{ "name": "Essential works on AI and human capability", "description": "Essential works on human capability in the age of AI. Inclusion does not mean endorsement.", "url": "https://thesuperskills.com/research/essential-works", "compiledBy": "Rahim Hirji, The SuperSkills Intelligence Company", "licence": "CC BY 4.0", "lastReviewed": "2026-08-26", "count": 84, "classifications": { "canonical": { "label": "Core", "definition": "Sources anyone entering this field should know. Roughly twenty-five of them." }, "strong": { "label": "Strong evidence", "definition": "Peer-reviewed work, large empirical datasets, or institutional analysis credible enough to build on." }, "perspective": { "label": "Important perspective", "definition": "A serious argument that shapes the debate. Not empirical proof, and not treated as such here." }, "foundation": { "label": "Foundation", "definition": "Older work on cognition, expertise, automation, judgement and skill acquisition. Most of this predates AI and explains it better than most writing about AI." }, "watch": { "label": "Watch", "definition": "A living programme or dataset whose conclusions will change. Track it rather than cite it once." } }, "works": [ { "id": "stanford-hai-index-2026", "citeAs": "https://thesuperskills.com/research/essential-works#stanford-hai-index-2026", "classification": "canonical", "themes": [ "technology capability", "jobs" ], "authors": "Stanford HAI", "year": 2026, "title": "The 2026 AI Index Report", "publication": "Stanford Institute for Human-Centered AI, 9th edition", "type": "Institutional report", "url": "https://hai.stanford.edu/ai-index/2026-ai-index-report", "finding": "The most comprehensive annual account of AI capability, investment, adoption and economic effect, now with a labour chapter that engages directly with learning effects and early-career workers.", "evidenceStrength": "Compiled review. Aggregates many third-party sources; strong on breadth, not original research.", "superSkillsRead": "The master factual reference for what AI can actually do, and the right place to check a capability claim before repeating it. Its limitation is also its usefulness: it measures the machine side of the equation, the half this research does not cover. Use its public data rather than only its PDF.", "relatedPages": [ "/research/ai-workforce-strategy" ] }, { "id": "oecd-ai-and-skills-2026", "citeAs": "https://thesuperskills.com/research/essential-works#oecd-ai-and-skills-2026", "classification": "canonical", "themes": [ "skills", "jobs" ], "authors": "OECD", "year": 2026, "title": "AI and skills: What we know so far", "publication": "OECD policy brief, 5 June 2026", "type": "Policy brief", "url": "https://www.oecd.org/en/publications/ai-and-skills_f843b352-en/full-report.html", "finding": "Fewer than one percent of workers will need advanced AI-specific skills such as programming or model development, while most require digital, analytical and interpretive capability.", "evidenceStrength": "Institutional analysis. The headline figure derives from Green and Lamby (2023), built largely on job-posting data, so it describes demand as advertised.", "superSkillsRead": "The most quotable single finding for anyone being told to retrain as an AI specialist. It is the policy-grade version of the argument that tool fluency is table stakes rather than a career, and it comes from a body with no product to sell.", "relatedPages": [ "/research/why-learn-to-prompt-is-weak-career-advice", "/research/human-skills-in-the-age-of-ai" ] }, { "id": "microsoft-wti-2026", "citeAs": "https://thesuperskills.com/research/essential-works#microsoft-wti-2026", "classification": "canonical", "themes": [ "agency", "organisation design", "judgement" ], "authors": "Microsoft", "year": 2026, "title": "Agents, human agency, and the opportunity for every organization", "publication": "2026 Work Trend Index Annual Report, 5 May 2026", "type": "Institutional survey", "url": "https://www.microsoft.com/en-us/worklab/work-trend-index/agents-human-agency-and-the-opportunity-for-every-organization", "finding": "Trillions of Microsoft 365 signals plus 20,000 AI-using workers across ten markets, organised around human agency, judgement and the delegation of work to agents.", "evidenceStrength": "Institutional survey plus product telemetry. Large and current; the vendor sells the tools whose adoption it measures, and telemetry counts usage rather than value.", "superSkillsRead": "The closest major corporate research programme to this thesis, and worth watching for that reason. When the company with the most usage data starts organising its annual report around human agency and judgement allocation, the question has moved from the margins to the centre. Read it as a signal about the field as much as evidence about the world.", "relatedPages": [ "/research/ai-agents-and-human-judgement", "/research/how-should-leaders-respond-to-ai" ] }, { "id": "pwc-jobs-barometer-2026", "citeAs": "https://thesuperskills.com/research/essential-works#pwc-jobs-barometer-2026", "classification": "canonical", "themes": [ "early careers", "jobs", "skills" ], "authors": "PwC", "year": 2026, "title": "2026 Global AI Jobs Barometer: Two futures for jobs in an AI era", "publication": "PwC, 15 June 2026", "type": "Labour-market analysis", "url": "https://www.pwc.com/gx/en/news-room/press-releases/2026/pwc-2026-ai-jobs-barometer.html", "finding": "Analysing 2.4 million US entry-level jobs, entry-level roles most exposed to AI are seven times more likely to require traditionally senior human-intensive skills such as leadership and creativity; those roles grew 35 percent since 2019 while other entry-level roles shrank 10 percent.", "evidenceStrength": "Large-scale job-advertisement analysis across more than a billion adverts. Observed demand rather than stated preference, though adverts describe what employers ask for rather than what work requires. The entry-level finding is US-only.", "superSkillsRead": "The strongest external evidence for synthetic seniority anywhere. Entry-level roles are not disappearing so much as being seniorised, which means the junior work through which people used to become senior is being removed while the expectation of senior judgement arrives on day one. That is the missing rungs problem, measured.", "relatedPages": [ "/research/will-ai-replace-entry-level-jobs", "/research/synthetic-seniority", "/research/missing-rungs" ] }, { "id": "bastani-2025", "citeAs": "https://thesuperskills.com/research/essential-works#bastani-2025", "classification": "canonical", "themes": [ "learning", "education", "deskilling" ], "authors": "Bastani, H., Bastani, O., Sungu, A., Ge, H., Kabakci, O. and Mariman, R.", "year": 2025, "title": "Generative AI Without Guardrails Can Harm Learning: Evidence from High School Mathematics", "publication": "Proceedings of the National Academy of Sciences, 122(26)", "type": "Journal article", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4895486", "finding": "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 removed the harm.", "evidenceStrength": "Peer-reviewed study. What a guardrailed interface should look like for professional work. It was school mathematics over a bounded period.", "superSkillsRead": "The cleanest demonstration that the design of the tool, not the presence of AI, decides whether a person is taught or carried. The same model produced both the best and the worst learning outcome in one experiment.", "relatedPages": [ "/research/how-humans-learn-with-ai", "/research/chro-guide-to-ai" ], "gradedEntry": "https://thesuperskills.com/research/evidence#bastani-2025" }, { "id": "autor-thompson-2025", "citeAs": "https://thesuperskills.com/research/essential-works#autor-thompson-2025", "classification": "canonical", "themes": [ "jobs", "expertise", "skills" ], "authors": "Autor, D. and Thompson, N.", "year": 2025, "title": "Expertise", "publication": "NBER Working Paper 33941; Journal of the European Economic Association, 23(4), 1203-1271", "type": "Journal article", "url": "https://www.nber.org/papers/w33941", "finding": "Automation that removed the LESS expert tasks raised wages and reduced employment. Automation that removed the EXPERT tasks lowered wages and increased employment.", "evidenceStrength": "Peer-reviewed study. Anything measured about generative AI. The data ends in 2018, so this is a lens, not a forecast.", "superSkillsRead": "Gives the career question a testable form. Not whether AI will take your tasks, but whether removing them raises or lowers the expertise of what remains, which determines whether your role appreciates or commoditises.", "relatedPages": [ "/research/staying-valuable-in-the-age-of-ai", "/research/will-ai-replace-my-job" ], "gradedEntry": "https://thesuperskills.com/research/evidence#autor-thompson-2025" }, { "id": "lee-2025", "citeAs": "https://thesuperskills.com/research/essential-works#lee-2025", "classification": "canonical", "themes": [ "critical thinking", "judgement" ], "authors": "Lee, H.-P. et al.", "year": 2025, "title": "The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects from a Survey of Knowledge Workers", "publication": "Microsoft Research and Carnegie Mellon, CHI 2025", "type": "Journal article", "url": "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/", "finding": "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.", "evidenceStrength": "Peer-reviewed study. Causation. People who think differently may use AI differently, and a survey cannot separate the two.", "superSkillsRead": "The first large study of what AI does to professional thinking in real work. Its finding that thinking migrates to verification rather than disappearing is more precise, and more useful, than the deskilling headlines it generated.", "relatedPages": [ "/research/ai-and-critical-thinking", "/research/ai-and-human-judgement" ], "gradedEntry": "https://thesuperskills.com/research/evidence#lee-2025" }, { "id": "wef-foj-2025", "citeAs": "https://thesuperskills.com/research/essential-works#wef-foj-2025", "classification": "canonical", "themes": [ "skills", "jobs" ], "authors": "World Economic Forum", "year": 2025, "title": "The Future of Jobs Report 2025", "publication": "World Economic Forum, Geneva, January 2025", "type": "Institutional survey", "url": "https://www.weforum.org/publications/the-future-of-jobs-report-2025/", "finding": "Analytical thinking is the most valued core skill, and skills gaps are named the single biggest barrier to business transformation.", "evidenceStrength": "Institutional survey. What employers actually do. Stated skill preference and hiring behaviour diverge routinely.", "superSkillsRead": "The dataset everyone quotes. Useful for what employers say they want, which is a real signal, and not evidence of what they do.", "relatedPages": [ "/research/human-skills-in-the-age-of-ai", "/research/ai-workforce-strategy" ], "gradedEntry": "https://thesuperskills.com/research/evidence#wef-foj-2025" }, { "id": "vaccaro-2024", "citeAs": "https://thesuperskills.com/research/essential-works#vaccaro-2024", "classification": "canonical", "themes": [ "human-AI collaboration", "judgement" ], "authors": "Vaccaro, M., Almaatouq, A. and Malone, T.", "year": 2024, "title": "When combinations of humans and AI are useful: a systematic review and meta-analysis", "publication": "Nature Human Behaviour, 8, 2293-2303", "type": "Journal article", "url": "https://www.nature.com/articles/s41562-024-02024-1", "finding": "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 where the AI beat humans.", "evidenceStrength": "Peer-reviewed study. 