewed#
Howick, J., Moscrop, A., Mebius, A. et al. (2018). Effects of empathic and positive communication in healthcare consultations: a systematic review and meta-analysis
Journal of the Royal Society of Medicine, 111(7)
- Method
- Systematic review and meta-analysis of 28 randomised trials, 6,017 patients in total. Seven of the 28 tested empathic communication specifically; the remainder tested positive-expectation messaging.
- Finding
- 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.
- What it supports
- That empathic communication has a measurable effect on patient-reported outcomes.
- What it does not support
- 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.
Institutional survey#
Lorenzo, R., Voigt, N., Tsusaka, M., Krentz, M. and Abouzahr, K. (2018). How Diverse Leadership Teams Boost Innovation
Boston Consulting Group
- Method
- Survey of more than 1,700 companies across eight countries.
- Finding
- 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.
- What it supports
- That reported diversity and reported innovation revenue move together at scale.
- What it does not support
- 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.
Peer-reviewed#
Stillman, P. E., Fujita, K., Sheldon, O. and Trope, Y. (2018). From 'Me' to 'We': The Role of Construal Level in Promoting Maximized Joint Outcomes
Organizational Behavior and Human Decision Processes, 147
- Method
- Four laboratory and online experiments, pooled N approximately 691.
- Finding
- Prompting a higher level of construal led participants to choose options that maximised joint outcomes, including where doing so reduced their own payoff.
- What it supports
- That the level at which a problem is framed changes whether people optimise for themselves or for the whole.
- What it does not support
- Field behaviour. These are economic games with student and online samples.
Peer-reviewed#
Schlaegel, C., Richter, N. F. and Taras, V. (2021). Cultural intelligence and work-related outcomes: A meta-analytic examination of joint effects and incremental predictive validity
Journal of World Business, 56(4)
- Method
- Meta-analysis of 70 studies providing 80 independent samples, total N 18,359.
- Finding
- Cultural intelligence is moderately associated with work-related outcomes, with a reliability-corrected average effect of about .39.
- What it supports
- That the association is consistent across a large body of studies.
- What it does not support
- Causation, and it does not establish any single dimension as the strongest predictor. The paper is about joint effects across all four dimensions.
Peer-reviewed#
De Freitas, J., Uguralp, A. K., Oguz-Uguralp, Z., Paul, L. A., Tenenbaum, J. and Ullman, T. D. (2023). Self-orienting in human and machine learning
Nature Human Behaviour, 7
- Method
- Behavioural experiments with 124 human players across custom games, benchmarked against deep reinforcement learning agents.
- Finding
- 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.
- What it supports
- That rapid self-orientation after an unexpected change is currently a human advantage over the algorithms tested.
- What it does not support
- General workplace adaptability. These are simple custom games, the sample is modest, and the comparison is against specific algorithms rather than all AI approaches.
Compiled review#
Lightcast (2025). Beyond the Buzz: Developing the AI Skills Employers Actually Need
Lightcast, July 2025
- Method
- Analysis of over 1.3 billion job postings collected by Lightcast from more than 160,000 online sources, comparing postings that request AI skills with those that do not, with a focus on 2024 and growth since 2022. Five career areas are examined: marketing and PR, HR, finance, science and research, and education and training.
- Finding
- 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 non-tech roles grew 800% since 2022; AI skill demand grew 66% in HR postings and 40% in finance, and 8% of marketing and PR postings request AI skills with 50% annual growth. The highest wage premiums appear in non-technical occupations such as architects and lawyers.
- What it supports
- That employer demand for AI skills has spread beyond technical occupations and is paired with higher advertised pay, especially where AI is combined with domain expertise.
- What it does not support
- 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 sells labour market data and skills taxonomies.
Institutional synthesis#
World Economic Forum (2025). New Economy Skills: Unlocking the Human Advantage
World Economic Forum white paper, 3 December 2025. Publication page read at source 26 September 2026; the report PDF is served from a host that refuses automated fetching and was not read
- Method
- Synthesis drawing on data from education industry and workforce technology providers, a review of existing research, and expert consultations. Proposes a framework for assessing, developing and credentialling human skills.
- Finding
- Argues that human-centric skills, the capabilities that let individuals, organisations and societies adapt to change and lead transformation, are central to business and economic performance, and that the open question is no longer whether they matter but how they are developed, assessed and credentialled.
- What it supports
- That the assessment and credentialling of human capability is now an active question for a standing international body, and that the gap it names is the same one this research names from the organisational side.
- What it does not support
- That any assessment method works. No instrument is validated in the publication page, no measurement is reported, and the detail of the proposed framework was not read because the report host refuses automated access.
