Essentials: inequality
Will AI widen inequality? The key papers, numbers, people and newsletters, sized to the time you have. The full question page. Picks are chosen for their weight in the debate and span its sides; they are refreshed every two weeks, last on 29 Oct 2026. The reasons are AI-written and evidence-cited; no rankings or scores.
Five papers, the numbers with their exact quotes, five people and three sources.
Papers to read
- Experimental evidence on the productivity effects of generative artificial intelligenceShakked Noy and Whitney Zhang · Science · 2023 · Evidence against the claim
ChatGPT helped weaker writers most, so the gap between workers narrowed in this experiment.
A preregistered experiment assigning 453 college-educated professionals writing tasks found that access to ChatGPT cut average completion time by 40% and raised output quality by 18%, with the gap between weaker and stronger writers narrowing. Two-thirds of exposed workers reported using ChatGPT in their actual jobs weeks later, making this one of the earliest rigorous demonstrations that generative AI both raises output and reduces within-occupation inequality for a common white-collar task.
- Generative AI at WorkErik Brynjolfsson et al. · The Quarterly Journal of Economics · 2025
Novice and lower-skilled support agents gained 34% per hour, the most experienced little; gains flowed to the less experienced.
Studying the staggered rollout of a generative AI assistant to 5,172 customer-support agents, finds it raised productivity (issues resolved per hour) by 15% on average, with the biggest gains for less experienced and lower-skilled workers, who saw both faster and higher-quality output; the most experienced agents saw little speed gain and slightly lower quality. Because the assistant appears to transfer top performers' know-how to novices, the paper is widely cited as evidence that generative AI can compress skill gaps within an occupation rather than simply replacing workers.
- The Uneven Impact of Generative AI on Entrepreneurial PerformanceNicholas G. Otis et al. · SSRN · 2024
A field experiment with entrepreneurs in Kenya: an AI mentor helped high performers and hurt low performers.
- Artificial intelligence and within-firm wage gaps: Evidence from China’s manufacturing sectorYanhui Yin et al. · International Labour Review · 2026 · Evidence for the claim
Firm-level evidence from Chinese manufacturing that AI adoption widens the wage gap between executives and ordinary staff.
Using firm-level text mining to measure AI adoption among Chinese manufacturers, the authors find AI transformation significantly widens the wage gap between executives and ordinary employees. They attribute this to skill-biased productivity gains and to a falling employment share of low-skilled routine workers, with stronger effects in capital-intensive and more digitised firms.
- Scenarios for the Transition to AGIAnton Korinek and Donghyun Suh · arXiv · 2024 · Mixed evidence
A model of when automation raises wages and when it lets them collapse, depending on how far it reaches.
In a task-based model where AI automates ever more complex tasks, wages depend on a race between automation and capital accumulation. If task complexity has an unbounded tail, wages can rise indefinitely; if full automation is reached, wages collapse, and they may fall earlier if automation outpaces capital accumulation.
Of these 3 papers: 1 with evidence against, 2 not coded on a side.
Of these 5 papers: 1 with evidence for the claim, 1 with evidence against, 1 mixed, 2 not coded on a side.
Of these 5 papers: 1 with evidence for the claim, 1 with evidence against, 1 mixed, 2 not coded on a side.
Figures to know
- +34%Issues resolved per hour, novice and low-skilled support agents
Erik Brynjolfsson et al., 2025 · Inequality“including a 34% improvement for novice and low-skilled workers but with minimal impact on experienced and highly skilled workers”
- +43%Task scores of consultants below average performance
Fabrizio Dell’Acqua et al., 2023 · Inequality“those below the average performance threshold increasing by 43% and those above increasing by 17%”
- +17%Task scores of consultants above average performance
Fabrizio Dell’Acqua et al., 2023 · Inequality“those below the average performance threshold increasing by 43% and those above increasing by 17%”
- +15%Business performance of high-performing entrepreneurs with an AI assistant
Nicholas G. Otis et al., 2024 · Inequality“high performers benefited by just over 15% from the AI assistant”
- −8%Business performance of low-performing entrepreneurs with an AI assistant
Nicholas G. Otis et al., 2024 · Inequality“low performers did about 8% worse”
Every number is quoted from its source and linked to the paper; estimates differ in method and horizon, so they are not directly comparable.
People to follow
- Daron Acemoglu agreesEconomist, Massachusetts Institute of Technology
Argues AI could widen the gap between capital and labour income even without lower wages.
Key work: Robots and Jobs: Evidence from US Labor Markets; The Race between Man and Machine: Implications of Technology for Growth, Factor Shares, and Employment
- David Autor it's complicatedEconomist, Massachusetts Institute of Technology
Author of work on the polarization of the US labour market and the growth of low-skill service jobs.
Key work: Why Are There Still So Many Jobs? The History and Future of Workplace Automation; The Growth of Low-Skill Service Jobs and the Polarization of the US Labor Market
- David J. Deming it's complicatedEconomist, Harvard University; Danoff Dean of Harvard College
Studies skills, education and the labour market, including how fast generative AI is being adopted.
Key work: The Rapid Adoption of Generative AI; Four Facts about Human Capital
- Anton Korinek agreesEconomist, University of Virginia
Co-author with Stiglitz of work on AI's implications for income distribution and unemployment.
Key work: Artificial Intelligence and Its Implications for Income Distribution and Unemployment; Excessive Volatility in Capital Flows: A Pigouvian Taxation Approach
- Rembrand Koning leans agreeAssociate professor, Harvard Business School
Harvard Business School researcher and co-author of the Kenyan entrepreneur experiment.
Key work: The Uneven Impact of Generative Artificial Intelligence on Entrepreneurial Performance: Evidence from a Field Experiment in Kenya; Global Evidence on Gender Gaps and Generative AI
Of these 3 people: 1 agrees, 2 it's complicated. Documented, cited positions only.
Of these 5 people: 2 agrees, 1 leans agree, 2 it's complicated. Documented, cited positions only.
Of these 5 people: 2 agrees, 1 leans agree, 2 it's complicated. Documented, cited positions only.
Newsletters and sources
- Paul KrugmanNewsletter · 5 posts in this guide
Regular posts on AI, labour and the distribution of income.
- Theory of the Game (Reid Hoffman)Newsletter · 1 posts in this guide
A technology investor's newsletter arguing for broadly shared gains from AI.
- Our World in Data: Artificial IntelligenceOur World in Data · dataset · outside this guide
Charts and data on AI across countries, free to reuse.
- Stanford AI Index ReportStanford Institute for Human-Centered AI · report · outside this guide
A yearly report with data on AI adoption, jobs and investment across countries.