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.", "superSkillsRead": "The most under-absorbed finding in the field. It is the empirical case against treating a human in the loop as a control, and it should change how oversight is designed rather than merely how it is described.", "relatedPages": [ "/research/human-ai-decision-making", "/research/how-should-leaders-respond-to-ai" ], "gradedEntry": "https://thesuperskills.com/research/evidence#vaccaro-2024" }, { "id": "eloundou-2024", "citeAs": "https://thesuperskills.com/research/essential-works#eloundou-2024", "classification": "canonical", "themes": [ "jobs", "skills" ], "authors": "Eloundou, T., Manning, S., Mishkin, P. and Rock, D.", "year": 2024, "title": "GPTs are GPTs: Labor market impact potential of LLMs", "publication": "Science, 384(6702), 1306-1308", "type": "Journal article", "url": "https://arxiv.org/abs/2303.10130", "finding": "Around 80 percent of US workers could have at least 10 percent of tasks affected; about 19 percent could see at least half affected.", "evidenceStrength": "Peer-reviewed study. 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.", "superSkillsRead": "The most misquoted number in the field. It measures task exposure, not displacement, and the authors say so explicitly. Included partly so that the distinction is easy to point at.", "relatedPages": [ "/research/will-ai-replace-my-job" ], "gradedEntry": "https://thesuperskills.com/research/evidence#eloundou-2024" }, { "id": "yin-2024", "citeAs": "https://thesuperskills.com/research/essential-works#yin-2024", "classification": "canonical", "themes": [ "agency", "ethics" ], "authors": "Yin, Y., Jia, N. and Wakslak, C. J.", "year": 2024, "title": "AI can help people feel heard, but an AI label diminishes this impact", "publication": "PNAS, 121(14), e2319112121", "type": "Journal article", "url": "https://pure.psu.edu/en/publications/ai-can-help-people-feel-heard-but-an-ai-label-diminishes-this-imp/", "finding": "AI-generated replies made recipients feel MORE heard than replies from untrained humans, and labelling the reply as AI removed the advantage.", "evidenceStrength": "Peer-reviewed study. That the label effect is stable. Norms around disclosed AI assistance are moving, and nobody has measured this over time.", "superSkillsRead": "The most clarifying study in the debate about what stays human. The words were not the thing being valued; the identity of whoever wrote them was.", "relatedPages": [ "/research/outsourced-recognition", "/research/what-stays-human" ], "gradedEntry": "https://thesuperskills.com/research/evidence#yin-2024" }, { "id": "mollick-co-intelligence", "citeAs": "https://thesuperskills.com/research/essential-works#mollick-co-intelligence", "classification": "canonical", "themes": [ "human-AI collaboration", "skills" ], "authors": "Mollick, E.", "year": 2024, "title": "Co-Intelligence: Living and Working with AI", "publication": "Portfolio", "type": "Book", "url": "https://www.penguinrandomhouse.com/books/741805/co-intelligence-by-ethan-mollick/", "finding": "Sets out practical principles for working with AI, including always inviting it to the table and treating it as a person while remembering it is not one.", "evidenceStrength": "Important perspective, grounded in the author's own experiments rather than controlled research.", "superSkillsRead": "The best mainstream book on practical human-AI collaboration, and the most useful single recommendation for someone starting out. Where this research differs is emphasis: Mollick is strongest on how to get value from the tool today, and lighter on what habitual use does to capability over years.", "relatedPages": [ "/research/using-ai-without-dependency" ] }, { "id": "vallor-ai-mirror", "citeAs": "https://thesuperskills.com/research/essential-works#vallor-ai-mirror", "classification": "canonical", "themes": [ "ethics", "agency" ], "authors": "Vallor, S.", "year": 2024, "title": "The AI Mirror: How to Reclaim Our Humanity in an Age of Machine Thinking", "publication": "Oxford University Press", "type": "Book", "url": "https://academic.oup.com/book/56292", "finding": "Argues that AI systems are mirrors reflecting our past rather than minds, and that the danger is our shrinking conception of what humans are.", "evidenceStrength": "Important perspective. Philosophy, not evidence.", "superSkillsRead": "The closest philosophical work to the question underneath this research. Vallor's concern is not that machines will surpass us but that we will define ourselves downwards to match them, which is the same worry as capability debt stated at the level of the species rather than the organisation.", "relatedPages": [ "/research/what-stays-human" ] }, { "id": "beane-skill-code", "citeAs": "https://thesuperskills.com/research/essential-works#beane-skill-code", "classification": "canonical", "themes": [ "expertise", "learning", "early careers" ], "authors": "Beane, M.", "year": 2024, "title": "The Skill Code: How to Save Human Ability in an Age of Intelligent Machines", "publication": "Harper Business", "type": "Book", "url": "https://harpercollins.co.uk/products/the-skill-code-how-to-save-human-ability-in-an-age-of-intelligent-machines-matt-beane", "finding": "Draws on field research into how novices actually learn from experts, and shows that intelligent machines are quietly removing the apprenticeship through which skill has always been transmitted.", "evidenceStrength": "Important perspective, built on the author's own ethnographic field research including robotic surgery.", "superSkillsRead": "The closest book to the missing-rungs argument, and essential reading for anyone interested in it. Beane arrived at the same problem from a different direction, through direct observation of surgical residents losing hands-on time to robotic systems, the same mechanism this research describes in knowledge work.", "relatedPages": [ "/research/missing-rungs", "/research/the-missed-reps-of-ai-seniority", "/research/how-humans-learn-with-ai" ] }, { "id": "autor-middle-class-2024", "citeAs": "https://thesuperskills.com/research/essential-works#autor-middle-class-2024", "classification": "canonical", "themes": [ "jobs", "expertise", "skills" ], "authors": "Autor, D.", "year": 2024, "title": "Applying AI to Rebuild Middle Class Jobs", "publication": "NBER Working Paper 32140, February 2024", "type": "Working paper", "url": "https://www.nber.org/papers/w32140", "finding": "Argues that AI, unlike previous automation, can extend expert decision-making to a wider set of workers, potentially rebuilding middle-skill work rather than hollowing it out further.", "evidenceStrength": "Working paper, not peer-reviewed. An argument built on economic reasoning and historical analogy rather than measurement of generative AI itself.", "superSkillsRead": "The strongest rigorous optimistic case in the field, and a genuine challenge to this research rather than a soft counterpoint. If Autor is right, AI restores the middle by putting expert judgement within reach of more people. If the capability argument on this site is right, that only holds where the expertise is genuinely acquired rather than borrowed from the machine. Both cannot be fully true, and that tension is worth sitting with.", "relatedPages": [ "/research/staying-valuable-in-the-age-of-ai", "/research/will-ai-replace-my-job" ] }, { "id": "acemoglu-macro-2024", "citeAs": "https://thesuperskills.com/research/essential-works#acemoglu-macro-2024", "classification": "canonical", "themes": [ "jobs", "organisation design" ], "authors": "Acemoglu, D.", "year": 2024, "title": "The Simple Macroeconomics of AI", "publication": "NBER Working Paper 32487; published in Economic Policy, 40(121), 2025, 13-58", "type": "Working paper", "url": "https://www.nber.org/papers/w32487", "finding": "Estimates that AI's effect on total factor productivity over ten years will be modest, roughly an order of magnitude smaller than the most quoted forecasts.", "evidenceStrength": "Peer-reviewed in its published form. Assumption-driven macro estimate; Acemoglu is explicit that the numbers depend on which tasks prove genuinely automatable.", "superSkillsRead": "The necessary corrective to every consultancy trillion-dollar figure, including several elsewhere in this Canon. It also complicates the urgency this research assumes: if the economic effect is small and slow, the capability effect may be too. Worth reading immediately after any report promising transformation.", "relatedPages": [ "/research/ai-workforce-strategy", "/research/how-should-leaders-respond-to-ai" ] }, { "id": "dellacqua-2023", "citeAs": "https://thesuperskills.com/research/essential-works#dellacqua-2023", "classification": "canonical", "themes": [ "human-AI collaboration", "judgement", "expertise" ], "authors": "Dell'Acqua, F. et al.", "year": 2023, "title": "Navigating the Jagged Technological Frontier: Field Experimental Evidence of the Effects of AI on Knowledge Worker Productivity and Quality", "publication": "Harvard Business School and BCG working paper", "type": "Working paper", "url": "https://papers.ssrn.com/sol3/papers.cfm?abstract_id=4573321", "finding": "Inside the frontier, AI-assisted consultants were dramatically better and faster. Outside it, they performed worse than consultants with no AI at all.", "evidenceStrength": "Working paper, not peer-reviewed. Where the frontier runs in your domain. That is local and must be learned.", "superSkillsRead": "The jagged frontier is the single most useful mental model for anyone using AI at work. It also shows the cost of not knowing where the frontier runs, which is the capability that does not commoditise.", "relatedPages": [ "/research/human-ai-decision-making", "/research/why-learn-to-prompt-is-weak-career-advice" ], "gradedEntry": "https://thesuperskills.com/research/evidence#dellacqua-2023" }, { "id": "brynjolfsson-2023", "citeAs": "https://thesuperskills.com/research/essential-works#brynjolfsson-2023", "classification": "canonical", "themes": [ "jobs", "expertise", "early careers" ], "authors": "Brynjolfsson, E., Li, D. and Raymond, L.", "year": 2023, "title": "Generative AI at Work", "publication": "NBER Working Paper 31161; Quarterly Journal of Economics, 2025", "type": "Journal article", "url": "https://www.nber.org/papers/w31161", "finding": "Productivity rose 14 percent on average, 34 percent for the newest and least experienced staff, and barely at all for the most skilled.", "evidenceStrength": "Peer-reviewed study. Whether those novices became experts. It measures output, not development, over months rather than years.", "superSkillsRead": "Establishes that AI raises the floor far more than the ceiling. Read alongside the missing-rungs argument, it poses the question the study itself does not answer: whether the novices it lifted ever became experts.", "relatedPages": [ "/research/how-humans-learn-with-ai", "/research/staying-valuable-in-the-age-of-ai" ], "gradedEntry": "https://thesuperskills.com/research/evidence#brynjolfsson-2023" }, { "id": "oecd-learning-compass-2030", "citeAs": "https://thesuperskills.com/research/essential-works#oecd-learning-compass-2030", "classification": "canonical", "themes": [ "frameworks", "education", "agency", "learning" ], "authors": "OECD", "year": 2019, "title": "OECD Learning Compass 2030", "publication": "OECD Future of Education and Skills 2030 project", "type": "Competency framework", "url": "https://www.oecd.org/en/data/tools/oecd-learning-compass-2030.html", "finding": "A framework placing student agency, and the capacity to set goals, reflect and act responsibly, at the centre of education, alongside knowledge, skills, attitudes and values.", "evidenceStrength": "Official framework developed through multi-country expert consultation. Evolving rather than a single dated report; normative, not empirical.", "superSkillsRead": "The most important framework on this list for anyone arguing about human capability, because it put agency at the centre years before AI made the argument urgent. Its anticipation-action-reflection cycle is a considerably more rigorous account of learning to learn than most current writing on the subject.", "relatedPages": [ "/research/how-humans-learn-with-ai", "/research/human-skills-in-the-age-of-ai" ] }, { "id": "agrawal-prediction-machines", "citeAs": "https://thesuperskills.com/research/essential-works#agrawal-prediction-machines", "classification": "canonical", "themes": [ "judgement", "organisation design" ], "authors": "Agrawal, A., Gans, J. and