Working paper#
Autor, D., Rodchenko, T., Martin, J., Iscenko, Z., Strand, S., Pearl, D. and Ferere, M. (2026). Does AI Assistance Enhance or Erode Expertise? Evidence from a Three-Month Field Experiment in Patent Drafting
NBER Working Paper 35720, issued September 2026, with the authors' account on the Google Research blog of 7 October 2026. Abstract page and blog read at source 9 October 2026; the full paper was not read
- Method
- Pre-registered randomised controlled trial (AEA RCT Registry AEARCTR-0015823). 133 practising patent lawyers at eleven US intellectual property firms over three months; two-thirds at each firm were given a then-unreleased Google AI patent drafting assistant and the rest some AI training and no tool. All work was scored by blinded expert patent attorneys. At three months every lawyer redlined a patent application without AI. Juniors are lawyers with under seven years of experience.
- Finding
- With AI, quality on benchmark drafting tasks rose by 0.34 SD at 10 days (p = 0.03) and 0.38 SD at 90 days (p = 0.01), 'with larger gains among junior lawyers'. On the unassisted task 'Treated lawyers outperformed controls by 0.32 SD (p = 0.04), but this advantage was concentrated entirely among senior lawyers (0.45 SD, p = 0.02).' 'Junior lawyers showed no average gain; their scores instead bifurcated, with sharply fewer mediocre scores offset by more poor and more good ones.' 'The largest gains from AI thus accrued to the lawyers who retained the least.'
- What it supports
- That in one profession, scored by blinded graders, the gain in work delivered with AI and the gain in unaided judgement came apart, and that the unaided gain went to the lawyers with more experience.
- What it does not support
- That AI harms juniors: no group did worse on average. Anything beyond 133 lawyers at top-tier firms over three months; the sizes of the junior and senior groups are not on the pages read. Google paid the direct costs, five of the seven authors are Google employees and a sixth is a paid contractor, the tool was Google's own and the firms do business with Google. Not peer-reviewed.
Simulation model#
Catrinescu, I. (2026). The Capability Transfer: Cohort Reversal, Codified Expertise, and the Adaptability Tax of AI Transformation
arXiv 2610.33102, working paper, submitted 22 October 2026, also on SSRN. Abstract read at source 30 October 2026; the full text was rate-limited
- Method
- A two-cohort model of juniors and seniors extending Caosun and Aral's 2026 single-worker model, calibrated to a baseline scenario; the author states two key parameters are unmeasured and presents the results as scenarios rather than forecasts.
- Finding
- Because junior work is the most codified, demand for juniors falls first; the senior stock falls later and more severely because the pipeline stops replacing experts who leave. In the baseline, junior demand falls about a fifth below its 2022 level before mostly recovering, the senior stock falls by about a quarter and later halves, and developing an expert internally takes about nine years instead of five. A manager with a three-year horizon adopts AI in a way that cuts the wage bill sharply and leaves the firm less capable on work AI cannot do. Three proposed indicators: unaided practice juniors get on non-automatable work, senior time on such work, and years to develop an expert.
- What it supports
- That a formal model of the expertise pipeline produces the sequence this research describes, juniors first and seniors later and worse, and offers three things an organisation could count.
- What it does not support
- Any measured outcome: the parameters that drive the result are unmeasured by the author's own statement, and the paper is unreviewed.
Working paper#
Herrera-Poyatos, D., Peláez-González, C., Zuheros, C. and Herrera, F. (2026). What Human-AI Interaction Leaves Behind: Residual Human States. A Trajectory-Based Framework and Multi-Study Analysis
arXiv:2610.18765 (cs.HC), preprint, submitted 13 October 2026 and listed the following week. Read at source 23 October 2026
- Method
- A conceptual framework with four propositions, applied in a re-analysis of four published studies: a medical training study, a diagnostic learning study, a classroom AI tutoring experiment and a longitudinal re-analysis of personalised AI dialogue. No new data.
- Finding
- Assisted performance does not determine the state a person carries forward. In one case generative AI tutoring 'substantially improved assisted practice performance' 'without a corresponding detectable gain on a subsequent unaided exam'; in another, a history of AI interaction transmitted 'a systematic decision bias' to later unaided cases. The authors propose 'residual human states' as the thing to measure, with propositions of multidimensionality, interaction-policy dependence, horizon dissociation and temporal persistence.
- What it supports
- That the gap between assisted output and unaided capability can be seen in several existing studies when they are read for what remains after the tool is withdrawn, and that a framework now exists for measuring it.
- What it does not support
- Anything new empirically: it is a re-reading of others' studies, not peer-reviewed, and the affiliations and the original studies' details are as the abstract gives them.
Argued perspective#
Kärki, K. (2026). Who is Responsible for Tackling Professional Deskilling?
Human Affairs, 23 September 2026, doi 10.1515/humaff-2026-0080; University of Helsinki release via Newswise, 27 October 2026. Read through the release 30 October 2026
- Method
- A philosophical analysis in a peer-reviewed journal, applying theories of moral responsibility to skill loss from AI use. No empirical data. Read through the university's press release rather than the paper.