Goldfarb, A.", "year": 2018, "title": "Prediction Machines: The Simple Economics of Artificial Intelligence", "publication": "Harvard Business Review Press, updated edition 2022", "type": "Book", "url": "https://store.hbr.org/product/prediction-machines-updated-and-expanded-the-simple-economics-of-artificial-intelligence/10598", "finding": "AI reduces the cost of prediction. When prediction becomes cheap, the value of its complements rises, and the most important complement is human judgement.", "evidenceStrength": "Important perspective, built on economic reasoning rather than empirical measurement.", "superSkillsRead": "The cleanest economic articulation of why judgement becomes more valuable rather than less. It is the argument this research makes, arrived at from price theory rather than from capability. It is the version to give a sceptical executive.", "relatedPages": [ "/research/ai-and-human-judgement", "/research/staying-valuable-in-the-age-of-ai" ] }, { "id": "nap-how-people-learn-ii", "citeAs": "https://thesuperskills.com/research/essential-works#nap-how-people-learn-ii", "classification": "canonical", "themes": [ "learning", "education" ], "authors": "National Academies of Sciences, Engineering, and Medicine", "year": 2018, "title": "How People Learn II: Learners, Contexts, and Cultures", "publication": "National Academies Press", "type": "Research synthesis", "url": "https://nap.nationalacademies.org/catalog/24783/how-people-learn-ii-learners-contexts-and-cultures", "finding": "A consensus synthesis of learning science, covering how context, culture, motivation and prior knowledge shape what is actually learned.", "evidenceStrength": "Consensus report by an expert committee, the strongest form of evidence synthesis available in education research.", "superSkillsRead": "The reference to reach for before accepting any claim about AI and learning. Its central lesson, that learning is situated and depends on far more than information transfer, is why an AI tutor that answers well is not the same as one that teaches.", "relatedPages": [ "/research/how-humans-learn-with-ai" ] }, { "id": "carr-glass-cage", "citeAs": "https://thesuperskills.com/research/essential-works#carr-glass-cage", "classification": "canonical", "themes": [ "deskilling", "judgement" ], "authors": "Carr, N.", "year": 2014, "title": "The Glass Cage: Automation and Us", "publication": "W. W. Norton", "type": "Book", "url": "https://wwnorton.co.uk/books/9780393240764-the-glass-cage/formats", "finding": "Argues that automation degrades the skills and attention of the people it assists, drawing on aviation, medicine and architecture.", "evidenceStrength": "Important perspective. Journalistic synthesis of research rather than original evidence.", "superSkillsRead": "The historical warning, written before generative AI and largely vindicated by it. Carr's value is that he made the argument when it was contrarian and about autopilots rather than chatbots, and the current wave of AI-and-thinking writing so often reads as a rediscovery of it.", "relatedPages": [ "/research/ai-and-human-judgement", "/research/capability-debt" ] }, { "id": "bainbridge-1983", "citeAs": "https://thesuperskills.com/research/essential-works#bainbridge-1983", "classification": "canonical", "themes": [ "deskilling", "expertise", "judgement", "organisation design" ], "authors": "Bainbridge, L.", "year": 1983, "title": "Ironies of Automation", "publication": "Automatica, 19(6), 775-779", "type": "Journal article", "url": "https://ckrybus.com/static/papers/Bainbridge_1983_Automatica.pdf", "finding": "In the author's words, the more advanced a control system is, the more crucial the contribution of the human operator may become; and by taking away the easy parts of the task, automation can make the difficult parts harder.", "evidenceStrength": "Foundational theoretical paper, extensively validated in aviation and process control over four decades.", "superSkillsRead": "The single most important source on this list, and the least cited in current debate. Bainbridge described the missing-rungs and missed-reps problem in 1983, in process control: automation removes the routine work through which operators stayed practised, then depends on those same operators for the hard cases. Everything written about AI and deskilling since is a restatement of this paper, usually without knowing it.", "relatedPages": [ "/research/capability-debt", "/research/the-missed-reps-of-ai-seniority", "/research/missing-rungs" ] }, { "id": "polanyi-1966", "citeAs": "https://thesuperskills.com/research/essential-works#polanyi-1966", "classification": "canonical", "themes": [ "expertise", "learning" ], "authors": "Polanyi, M.", "year": 1966, "title": "The Tacit Dimension", "publication": "University of Chicago Press, current edition 2009 with a foreword by Amartya Sen", "type": "Book", "url": "https://press.uchicago.edu/ucp/books/book/chicago/T/bo6035368.html", "finding": "We can know more than we can tell. A large part of expert knowledge cannot be made explicit, and is acquired through practice rather than instruction.", "evidenceStrength": "Foundational philosophy, not empirical work.", "superSkillsRead": "The reason expertise cannot be written into a prompt. If the most valuable part of what an expert knows cannot be articulated, then a system trained on articulated knowledge is missing the part that matters, and a novice who never practises never acquires it.", "relatedPages": [ "/research/what-stays-human", "/research/how-humans-learn-with-ai" ] }, { "id": "mckinsey-agents-robots-2026", "citeAs": "https://thesuperskills.com/research/essential-works#mckinsey-agents-robots-2026", "classification": "strong", "themes": [ "jobs", "organisation design" ], "authors": "McKinsey Global Institute", "year": 2026, "title": "Agents, robots, and us: How AI reshapes work and skills in Europe", "publication": "McKinsey Global Institute, May 2026", "type": "Institutional analysis", "url": "https://www.mckinsey.com/mgi/our-research/agents-robots-and-us-how-ai-reshapes-work-and-skills-in-europe", "finding": "Models how agentic AI and robotics together reshape task composition and skill demand across Europe.", "evidenceStrength": "Institutional modelling. Outcomes. Task exposure modelling has consistently over-predicted the pace of realised change.", "superSkillsRead": "The most current institutional modelling of the agentic shift. Task-exposure modelling has consistently over-predicted pace.", "relatedPages": [ "/research/ai-agents-and-human-judgement", "/research/ai-workforce-strategy" ], "gradedEntry": "https://thesuperskills.com/research/evidence#mckinsey-agents-robots-2026" }, { "id": "allianz-labour-2026", "citeAs": "https://thesuperskills.com/research/essential-works#allianz-labour-2026", "classification": "strong", "themes": [ "jobs" ], "authors": "Allianz Research", "year": 2026, "title": "Happy Labor Day? How geopolitics, immigration and AI will reshape work", "publication": "Allianz Trade, 30 April 2026", "type": "Institutional analysis", "url": "https://www.allianz-trade.com/en_global/news-insights/economic-insights/Happy-labor-day-How-geopolitics-immigration-AI-reshape-work.html", "finding": "Models the combined effect of AI, demographics and migration on labour supply and task composition.", "evidenceStrength": "Institutional modelling. Measured effects. Like all exposure modelling, it is an estimate of what could be affected.", "superSkillsRead": "A cross-country modelling view from outside the consultancy sector, which is worth having for that reason alone.", "relatedPages": [ "/research/will-ai-replace-my-job" ], "gradedEntry": "https://thesuperskills.com/research/evidence#allianz-labour-2026" }, { "id": "ilo-changing-landscape-2026", "citeAs": "https://thesuperskills.com/research/essential-works#ilo-changing-landscape-2026", "classification": "strong", "themes": [ "skills", "jobs", "agency" ], "authors": "ILO, ETF, Cedefop, Eurofound, European Commission and UNESCO", "year": 2026, "title": "Changing landscape of skills in the age of AI", "publication": "Joint publication, 13 August 2026", "type": "Institutional report", "url": "https://www.ilo.org/publications/changing-landscape-skills-age-ai", "finding": "A six-agency synthesis on how AI is reshaping skill demand, with higher-order cognitive skills, socioemotional skills, adaptability and human agency as organising themes.", "evidenceStrength": "Institutional synthesis. Six agencies, so unusually broad, but a synthesis rather than new primary data.", "superSkillsRead": "Notable less for new evidence than for the language. When six intergovernmental bodies converge on human agency and higher-order cognition as the organising frame, the territory this research occupies has become the mainstream policy question. Cite it as ILO and others, not ILO alone.", "relatedPages": [ "/research/human-skills-in-the-age-of-ai" ] }, { "id": "oecd-skills-ai-age-2026", "citeAs": "https://thesuperskills.com/research/essential-works#oecd-skills-ai-age-2026", "classification": "strong", "themes": [ "skills", "education" ], "authors": "OECD", "year": 2026, "title": "Skills in the AI age", "publication": "OECD Artificial Intelligence Papers No. 60, July 2026", "type": "Policy analysis", "url": "https://www.oecd.org/en/publications/skills-in-the-ai-age_972bd15e-en/full-report/component-4.html", "finding": "A policy synthesis on how AI changes skill requirements across economies, rather than a forecast of job losses.", "evidenceStrength": "Institutional analysis, methodologically documented, not peer-reviewed.", "superSkillsRead": "The serious policy treatment of skills, as distinct from the consultancy treatment. Useful precisely because it resists the headline framing that dominates this subject.", "relatedPages": [ "/research/human-skills-in-the-age-of-ai" ] }, { "id": "bcg-deskilling-2026", "citeAs": "https://thesuperskills.com/research/essential-works#bcg-deskilling-2026", "classification": "strong", "themes": [ "deskilling", "judgement", "leadership" ], "authors": "BCG Henderson Institute", "year": 2026, "title": "When Everyone Uses AI, Companies Risk Losing Critical Skills", "publication": "Boston Consulting Group, 17 June 2026", "type": "Executive survey", "url": "https://www.bcg.com/publications/2026/when-everyone-uses-ai-companies-risk-critical-skills", "finding": "Half of the executives surveyed report already observing deskilling in their organisations, and more than sixty percent expect it to become a material threat within three to five years.", "evidenceStrength": "Executive survey, and the base is small: seventy C-suite and senior leaders. Directionally interesting, statistically light.", "superSkillsRead": "The most direct external corroboration of capability debt available, and it must be cited honestly. Seventy executives is a signal that senior people are noticing something, not a measurement of how widespread it is. Anyone quoting the fifty percent without the base of seventy is overstating it.", "relatedPages": [ "/research/capability-debt", "/research/ai-workforce-strategy" ] }, { "id": "anthropic-81k-2026", "citeAs": "https://thesuperskills.com/research/essential-works#anthropic-81k-2026", "classification": "strong", "themes": [ "jobs", "early careers" ], "authors": "Massenkoff, M. and Huang, S. (Anthropic)", "year": 2026, "title": "What 81,000 people told us about the economics of AI", "publication": "Anthropic, 22 April 2026", "type": "Survey", "url": "https://www.anthropic.com/research/81k-economics", "finding": "A survey of 80,508 respondents on productivity, displacement anxiety and the economics of AI use.", "evidenceStrength": "Very large sample, self-selected towards AI users. Strong on scale, weak on representativeness.", "superSkillsRead": "Useful mainly for early-career anxiety and perceived displacement, which are real economic forces regardless of whether the displacement itself materialises. Read alongside the Danish administrative data, which finds no measured earnings effect at all.", "relatedPages": [ "/research/will-ai-replace-my-job" ] }, { "id": "pearson-aws-ai-readiness-2026", "citeAs": "https://thesuperskills.com/research/essential-works#pearson-aws-ai-readiness-2026", "classification": "strong", "themes": [ "education", "early careers", "skills" ], "authors": "Pearson and AWS", "year": 2026, "title": "AI Readiness: Building the Bridge from Higher Education to Work", "publication": "Pearson and Amazon Web Services, 13 April 2026", "type": "Institutional survey", "url": "https://plc.pearson.com/en-GB/news-and-insights/news/new-pearson-and-aws-global-research-53-employers-struggle-find-ai-ready", "finding": "More than 2,700 responses from learners, higher-education leaders and employers across six countries, framing readiness as technical fluency combined with problem solving, communication, collaboration, adaptability and human-AI judgement.", "evidenceStrength": "Institutional survey, multi-country, commissioned by organisations that sell the remedy.", "superSkillsRead": "The most directly relevant source on the education-to-work transition, which is where the missing rungs problem begins. Its definition of readiness is notable: judgement and adaptability sit alongside technical fluency rather than behind it.", "relatedPages": [ "/research/will-ai-replace-entry-level-jobs", "/research/how-humans-learn-with-ai" ] }, { "id": "pearson-cognizant-workforce-pulse-2026", "citeAs": "https://thesuperskills.com/research/essential-works#pearson-cognizant-workforce-pulse-2026", "classification": "strong", "themes": [ "leadership", "skills", "organisation design" ], "authors": "Pearson and Cognizant", "year": 2026, "title": "The AI Workforce Pulse: The Adaptability Imperative", "publication": "Pearson and Cognizant, June 2026", "type": "Institutional survey", "url": "https://www.pearson.com/power-of-learning/ai-workforce-pulse.html", "finding": "A survey of 750 HR leaders across the US, UK and India, fielded March to April 2026 among directors and above at organisations of 1,000 or more employees.", "evidenceStrength": "Institutional survey, decent sample of senior HR decision-makers, commissioned by vendors of the remedy.", "superSkillsRead": "Useful for what HR leadership currently believes about middle managers, early careers and the durability of human skills. Belief is not evidence, but in workforce strategy what leaders believe determines what gets funded.", "relatedPages": [ "/research/chro-guide-to-ai" ] }, { "id": "wef-human-advantage-2026", "citeAs": "https://thesuperskills.com/research/essential-works#wef-human-advantage-2026", "classification": "strong", "themes": [ "agency", "skills", "learning" ], "authors": "World Economic Forum with the McKinsey Health Institute", "year": 2026, "title": "The Human Advantage: Stronger Brains in the Age of AI", "publication": "World Economic Forum, 15 January 2026", "type": "Institutional report", "url": "https://www.weforum.org/publications/the-human-advantage-stronger-brains-in-the-age-of-ai/", "finding": "Frames human cognitive capability, or brain capital, as an economic asset to be actively maintained rather than assumed.", "evidenceStrength": "Institutional report, launched at Davos, co-produced with a consultancy health institute. Framing and synthesis rather than new primary evidence.", "superSkillsRead": "The clearest sign that the argument on this site has reached the mainstream institutional agenda. When the WEF publishes on maintaining human cognitive capability rather than on reskilling for new tools, the question has shifted from what people should learn to whether they retain the capacity to learn at all.", "relatedPages": [ "/research/human-skills-in-the-age-of-ai", "/research/capability-debt" ] }, { "id": "humlum-vestergaard-2025", "citeAs": "https://thesuperskills.com/research/essential-works#humlum-vestergaard-2025", "classification": "strong", "themes": [ "jobs", "organisation design" ], "authors": "Humlum, A. and Vestergaard, E.", "year": 2025, "title": "Still Waters, Rapid Currents: Early Labor Market Transformation under Generative AI", "publication": "NBER Working Paper 33777, revised March 2026", "type": "Working paper", "url": "https://www.nber.org/papers/w33777", "finding": "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.", "evidenceStrength": "Working paper, not peer-reviewed. That the same holds elsewhere. Denmark is high-trust, high-wage and heavily unionised, and two years is early.", "superSkillsRead": "The best available answer to what has actually happened rather than what might. Its real lesson is that work reorganises long before earnings move, so pay data is the slowest signal a leader can watch.", "relatedPages": [ "/research/ai-workforce-strategy", "/research/will-ai-replace-my-job" ], "gradedEntry": "https://thesuperskills.com/research/evidence#humlum-vestergaard-2025" }, { "id": "gerlich-2025", "citeAs": "https://thesuperskills.com/research/essential-works#gerlich-2025", "classification": "strong", "themes": [ "critical thinking", "deskilling" ], "authors": "Gerlich, M.", "year": 2025, "title": "AI Tools in Society: Impacts on Cognitive Offloading and the Future of Critical Thinking", "publication": "Societies, 15(1), 6", "type": "Journal article", "url": "https://www.mdpi.com/2075-4698/15/1/6", "finding": "A negative correlation between frequent AI use and critical-thinking scores, mediated by cognitive offloading, strongest among the youngest users.", "evidenceStrength": "Peer-reviewed study. Causation, and it carries a published correction (Societies 2025, 15(9), 252) which anyone citing it should read alongside.", "superSkillsRead": "Widely cited, and worth citing carefully. It shows correlation with a plausible mechanism, and it carries a published correction that most people quoting it have not read.", "relatedPages": [ "/research/ai-and-critical-thinking" ], "gradedEntry": "https://thesuperskills.com/research/evidence#gerlich-2025" }, { "id": "deloitte-hct-2025", "citeAs": "https://thesuperskills.com/research/essential-works#deloitte-hct-2025", "classification": "strong", "themes": [ "organisation design", "leadership" ], "authors": "Deloitte", "year": 2025, "title": "2025 Global Human Capital Trends", "publication": "Deloitte Insights", "type": "Institutional survey", "url": "https://www.deloitte.com/us/en/insights/topics/talent/human-capital-trends/2025.html", "finding": "Frames the worker-organisation relationship as a set of unresolved tensions rather than a set of solved problems.", "evidenceStrength": "Institutional survey. Causal claims. It is a sentiment survey by a firm that sells the remedies it recommends, and should be read with that in view.", "superSkillsRead": "Large-scale practitioner sentiment, published by a firm that sells the remedies it recommends. Both facts matter.", "relatedPages": [ "/research/ai-workforce-strategy", "/research/chro-guide-to-ai" ], "gradedEntry": "https://thesuperskills.com/research/evidence#deloitte-hct-2025" }, { "id": "microsoft-wti-2025", "citeAs": "https://thesuperskills.com/research/essential-works#microsoft-wti-2025", "classification": "strong", "themes": [ "organisation design", "agency" ], "authors": "Microsoft and LinkedIn", "year": 2025, "title": "2025 Work Trend Index Annual Report: The Year the Frontier Firm is Born", "publication": "Microsoft WorkLab, April 2025", "type": "Institutional survey", "url": "https://www.microsoft.com/en-us/worklab/work-trend-index/2025-the-year-the-frontier-firm-is-born", "finding": "Describes the emergence of firms organised around human-agent teams and a shift towards workers managing AI agents.", "evidenceStrength": "Institutional survey. Independence. Microsoft sells the tools whose adoption it is measuring, and telemetry measures usage rather than value.", "superSkillsRead": "Kept as the longitudinal predecessor to the 2026 edition, which moved much closer to the human-agency question.", "relatedPages": [ "/research/ai-agents-and-human-judgement", "/research/ai-workforce-strategy" ], "gradedEntry": "https://thesuperskills.com/research/evidence#microsoft-wti-2025" }, { "id": "pwc-jobs-barometer-2025", "citeAs": "https://thesuperskills.com/research/essential-works#pwc-jobs-barometer-2025", "classification": "strong", "themes": [ "jobs", "early careers" ], "authors": "PwC", "year": 2025, "title": "PwC Global AI Jobs Barometer 2025", "publication": "PwC, 2025", "type": "Institutional analysis", "url": "https://www.pwc.com/gx/en/issues/artificial-intelligence/job-barometer/2025/report.pdf", "finding": "Reports wage premiums for AI skills and shifting skill requirements in AI-exposed occupations.", "evidenceStrength": "Institutional modelling. Causation, and job advertisements describe what employers ask for rather than what the work requires.", "superSkillsRead": "Kept as the longitudinal predecessor to the 2026 edition, whose entry-level finding is the more striking one.", "relatedPages": [ "/research/will-ai-replace-entry-level-jobs" ], "gradedEntry": "https://thesuperskills.com/research/evidence#pwc-jobs-barometer-2025" }, { "id": "cui-2025", "citeAs": "https://thesuperskills.com/research/essential-works#cui-2025", "classification": "strong", "themes": [ "jobs", "expertise", "human-AI collaboration" ], "authors": "Cui, Z. K., Demirer, M., Jaffe, S., Musolff, L., Peng, S. and Salz, T.", "year": 2025, "title": "The Effects of Generative AI on High-Skilled Work: Evidence from Three Field Experiments with Software Developers", "publication": "Management Science, 2025", "type": "Field experiment", "url": "https://www.microsoft.com/en-us/research/publication/the-effects-of-generative-ai-on-high-skilled-work-evidence-from-three-field-experiments-with-software-developers/", "finding": "Across three randomised experiments and 4,867 developers at Microsoft, Accenture and a Fortune 100 company, completed tasks rose 26.08 percent, with larger gains among less experienced developers.", "evidenceStrength": "Randomised controlled trials at real firms, which is rare and valuable. The headline estimate carries a standard error of 10.3 percent, so the true effect is wide.", "superSkillsRead": "The strongest evidence that AI gains are real in high-skilled work, and it repeats the pattern found in customer support: the least experienced gain most. Quote the 26 percent with its standard error, or not at all.", "relatedPages": [ "/research/staying-valuable-in-the-age-of-ai" ] }, { "id": "dillon-2025", "citeAs": "https://thesuperskills.com/research/essential-works#dillon-2025", "classification": "strong", "themes": [ "organisation design", "jobs" ], "authors": "Dillon, E., Jaffe, S., Immorlica, N. and Stanton, C.", "year": 2025, "title": "Shifting Work Patterns with Generative AI", "publication": "Microsoft Research; NBER Working Paper 33795", "type": "Field experiment", "url": "https://www.microsoft.com/en-us/research/publication/shifting-work-patterns-with-generative-ai/", "finding": "A six-month randomised field experiment across roughly 6,000 knowledge workers found users spent about three fewer hours per week on email, with no significant change in time spent in meetings.", "evidenceStrength": "Randomised, six months, cross-industry, measuring behaviour rather than self-report. Among the best-designed studies in this literature.", "superSkillsRead": "Important because it measures what changed rather than what people said changed, and because the answer is modest and specific. Time moved out of email and did not move out of meetings, which is a useful corrective to claims of wholesale transformation.", "relatedPages": [ "/research/ai-workforce-strategy" ] }, { "id": "ilo-genai-jobs-2025", "citeAs": "https://thesuperskills.com/research/essential-works#ilo-genai-jobs-2025", "classification": "strong", "themes": [ "jobs", "skills" ], "authors": "International Labour Organization (with NASK)", "year": 2025, "title": "Generative AI and Jobs: A Refined Global Index of Occupational Exposure", "publication": "ILO Working Paper 140, 20 May 2025", "type": "Working