- Finding
- 'One of the problems with deskilling is that professionals have difficulty noticing it in their everyday lives.' Theories of responsibility hold an agent responsible for actions done with awareness, so 'It's therefore unclear whether we can be held responsible for the insidious deterioration of our skills'; responsibility is societal and structural. Recommendations: make professionals aware; test core skills regularly, including in seasoned staff; restrict some tasks to humans only; workplace AI guidelines that consider skill decay; stages in professional education during which AI use is prohibited; financial incentives to preserve skills; legislation where deskilling threatens critical processes; professional communities to define the core skills to be kept in human hands. Pilots practising autopilot failure is the model.
- What it supports
- That a peer-reviewed philosophical argument places the duty to prevent AI deskilling on professions and employers rather than individuals, on the ground that the loss is hard to notice from inside.
- What it does not support
- That deskilling occurs or how much: the paper argues, and reports no measurement. Its wording beyond the quoted lines is known here only as the release gave it.
Institutional survey#
Pearson, with Amazon Web Services (2026). AI Readiness: Building the Bridge from Higher Education to Work
Pearson, April 2026
- Method
- Survey of 2,711 undergraduate students, university educators and administrators, and employers from small, mid-sized and large firms in Brazil, Malaysia, Saudi Arabia, the United States, the United Kingdom and Vietnam. Fieldwork dates are not stated; survey data is combined with secondary sources from the World Economic Forum, Stanford HAI and UNESCO.
- Finding
- 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 communication and collaboration (50%) and adaptability (45%) as their highest priorities for graduates, and 42% of employers cite lack of hands-on experience with workplace AI tools as a major barrier. 63% of all respondents view a degree as more essential than five years ago.
- What it supports
- That across six countries employers and universities report different views of graduate readiness, and that employers name human skills ahead of AI tool skills.
- What it does not support
- 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 measure of actual graduate performance.
Compiled review#
Storey, M.-A. (2026). From Technical Debt to Cognitive and Intent Debt: Rethinking Software Health in the Age of AI
ACM Queue, preprint at arXiv 2603.22106, March 2026
- Method
- Conceptual synthesis by a Canada Research Chair at the University of Victoria, drawing on Naur (1985) on programming as theory building, Cunningham (1993) on technical debt, and the author's own empirical work with Starr on developer confusion. Illustrated with a single teaching anecdote rather than a study. Proposes a framework; presents no new data.
- Finding
- 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 constraints. Argues generative AI may reduce technical debt while accelerating the other two, because code can now be produced faster than a team can build the understanding needed to change it safely.
- What it supports
- That the debt metaphor has been extended into a structured framework by a serious researcher, and that cognitive debt now carries a team-level meaning distinct from the individual-level one in Kosmyna et al. The distinction is the author's own and she states it explicitly.
- What it does not support
- 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 not appear on the MIT Media Lab's own publications list for that author; the citation is not relied on here.
Working paper#
Yi Duan, Ying Liu, Zirui Tang, Haodong Chen, Jun Zhou, Yumou Liu and 27 others (2026). The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement
arXiv 2609.11873, submitted 10 September 2026
- Method
- A 33-author preprint proposing a five-level roadmap of self-improvement autonomy (L1 improvement execution, L2 improvement strategy, L3 experience acquisition, L4 environment adaptation, L5 recursive inheritance), a review of industry systems against it, and a Headroom-Closed Index that normalises benchmark scores between an entry-year frontier (0) and perfect performance (100). Not peer-reviewed at the time of entry.
- Finding
- 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 bounded tasks. The paper places current industrial systems at levels L1 to L4 and describes genuine recursive meta-improvement (L5) as largely unrealised. In the systems it reviews, 'experts retain control over consequential corrections and deployment'.
- What it supports
- That a large group of researchers has written down what recursive self-improvement would mean, level by level, and by their own measure finds the field short of the last level. The index is a useful vocabulary for saying how far a capability has closed rather than whether it has arrived.
- What it does not support
- 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 the authors' construction and its entry-year frontiers are choices. The paper is a roadmap by people who want the destination, which is a reason to read it and a reason to read it carefully.
Where this evidence is used
The research that rests most on this section:
- Curiosity
- Models of big picture thinking
- Principled Innovation
- Change Readiness
- Capability debt
- Big Picture Thinking
- Empathy
- Global Adaptability
The other sections
- Judgement, thinking and offloading 79
- Learning, practice and expertise 92
- Human and AI collaboration 44
- Empathy, creativity and what stays human 36
- Work, jobs and the labour market 94
- The institutional record 180
- Professions and sectors 25
- Frontline, clinical and physical work 24
- The international picture 29
The whole base is also machine-readable at evidence.json under CC BY 4.0.
Reference · SS-2026-446 · How I use AI here
Cite this page
Hirji, R. (2026). The seven human capabilities. The SuperSkills evidence base, SS-2026-446. https://thesuperskills.com/research/evidence/human-capabilities. Last reviewed 3 November 2026.
An evidence review by Rahim Hirji, not peer-reviewed research. For a material claim, cite the underlying study as well; every study here carries its own permanent link.
How citations and IDs work