paper", "url": "https://www.ilo.org/publications/generative-ai-and-jobs-refined-global-index-occupational-exposure", "finding": "A refined global index of occupational exposure to generative AI, updating the ILO's 2023 analysis with a more internationally balanced view of augmentation versus automation.", "evidenceStrength": "Institutional working paper with published methodology. Exposure modelling, so it estimates what could be affected rather than what will be.", "superSkillsRead": "The best correction to the US-centric bias running through most of this literature. Exposure looks very different in economies with different occupational structures, and the ILO's finding that augmentation dominates automation in most contexts is the more defensible reading of the same underlying data.", "relatedPages": [ "/research/will-ai-replace-my-job" ] }, { "id": "oecd-ai-capability-indicators-2025", "citeAs": "https://thesuperskills.com/research/essential-works#oecd-ai-capability-indicators-2025", "classification": "strong", "themes": [ "technology capability", "skills" ], "authors": "OECD", "year": 2025, "title": "OECD AI Capability Indicators: Technical Report", "publication": "OECD Publishing, Paris, November 2025", "type": "Institutional analysis", "url": "https://www.oecd.org/en/publications/oecd-ai-capability-indicators-technical-report_9cdb3dd1-en.html", "finding": "Builds nine indicator scales that rate AI capability against human ability levels, so machine and human performance can be compared on the same dimensions.", "evidenceStrength": "Institutional methodology report with expert elicitation. New and not yet widely validated; the scales are judgement-based rather than benchmark-derived.", "superSkillsRead": "The most serious attempt anywhere to answer what AI can do relative to a person, on comparable scales, rather than through benchmark scores that mean nothing to a manager. If it matures, it becomes the instrument this whole field has lacked.", "relatedPages": [ "/research/what-stays-human", "/research/human-ai-decision-making" ] }, { "id": "doshi-hauser-2024", "citeAs": "https://thesuperskills.com/research/essential-works#doshi-hauser-2024", "classification": "strong", "themes": [ "agency", "organisation design" ], "authors": "Doshi, A. R. and Hauser, O. P.", "year": 2024, "title": "Generative AI enhances individual creativity but reduces the collective diversity of novel content", "publication": "Science Advances, 10(28)", "type": "Journal article", "url": "https://discovery.ucl.ac.uk/id/eprint/10195027/", "finding": "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.", "evidenceStrength": "Peer-reviewed study. That this generalises beyond one short creative task with one form of assistance.", "superSkillsRead": "Individual creative quality and collective creative range move in opposite directions. A social dilemma in which everyone is individually right and the result is duller.", "relatedPages": [ "/research/what-stays-human" ], "gradedEntry": "https://thesuperskills.com/research/evidence#doshi-hauser-2024" }, { "id": "bick-2024", "citeAs": "https://thesuperskills.com/research/essential-works#bick-2024", "classification": "strong", "themes": [ "jobs", "organisation design" ], "authors": "Bick, A., Blandin, A. and Deming, D. J.", "year": 2024, "title": "The Rapid Adoption of Generative AI", "publication": "NBER Working Paper 32966", "type": "Working paper", "url": "https://www.nber.org/papers/w32966", "finding": "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.", "evidenceStrength": "Working paper, not peer-reviewed. Quality of use. Self-reported use counts any use at all.", "superSkillsRead": "Enormous reach, thin penetration. The gap between how many people use AI and how few hours it touches is the whole of the adoption-versus-redesign argument.", "relatedPages": [ "/research/ai-workforce-strategy", "/research/why-learn-to-prompt-is-weak-career-advice" ], "gradedEntry": "https://thesuperskills.com/research/evidence#bick-2024" }, { "id": "mckinsey-new-future-of-work-2024", "citeAs": "https://thesuperskills.com/research/essential-works#mckinsey-new-future-of-work-2024", "classification": "strong", "themes": [ "jobs", "skills" ], "authors": "McKinsey Global Institute", "year": 2024, "title": "A new future of work: The race to deploy AI and raise skills in Europe and beyond", "publication": "McKinsey Global Institute, May 2024", "type": "Institutional analysis", "url": "https://www.mckinsey.com/mgi/our-research/a-new-future-of-work-the-race-to-deploy-ai-and-raise-skills-in-europe-and-beyond", "finding": "Projects large-scale occupational transitions and rising demand for social, emotional and higher cognitive skills.", "evidenceStrength": "Institutional modelling. What will happen. Scenario models are assumption-driven, and McKinsey's prior transition estimates have moved substantially between editions.", "superSkillsRead": "A transparent scenario model, useful for direction. Its prior transition estimates have moved substantially between editions.", "relatedPages": [ "/research/ai-workforce-strategy", "/research/will-ai-replace-my-job" ], "gradedEntry": "https://thesuperskills.com/research/evidence#mckinsey-new-future-of-work-2024" }, { "id": "imf-genai-work-2024", "citeAs": "https://thesuperskills.com/research/essential-works#imf-genai-work-2024", "classification": "strong", "themes": [ "jobs" ], "authors": "Cazzaniga, M. et al. (International Monetary Fund)", "year": 2024, "title": "Gen-AI: Artificial Intelligence and the Future of Work", "publication": "IMF Staff Discussion Note 2024/001, 14 January 2024", "type": "Institutional analysis", "url": "https://www.imf.org/en/publications/staff-discussion-notes/issues/2024/01/14/gen-ai-artificial-intelligence-and-the-future-of-work-542379", "finding": "Estimates AI exposure across advanced, emerging and low-income economies, and analyses the implications for inequality within and between countries.", "evidenceStrength": "Institutional macro analysis, not peer-reviewed. Built on occupational exposure measures with the usual limits.", "superSkillsRead": "The macroeconomic counterweight to company surveys, and the only major source here that takes inequality between countries seriously. Its uncomfortable finding is that advanced economies face more exposure and are better placed to benefit, which widens rather than narrows global gaps.", "relatedPages": [ "/research/will-ai-replace-my-job" ] }, { "id": "daugherty-wilson-human-machine", "citeAs": "https://thesuperskills.com/research/essential-works#daugherty-wilson-human-machine", "classification": "strong", "themes": [ "human-AI collaboration", "organisation design" ], "authors": "Daugherty, P. R. and Wilson, H. J.", "year": 2024, "title": "Human + Machine: Reimagining Work in the Age of AI", "publication": "Harvard Business Review Press, updated and expanded edition", "type": "Book", "url": "https://store.hbr.org/product/human-machine-updated-and-expanded-reimagining-work-in-the-age-of-ai/10724", "finding": "Sets out the missing middle, where humans and machines collaborate rather than substitute, and names hybrid roles including trainer, explainer and sustainer.", "evidenceStrength": "Important perspective from consultancy practice, with case material rather than controlled evidence.", "superSkillsRead": "The most usable role taxonomy for human-AI work, and it predates the generative wave by six years. Cite the 2024 expanded edition, which adds generative AI; the 2018 original is often quoted for claims it does not make.", "relatedPages": [ "/research/human-ai-decision-making", "/research/ai-workforce-strategy" ] }, { "id": "unesco-ai-competency-teachers-2024", "citeAs": "https://thesuperskills.com/research/essential-works#unesco-ai-competency-teachers-2024", "classification": "strong", "themes": [ "frameworks", "education" ], "authors": "Miao, F. and Cukurova, M. (UNESCO)", "year": 2024, "title": "AI competency framework for teachers", "publication": "UNESCO, Paris", "type": "Competency framework", "url": "https://unesdoc.unesco.org/ark:/48223/pf0000391104", "finding": "Defines what teachers need to know and be able to do with AI, across a human-centred mindset, ethics, foundations, pedagogy and professional learning.", "evidenceStrength": "Official UNESCO framework, expert-developed and internationally consulted. Normative guidance rather than evidence.", "superSkillsRead": "Notable for leading with a human-centred mindset and ethics rather than tool skills, which is the opposite ordering from most corporate AI-literacy training. A useful model for what capability-first rather than tool-first development looks like in practice.", "relatedPages": [ "/research/how-humans-learn-with-ai", "/research/chro-guide-to-ai" ] }, { "id": "ayers-2023", "citeAs": "https://thesuperskills.com/research/essential-works#ayers-2023", "classification": "strong", "themes": [ "agency", "ethics" ], "authors": "Ayers, J. W. et al.", "year": 2023, "title": "Comparing Physician and Artificial Intelligence Chatbot Responses to Patient Questions Posted to a Public Social Media Forum", "publication": "JAMA Internal Medicine, 183(6), 589-596", "type": "Journal article", "url": "https://pure.johnshopkins.edu/en/publications/comparing-physician-and-artificial-intelligence-chatbot-responses/", "finding": "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.", "evidenceStrength": "Peer-reviewed study. That a machine can care for anyone. Doctors answering strangers free of charge between patients are not doing the job they trained for.", "superSkillsRead": "The study that should end the claim that empathy is safe from AI. On the observable performance of empathy in text, the machine already wins comfortably.", "relatedPages": [ "/research/what-stays-human", "/research/outsourced-recognition" ], "gradedEntry": "https://thesuperskills.com/research/evidence#ayers-2023" }, { "id": "zamfirescu-pereira-2023", "citeAs": "https://thesuperskills.com/research/essential-works#zamfirescu-pereira-2023", "classification": "strong", "themes": [ "skills", "human-AI collaboration" ], "authors": "Zamfirescu-Pereira, J. D., Wong, R. Y., Hartmann, B. and Yang, Q.", "year": 2023, "title": "Why Johnny Can't Prompt: How Non-AI Experts Try (and Fail) to Design LLM Prompts", "publication": "CHI 2023", "type": "Journal article", "url": "https://dl.acm.org/doi/10.1145/3544548.3581388", "finding": "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.", "evidenceStrength": "Peer-reviewed study. That this persists. It used 2023 models, and providers are actively engineering the difficulty away.", "superSkillsRead": "The strongest case for teaching prompting, included on a site that argues prompting is weak career advice. It shows the difficulty is real; it does not show the difficulty is durable.", "relatedPages": [ "/research/why-learn-to-prompt-is-weak-career-advice" ], "gradedEntry": "https://thesuperskills.com/research/evidence#zamfirescu-pereira-2023" }, { "id": "wef-foj-2023", "citeAs": "https://thesuperskills.com/research/essential-works#wef-foj-2023", "classification": "strong", "themes": [ "skills", "jobs" ], "authors": "World Economic Forum", "year": 2023, "title": "The Future of Jobs Report 2023", "publication": "World Economic Forum, Geneva, April 2023", "type": "Institutional survey", "url": "https://www.weforum.org/publications/the-future-of-jobs-report-2023/", "finding": "Analytical thinking leads, with creative thinking second, and a growing emphasis on self-efficacy skills.", "evidenceStrength": "Institutional survey. Considered judgement on generative AI. It was fielded too early for that.", "superSkillsRead": "The first post-ChatGPT edition, fielded five months after launch and therefore too early for considered judgement on it.", "relatedPages": [ "/research/human-skills-in-the-age-of-ai" ], "gradedEntry": "https://thesuperskills.com/research/evidence#wef-foj-2023" }, { "id": "oecd-skills-outlook-2023", "citeAs": "https://thesuperskills.com/research/essential-works#oecd-skills-outlook-2023", "classification": "strong", "themes": [ "skills", "education" ], "authors": "OECD", "year": 2023, "title": "OECD Skills Outlook 2023: Skills for a Resilient Green and Digital Transition", "publication": "OECD Publishing, Paris, November 2023", "type": "Institutional analysis", "url": "https://www.oecd.org/en/publications/oecd-skills-outlook-2023_27452f29-en.html", "finding": "Analyses the skills required for green and digital transitions across member economies.", "evidenceStrength": "Institutional modelling. Anything AI-specific and current. The underlying data collection predates the generative-AI period.", "superSkillsRead": "A methodologically transparent cross-national baseline, from data collected before the generative-AI period.", "relatedPages": [ "/research/human-skills-in-the-age-of-ai" ], "gradedEntry": "https://thesuperskills.com/research/evidence#oecd-skills-outlook-2023" }, { "id": "nfer-skills-imperative-2035", "citeAs": "https://thesuperskills.com/research/essential-works#nfer-skills-imperative-2035", "classification": "strong", "themes": [ "skills", "education" ], "authors": "NFER", "year": 2023, "title": "The Skills Imperative 2035: An analysis of the demand for skills in the labour market in 2035 (Working Paper 3)", "publication": "National Foundation for Educational Research, with the University of Sheffield, funded by the Nuffield Foundation, May 2023", "type": "Institutional analysis", "url": "https://www.nfer.ac.uk/publications/the-skills-imperative-2035-an-analysis-of-the-demand-for-skills-in-the-labour-market-in-2035/", "finding": "Projects rising demand for a set of essential employment skills in the UK to 2035.", "evidenceStrength": "Institutional modelling. Precision. It maps US skill data onto UK occupations, and a revised working paper corrects coding errors in the underlying labour force survey.", "superSkillsRead": "A rare UK-specific, independently funded projection. Maps US skill data onto UK occupations, which is its main limitation.", "relatedPages": [ "/research/human-skills-in-the-age-of-ai" ], "gradedEntry": "https://thesuperskills.com/research/evidence#nfer-skills-imperative-2035" }, { "id": "nist-ai-rmf", "citeAs": "https://thesuperskills.com/research/essential-works#nist-ai-rmf", "classification": "strong", "themes": [ "governance", "organisation design", "ethics" ], "authors": "National Institute of Standards and Technology", "year": 2023, "title": "Artificial Intelligence Risk Management Framework (AI RMF 1.0)", "publication": "NIST AI 100-1, 26 January 2023; Generative AI Profile, NIST AI 600-1, July 2024", "type": "Governance framework", "url": "https://www.nist.gov/itl/ai-risk-management-framework", "finding": "A voluntary framework organised around governing, mapping, measuring and managing AI risk, with human oversight and accountability as explicit functions.", "evidenceStrength": "Official framework, widely adopted as a de facto standard. Voluntary and process-oriented; it describes what to do rather than evidencing what works. Currently under revision.", "superSkillsRead": "The practical vocabulary for the accountability argument, and the document a board will already recognise. Its weakness is instructive: it can be satisfied procedurally, by an organisation that documents oversight without exercising it, the failure mode this research is concerned with.", "relatedPages": [ "/research/human-ai-decision-making", "/research/chro-guide-to-ai" ] }, { "id": "ilo-genai-jobs-2023", "citeAs": "https://thesuperskills.com/research/essential-works#ilo-genai-jobs-2023", "classification": "strong", "themes": [ "jobs", "skills" ], "authors": "Gmyrek, P., Berg, J. and Bescond, D. (International Labour Organization)", "year": 2023, "title": "Generative AI and Jobs: A Global Analysis of Potential Effects on Job Quantity and Quality", "publication": "ILO Working Paper 96, 21 August 2023", "type": "Working paper", "url": "https://www.ilo.org/publications/generative-ai-and-jobs-global-analysis-potential-effects-job-quantity-and", "finding": "Finds that most jobs are more likely to be augmented than automated, with clerical work the most exposed category and a markedly larger effect for women in higher-income countries.", "evidenceStrength": "Institutional working paper with published methodology. Exposure analysis rather than measured outcome; superseded in part by the 2025 refined index.", "superSkillsRead": "The first major analysis to conclude that augmentation dominates automation, and still among the most balanced. Its finding on clerical work and gendered exposure is one of the few in this literature that engages seriously with who bears the cost.", "relatedPages": [ "/research/will-ai-replace-my-job" ] }, { "id": "pearson-power-skills-2022", "citeAs": "https://thesuperskills.com/research/essential-works#pearson-power-skills-2022", "classification": "strong", "themes": [ "skills", "jobs" ], "authors": "Pearson", "year": 2022, "title": "Pearson Skills Outlook: Power Skills", "publication": "Pearson, November 2022", "type": "Labour-market analysis", "url": "https://plc.pearson.com/en-GB/news-and-insights/news/new-pearson-study-identifies-human-skills-power-skills-most-demand-worlds", "finding": "Analysis of 21.8 million job advertisements across the US, Canada, the UK and Australia found human capabilities dominating employer demand.", "evidenceStrength": "Very large observed-demand dataset. Note the date: this is a 2022 study forecasting demand to 2026, not a 2026 publication.", "superSkillsRead": "Valuable as pre-ChatGPT evidence. It establishes that human capabilities were already the scarce thing in the labour market before generative AI arrived, which undercuts the claim that this is a fashion created by the current wave.", "relatedPages": [ "/research/human-skills-in-the-age-of-ai" ] }, { "id": "ec-digcomp-2-2", "citeAs": "https://thesuperskills.com/research/essential-works#ec-digcomp-2-2", "classification": "strong", "themes": [ "frameworks", "skills", "education" ], "authors": "Vuorikari, R., Kluzer, S. and Punie, Y. (European Commission JRC)", "year": 2022, "title": "DigComp 2.2: The Digital Competence Framework for Citizens", "publication": "Publications Office of the European Union, JRC128415", "type": "Competency framework", "url": "https://publications.jrc.ec.europa.eu/repository/handle/JRC128415", "finding": "The European reference framework for digital competence, updated in 2022 with examples covering AI and datafication.", "evidenceStrength": "Official framework, expert-developed and widely adopted in European policy. A normative taxonomy rather than empirical evidence.", "superSkillsRead": "One of the taxonomies any human-capability framework is implicitly competing with, and worth knowing for that reason. Its 2022 AI additions are thin on judgement and accountability, and that gap is the one the SuperSkills framework addresses.", "relatedPages": [ "/research/human-skills-in-the-age-of-ai" ] }, { "id": "mckinsey-deltas-2021", "citeAs": "https://thesuperskills.com/research/essential-works#mckinsey-deltas-2021", "classification": "strong", "themes": [ "skills" ], "authors": "McKinsey and Company", "year": 2021, "title": "Defining the skills citizens will need in the future world of work", "publication": "McKinsey Public and Social Sector Practice, June 2021", "type": "Institutional survey", "url": "https://www.mckinsey.com/industries/public-sector/our-insights/defining-the-skills-citizens-will-need-in-the-future-world-of-work", "finding": "Identifies distinct elements of talent, the DELTAs, associated with employment, income and job satisfaction.", "evidenceStrength": "Institutional survey. Anything about AI. It was fielded in 2019, before the generative-AI period entirely.", "superSkillsRead": "An unusually large individual-level dataset on skills and outcomes, fielded in 2019 and therefore silent on AI.", "relatedPages": [ "/research/human-skills-in-the-age-of-ai" ], "gradedEntry": "https://thesuperskills.com/research/evidence#mckinsey-deltas-2021" }, { "id": "wef-foj-2020", "citeAs": "https://thesuperskills.com/research/essential-works#wef-foj-2020", "classification": "strong", "themes": [ "skills", "jobs" ], "authors": "World Economic Forum", "year": 2020, "title": "The Future of Jobs Report 2020", "publication": "World Economic Forum, Geneva, October 2020", "type": "Institutional survey", "url": "https://www.weforum.org/publications/the-future-of-jobs-report-2020/", "finding": "Named critical thinking and problem solving as leading skills, and forecast large-scale reskilling need.", "evidenceStrength": "Institutional survey. A clean read on AI. The 2020 edition is dominated by COVID-era disruption.", "superSkillsRead": "The pandemic edition. Read for what the field expected in 2020, and discounted for the disruption dominating it.", "relatedPages": [ "/research/human-skills-in-the-age-of-ai" ], "gradedEntry": "https://thesuperskills.com/research/evidence#wef-foj-2020" }, { "id": "macnamara-maitra-2019", "citeAs": "https://thesuperskills.com/research/essential-works#macnamara-maitra-2019", "classification": "strong", "themes": [ "expertise", "learning" ], "authors": "Macnamara, B. N. and Maitra, M.", "year": 2019, "title": "The role of deliberate practice in expert performance: revisiting Ericsson, Krampe and Tesch-Romer (1993)", "publication": "Royal Society Open Science, 6, 190327", "type": "Journal article", "url": "https://royalsocietypublishing.org/doi/10.1098/rsos.190327", "finding": "Accumulated practice explained considerably less of the difference between performers than the original is usually taken to claim.", "evidenceStrength": "Peer-reviewed study. That practice does not matter. It does; the simple dose-response reading is what fails.", "superSkillsRead": "Included deliberately because it complicates the argument this research relies on. Practice matters; the simple dose-response reading of Ericsson does not survive.", "relatedPages": [ "/research/how-humans-learn-with-ai" ], "gradedEntry": "https://thesuperskills.com/research/evidence#macnamara-maitra-2019" }, { "id": "wef-foj-2018", "citeAs": "https://thesuperskills.com/research/essential-works#wef-foj-2018", "classification": "strong", "themes": [ "skills", "jobs" ], "authors": "World Economic Forum", "year": 2018, "title": "The Future of Jobs Report 2018", "publication": "World Economic Forum, Geneva, September 2018", "type": "Institutional survey", "url": "https://www.weforum.org/publications/the-future-of-jobs-report-2018/", "finding": "Set out expected skill demand to 2022, with analytical thinking and innovation, active learning and creativity leading the list.", "evidenceStrength": "Institutional survey. A representative picture of employers. Respondents are drawn from a self-selected membership network.", "superSkillsRead": "Kept deliberately as a longitudinal comparator. Its predictions for 2022 are now checkable, which is a rarer thing than another forecast.", "relatedPages": [ "/research/human-skills-in-the-age-of-ai" ], "gradedEntry": "https://thesuperskills.com/research/evidence#wef-foj-2018" }, { "id": "ec-entrecomp", "citeAs": "https://thesuperskills.com/research/essential-works#ec-entrecomp", "classification": "strong", "themes": [ "frameworks", "skills" ], "authors": "Bacigalupo, M., Kampylis, P., Punie, Y. and Van den Brande, G. (European Commission JRC)", "year": 2016, "title": "EntreComp: The Entrepreneurship Competence Framework", "publication": "Publications Office of the European Union, JRC101581", "type": "Competency framework", "url": "https://publications.jrc.ec.europa.eu/repository/handle/JRC101581", "finding": "A framework of fifteen competences across ideas and opportunities, resources, and action, describing initiative and value creation as learnable capabilities.", "evidenceStrength": "Official framework, expert-developed. Normative rather than empirical.", "superSkillsRead": "The best existing treatment of initiative and opportunity recognition as trainable rather than innate. Useful as prior art for anyone arguing that human capabilities can be developed systematically rather than merely possessed.", "relatedPages": [ "/research/human-skills-in-the-age-of-ai" ] }, { "id": "susskind-future-professions", "citeAs": "https://thesuperskills.com/research/essential-works#susskind-future-professions", "classification": "strong", "themes": [ "expertise", "jobs", "organisation design" ], "authors": "Susskind, R. and Susskind, D.", "year": 2015, "title": "The Future of the Professions: How Technology Will Transform the Work of Human Experts", "publication": "Oxford University Press, updated edition", "type": "Book", "url": "https://global.oup.com/academic/product/the-future-of-the-professions-9780198841890", "finding": "Argues that the professions as currently organised are a solution to the problem of unevenly distributed expertise, and that technology dissolves the grand bargain that sustained them.", "evidenceStrength": "Important perspective with substantial case material across law, medicine, accountancy, education and consulting.", "superSkillsRead": "The most rigorous treatment of what happens to professional expertise specifically, written a decade ago and increasingly accurate. Read it alongside Polanyi: the Susskinds argue much professional work can be decomposed and routinised, and Polanyi explains what resists that.", "relatedPages": [ "/research/staying-valuable-in-the-age-of-ai", "/research/what-stays-human" ] }, { "id": "pearson-learning-gap-2026", "citeAs": "https://thesuperskills.com/research/essential-works#pearson-learning-gap-2026", "classification": "perspective", "themes": [ "learning", "organisation design" ], "authors": "Pearson", "year": 2026, "title": "Mind the Learning Gap: How We Can Achieve AI's Full Potential", "publication": "Pearson, January 2026", "type": "Institutional report", "url": "https://www.pearson.com/en-us/power-of-learning/ai-learning-gap.html", "finding": "Argues that AI productivity depends on learning and augmentation rather than deployment alone, estimating that augmentation could add 4.8 to 6.6 trillion dollars to the US economy by 2034.", "evidenceStrength": "Economic modelling by a learning company arguing for the importance of learning. The direction is well argued; the figure is an estimate with wide bounds.", "superSkillsRead": "Included as a serious argument rather than as evidence. Its core proposition, that the return on AI depends on whether people learn alongside it, is the constructive form of the capability-debt argument.", "relatedPages": [ "/research/ai-workforce-strategy" ] }, { "id": "kosmyna-2025", "citeAs": "https://thesuperskills.com/research/essential-works#kosmyna-2025", "classification": "perspective", "themes": [ "critical thinking" ], "authors": "Kosmyna, N. et al.", "year": 2025, "title": "Your Brain on ChatGPT: Accumulation of Cognitive Debt when Using an AI Assistant for Essay Writing Task", "publication": "MIT Media Lab preprint, arXiv:2506.08872", "type": "Working paper", "url": "https://arxiv.org/abs/2506.08872", "finding": "The LLM group showed the weakest brain connectivity and the lowest sense of ownership over their own writing.", "evidenceStrength": "Working paper, not peer-reviewed. Anything settled. 54 participants, a preprint, and reproducibility flagged by its own commentators. Treat claims of proof with suspicion.", "superSkillsRead": "Included because it is quoted constantly and rests on 54 participants in a preprint. A useful test of whether someone has read what they are citing.", "relatedPages": [ "/research/ai-and-critical-thinking" ], "gradedEntry": "https://thesuperskills.com/research/evidence#kosmyna-2025" }, { "id": "unicef-ai-children-2025", "citeAs": "https://thesuperskills.com/research/essential-works#unicef-ai-children-2025", "classification": "perspective", "themes": [ "education", "ethics" ], "authors": "UNICEF Innocenti", "year": 2025, "title": "Guidance on AI and Children, Version 3.0: Recommendations for AI policies and systems that uphold child rights", "publication": "UNICEF Innocenti, December 2025", "type": "Institutional analysis", "url": "https://www.unicef.org/innocenti/reports/policy-guidance-ai-children", "finding": "Sets out ten requirements and 48 recommendations for AI systems and policy affecting children.", "evidenceStrength": "Institutional modelling. Empirical claims about learning or capability effects. It is a normative guidance document.", "superSkillsRead": "A rights-based standard developed with children rather than about them, which almost nothing else in this space does.", "relatedPages": [ "/research/how-humans-learn-with-ai" ], "gradedEntry": "https://thesuperskills.com/research/evidence#unicef-ai-children-2025" }, { "id": "acemoglu-johnson-power-progress", "citeAs": "https://thesuperskills.com/research/essential-works#acemoglu-johnson-power-progress", "classification": "perspective", "themes": [ "ethics", "jobs", "organisation design" ], "authors": "Acemoglu, D. and Johnson, S.", "year": 2023, "title": "Power and Progress: Our Thousand-Year Struggle Over Technology and Prosperity", "publication": "PublicAffairs", "type": "Book", "url": "https://www.hachettebookgroup.com/titles/daron-acemoglu/power-and-progress/9781541702554/?lens=publicaffairs", "finding": "A thousand years of evidence that technology's gains are distributed according to choices about direction and power, not automatically by the technology itself.", "evidenceStrength": "Important perspective, historically grounded, contested by other economists.", "superSkillsRead": "The essential dissent. It challenges the assumption underneath most workforce writing, including some of this research: that if capability is protected, the benefits follow. Acemoglu and Johnson argue that who benefits is decided politically, and that is a claim this site does not answer.", "relatedPages": [ "/research/how-should-leaders-respond-to-ai" ] }, { "id": "frey-technology-trap", "citeAs": "https://thesuperskills.com/research/essential-works#frey-technology-trap", "classification": "perspective", "themes": [ "jobs", "ethics" ], "authors": "Frey, C. B.", "year": 2019, "title": "The Technology Trap: Capital, Labor, and Power in the Age of Automation", "publication": "Princeton University Press", "type": "Book", "url": "https://press.princeton.edu/books/hardcover/9780691172798/the-technology-trap", "finding": "A history of automation showing that periods of technological progress have frequently produced decades of falling wages and political backlash before broad gains arrived.", "evidenceStrength": "Important perspective, historically grounded. Frey is also co-author of the widely quoted and widely criticised 2013 estimate that 47 percent of US jobs were at risk.", "superSkillsRead": "The long view, and a warning against arguing from eventual outcomes. Even where technology ultimately raised living standards, the transition punished a generation. Read it as the reason the design choices this research argues for cannot be left to work themselves out.", "relatedPages": [ "/research/how-should-leaders-respond-to-ai" ] }, { "id": "dahmani-bohbot-2020", "citeAs": "https://thesuperskills.com/research/essential-works#dahmani-bohbot-2020", "classification": "foundation", "themes": [ "deskilling", "expertise" ], "authors": "Dahmani, L. and Bohbot, V. D.", "year": 2020, "title": "Habitual use of GPS negatively impacts spatial memory during self-guided navigation", "publication": "Scientific Reports, 10, 6310", "type": "Journal article", "url": "https://www.nature.com/articles/s41598-020-62877-0", "finding": "Habitual satnav users had worse spatial memory when navigating unaided, and heavier use over the following three years was associated with steeper decline.", "evidenceStrength": "Peer-reviewed study. That the same holds for reasoning. Spatial memory is not judgement, and the design is correlational.", "superSkillsRead": "The closest long-run analogue we have for what habitual reliance does to a human capability. Spatial memory is not judgement, and the analogy should carry weight without carrying certainty.", "relatedPages": [ "/research/what-is-cognitive-offloading", "/research/using-ai-without-dependency" ], "gradedEntry": "https://thesuperskills.com/research/evidence#dahmani-bohbot-2020" }, { "id": "logg-2019", "citeAs": "https://thesuperskills.com/research/essential-works#logg-2019", "classification": "foundation", "themes": [ "judgement", "human-AI collaboration" ], "authors": "Logg, J. M., Minson, J. A. and Moore, D. A.", "year": 2019, "title": "Algorithm Appreciation: People Prefer Algorithmic to Human Judgment", "publication": "Organizational Behavior and Human Decision Processes, 151, 90-103", "type": "Journal article", "url": "https://www.jennlogg.com/uploads/2/8/9/2/2892148/algorithm_appreciation__logg_minson_moore_2019_.pdf", "finding": "People often weight algorithmic advice MORE heavily than human advice. Domain experts are the notable exception.", "evidenceStrength": "Peer-reviewed study. Which tendency dominates in any given workplace.", "superSkillsRead": "Algorithm appreciation, the mirror image of aversion. Together they show miscalibration runs in both directions, with domain experts the notable exception.", "relatedPages": [ "/research/human-ai-decision-making" ], "gradedEntry": "https://thesuperskills.com/research/evidence#logg-2019" }, { "id": "risko-gilbert-2016", "citeAs": "https://thesuperskills.com/research/essential-works#risko-gilbert-2016", "classification": "foundation", "themes": [ "judgement", "critical thinking" ], "authors": "Risko, E. F. and Gilbert, S. J.", "year": 2016, "title": "Cognitive Offloading", "publication": "Trends in Cognitive Sciences, 20(9)", "type": "Journal article", "url": "https://www.cell.com/trends/cognitive-sciences/abstract/S1364-6613(16)30098-5", "finding": "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.", "evidenceStrength": "Peer-reviewed study. Nothing about generative AI specifically; it predates it.", "superSkillsRead": "The mechanism underneath everything on this site. Its most useful finding is that we offload when we judge a task to be hard, and that judgement is frequently wrong.", "relatedPages": [ "/research/what-is-cognitive-offloading", "/research/ai-and-human-judgement" ], "gradedEntry": "https://thesuperskills.com/research/evidence#risko-gilbert-2016" }, { "id": "dietvorst-2015", "citeAs": "https://thesuperskills.com/research/essential-works#dietvorst-2015", "classification": "foundation", "themes": [ "judgement", "human-AI collaboration" ], "authors": "Dietvorst, B. J., Simmons, J. P. and Massey, C.", "year": 2015, "title": "Algorithm Aversion: People Erroneously Avoid Algorithms After Seeing Them Err", "publication": "Journal of Experimental Psychology: General, 144(1)", "type": "Journal article", "url": "https://marketing.wharton.upenn.edu/wp-content/uploads/2016/10/Dietvorst-Simmons-Massey-2014.pdf", "finding": "After seeing an algorithm err, people abandon it even when it demonstrably outperforms them.", "evidenceStrength": "Peer-reviewed study. That this holds for conversational AI, which is far more recent and feels different to use.", "superSkillsRead": "Algorithm aversion. Trust in a model moves for reasons unrelated to its accuracy, so calibration cannot be mandated.", "relatedPages": [ "/research/human-ai-decision-making" ], "gradedEntry": "https://thesuperskills.com/research/evidence#dietvorst-2015" }, { "id": "sparrow-2011", "citeAs": "https://thesuperskills.com/research/essential-works#sparrow-2011", "classification": "foundation", "themes": [ "critical thinking" ], "authors": "Sparrow, B., Liu, J. and Wegner, D. M.", "year": 2011, "title": "Google Effects on Memory: Cognitive Consequences of Having Information at Our Fingertips", "publication": "Science, 333(6043)", "type": "Journal article", "url": "https://www.science.org/doi/10.1126/science.1207745", "finding": "When people expect information to remain available, they remember where to find it rather than the thing itself.", "evidenceStrength": "Peer-reviewed study. That total memory capability declines, or that the trade is net negative.", "superSkillsRead": "The Google effect. The first widely known demonstration that expected availability changes what we bother to encode.", "relatedPages": [ "/research/what-is-cognitive-offloading" ], "gradedEntry": "https://thesuperskills.com/research/evidence#sparrow-2011" }, { "id": "bjork-desirable-difficulties", "citeAs": "https://thesuperskills.com/research/essential-works#bjork-desirable-difficulties", "classification": "foundation", "themes": [ "learning", "education" ], "authors": "Bjork, E. L. and Bjork, R. A.", "year": 2011, "title": "Making Things Hard on Yourself, But in a Good Way: Creating Desirable Difficulties to Enhance Learning", "publication": "In Psychology and the Real World, Worth Publishers", "type": "Journal article", "url": "https://bjorklab.psych.ucla.edu/wp-content/uploads/sites/13/2016/04/EBjork_RBjork_2011.pdf", "finding": "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.", "evidenceStrength": "Peer-reviewed study. That AI-assisted work is equivalent to a fluent study condition. That is an inference, and it is ours.", "superSkillsRead": "Explains why AI-assisted work feels like learning and often is not. Learners systematically mistake fluency for understanding, and AI is a fluency machine.", "relatedPages": [ "/research/how-humans-learn-with-ai" ], "gradedEntry": "https://thesuperskills.com/research/evidence#bjork-desirable-difficulties" }, { "id": "parasuraman-manzey-2010", "citeAs": "https://thesuperskills.com/research/essential-works#parasuraman-manzey-2010", "classification": "foundation", "themes": [ "judgement", "human-AI collaboration" ], "authors": "Parasuraman, R. and Manzey, D. H.", "year": 2010, "title": "Complacency and Bias in Human Use of Automation: An Attentional Integration", "publication": "Human Factors, 52(3)", "type": "Journal article", "url": "https://journals.sagepub.com/doi/10.1177/0018720810376055", "finding": "Automation bias and complacency appear in novices and experts alike, resist training, and worsen under workload.", "evidenceStrength": "Peer-reviewed study. The size of the effect for generative AI, which is far less predictable than the automation studied here.", "superSkillsRead": "Automation bias is not a novelty of the AI era. It appears in novices and experts alike, resists training, and worsens under load. Policy alone will not fix it.", "relatedPages": [ "/research/human-ai-decision-making", "/research/ai-and-human-judgement" ], "gradedEntry": "https://thesuperskills.com/research/evidence#parasuraman-manzey-2010" }, { "id": "dzindolet-2003", "citeAs": "https://thesuperskills.com/research/essential-works#dzindolet-2003", "classification": "foundation", "themes": [ "judgement", "human-AI collaboration" ], "authors": "Dzindolet, M. T., Peterson, S. A., Pomranky, R. A., Pierce, L. G. and Beck, H. P.", "year": 2003, "title": "The role of trust in automation reliance", "publication": "International Journal of Human-Computer Studies, 58(6), 697-718", "type": "Journal article", "url": "https://doi.org/10.1016/S1071-5819(03)00038-7", "finding": "Trust mediates reliance on automated aids, and explaining why an aid might err restores trust even when that restored trust is not warranted.", "evidenceStrength": "Foundational experimental work.", "superSkillsRead": "A warning about explainability. Telling people why a system might be wrong makes them trust it more, not less, which means transparency measures can increase over-reliance rather than reduce it.", "relatedPages": [ "/research/human-ai-decision-making", "/research/ai-agents-and-human-judgement" ] }, { "id": "parasuraman-sheridan-wickens-2000", "citeAs": "https://thesuperskills.com/research/essential-works#parasuraman-sheridan-wickens-2000", "classification": "foundation", "themes": [ "human-AI collaboration", "organisation design" ], "authors": "Parasuraman, R., Sheridan, T. B. and Wickens, C. D.", "year": 2000, "title": "A Model for Types and Levels of Human Interaction with Automation", "publication": "IEEE Transactions on Systems, Man and Cybernetics, Part A, 30(3), 286-297", "type": "Journal article", "url": "https://ieeexplore.ieee.org/document/844354", "finding": "Sets out a four-stage model of what can be automated and ten levels of how far, providing a framework for deciding the degree of automation rather than treating it as binary.", "evidenceStrength": "Foundational framework, standard reference in human factors.", "superSkillsRead": "The intellectual groundwork for any delegation-boundary decision. It is twenty-six years old and considerably more rigorous than most current thinking about what to hand to an agent.", "relatedPages": [ "/research/ai-agents-and-human-judgement", "/research/human-ai-decision-making" ] }, { "id": "skitka-1999", "citeAs": "https://thesuperskills.com/research/essential-works#skitka-1999", "classification": "foundation", "themes": [ "judgement" ], "authors": "Skitka, L. J., Mosier, K. L. and Burdick, M.", "year": 1999, "title": "Does automation bias decision-making?", "publication": "International Journal of Human-Computer Studies, 51(5), 991-1006", "type": "Journal article", "url": "https://doi.org/10.1006/ijhc.1999.0252", "finding": "Demonstrates automation bias experimentally and separates it into errors of omission, where a person misses what the system did not flag, and errors of commission, where they act on a wrong recommendation.", "evidenceStrength": "Foundational experimental work.", "superSkillsRead": "The omission and commission distinction is the practical one. Most AI oversight is designed to catch commission errors, and omission errors are the ones nobody sees, because nothing appears on the screen to check.", "relatedPages": [ "/research/human-ai-decision-making" ] }, { "id": "klein-1998", "citeAs": "https://thesuperskills.com/research/essential-works#klein-1998", "classification": "foundation", "themes": [ "judgement", "expertise" ], "authors": "Klein, G.", "year": 1998, "title": "Sources of Power: How People Make Decisions", "publication": "MIT Press", "type": "Book", "url": "https://mitpress.mit.edu/9780262611466/sources-of-power/", "finding": "Experts in real conditions rarely compare options. They recognise a situation as typical and generate a workable course of action directly, a process Klein calls recognition-primed decision making.", "evidenceStrength": "Foundational naturalistic decision research, based on field studies of firefighters, nurses and military commanders.", "superSkillsRead": "The necessary counterweight to Kahneman. Where the heuristics literature treats intuition as a source of error, Klein shows it is the mechanism of expert performance under time pressure, and that it is built only through exposure to many real situations.", "relatedPages": [ "/research/ai-and-human-judgement", "/research/how-humans-learn-with-ai" ] }, { "id": "parasuraman-riley-1997", "citeAs": "https://thesuperskills.com/research/essential-works#parasuraman-riley-1997", "classification": "foundation", "themes": [ "judgement", "human-AI collaboration" ], "authors": "Parasuraman, R. and Riley, V.", "year": 1997, "title": "Humans and Automation: Use, Misuse, Disuse, Abuse", "publication": "Human Factors, 39(2), 230-253", "type": "Journal article", "url": "https://journals.sagepub.com/doi/10.1518/001872097778543886", "finding": "Establishes a taxonomy of how humans relate to automation: appropriate use, over-reliance, unwarranted rejection, and inappropriate deployment by designers.", "evidenceStrength": "Foundational review, widely replicated.", "superSkillsRead": "The vocabulary here is better than the vocabulary currently used about AI. Misuse and disuse are distinct failures with distinct remedies, and collapsing them into over-reliance loses the distinction that matters for design.", "relatedPages": [ "/research/human-ai-decision-making", "/research/design-versus-drift" ] }, { "id": "ericsson-1993", "citeAs": "https://thesuperskills.com/research/essential-works#ericsson-1993", "classification": "foundation", "themes": [ "expertise", "learning" ], "authors": "Ericsson, K. A., Krampe, R. T. and Tesch-Romer, C.", "year": 1993, "title": "The Role of Deliberate Practice in the Acquisition of Expert Performance", "publication": "Psychological Review, 100(3), 363-406", "type": "Journal article", "url": "https://eric.ed.gov/?id=EJ471947", "finding": "Sets out deliberate practice: effortful, targeted activity at the edge of current ability, with feedback, sustained over years.", "evidenceStrength": "Peer-reviewed study. How much of the difference between performers practice explains. See the Macnamara and Maitra re-examination below.", "superSkillsRead": "The theoretical basis for why removing repetitions matters. Read with the Macnamara re-examination, not instead of it.", "relatedPages": [ "/research/how-humans-learn-with-ai" ], "gradedEntry": "https://thesuperskills.com/research/evidence#ericsson-1993" }, { "id": "dreyfus-1986", "citeAs": "https://thesuperskills.com/research/essential-works#dreyfus-1986", "classification": "foundation", "themes": [ "expertise", "learning" ], "authors": "Dreyfus, H. L. and Dreyfus, S. E.", "year": 1986, "title": "Mind Over Machine: The Power of Human Intuition and Expertise in the Era of the Computer", "publication": "Free Press", "type": "Book", "url": "https://www.simonandschuster.com/books/Mind-Over-Machine/Hubert-Dreyfus/9780029080610", "finding": "Sets out a five-stage progression from novice to expert, in which rule-following gives way to situational, intuitive judgement built from accumulated experience.", "evidenceStrength": "Foundational theory, contested in parts, enormously influential.", "superSkillsRead": "The ladder itself. If expertise is a progression through stages, each requiring the experience of the one below, then removing the early stages does not accelerate the climb. It removes it.", "relatedPages": [ "/research/missing-rungs", "/research/synthetic-seniority" ] }, { "id": "anthropic-economic-index-2026", "citeAs": "https://thesuperskills.com/research/essential-works#anthropic-economic-index-2026", "classification": "watch", "themes": [ "jobs", "human-AI collaboration" ], "authors": "Anthropic", "year": 2026, "title": "Anthropic Economic Index: Cadences", "publication": "Anthropic, June 2026 (third in the 2026 series, after Economic primitives and Learning curves)", "type": "Living dataset", "url": "https://www.anthropic.com/research/economic-index-june-2026-report", "finding": "Analyses what people actually do with AI, by task and occupation, from real usage rather than survey response.", "evidenceStrength": "Observed behaviour at scale, which almost nothing else in this field has. Limited to one provider's users, who are not representative of workers generally.", "superSkillsRead": "Track this rather than cite it once. Nearly every other source here asks people what they think they do with AI; this observes what they actually do, which is a different and better question. The underlying data is published, so it can be analysed independently rather than only quoted.", "relatedPages": [ "/research/ai-workforce-strategy" ] }, { "id": "stanford-hai-index-2025", "citeAs": "https://thesuperskills.com/research/essential-works#stanford-hai-index-2025", "classification": "watch", "themes": [ "technology capability" ], "authors": "Stanford HAI", "year": 2025, "title": "The 2025 AI Index Report", "publication": "Stanford Institute for Human-Centered AI, April 2025", "type": "Compiled review", "url": "https://hai.stanford.edu/ai-index/2025-ai-index-report", "finding": "The most comprehensive annual account of AI capability, investment, adoption and public attitudes.", "evidenceStrength": "Compiled review. Much about human capability. It measures the machine side of the equation, which is precisely the gap this evidence base exists to fill.", "superSkillsRead": "Superseded by the 2026 edition, retained for comparison. Measures the machine, not the humans working alongside it.", "relatedPages": [ "/research/ai-workforce-strategy" ], "gradedEntry": "https://thesuperskills.com/research/evidence#stanford-hai-index-2025" } ] }