Evidence (4148 claims)
Search and filter individual claims pulled from the papers. Looking for a specific finding ("what's the effect on wages?"), you're in the right place. Want to compare whole outcome categories against each other instead? Use the Evidence Explorer.
The board below groups claims two ways: by broad theme (nine paper-level topics) and by outcome category (the 34 claim-level outcomes that the Explorer and Syntheses also use).
Browse by theme
Nine broad, paper-level topics. Click one to filter the claims below.
Adoption
20058 claims
Filter claims →
Productivity
17184 claims
Filter claims →
Governance
16099 claims
Filter claims →
Human-AI Collaboration
16034 claims
Filter claims →
Innovation
10501 claims
Filter claims →
Org Design
10496 claims
Filter claims →
Labor Markets
6444 claims
Filter claims →
Skills & Training
5385 claims
Filter claims →
Inequality
4148 claims
Filtered →
Claims by outcome category
Counts by direction of finding. These are the same 34 outcome categories the Explorer compares and the Syntheses are written for. A linked row has a published synthesis.
| Outcome | Positive | Negative | Mixed | Null | Total |
|---|---|---|---|---|---|
| Other | 1820 | 479 | 278 | 1820 | 4588 |
| Organizational Efficiency | 2711 | 616 | 401 | 173 | 3922 |
| Governance & Regulation | 2075 | 886 | 459 | 246 | 3714 |
| Technology Adoption Rate | 1467 | 530 | 258 | 206 | 2488 |
| Decision Quality | 1281 | 496 | 289 | 152 | 2228 |
| Output Quality | 1227 | 447 | 207 | 138 | 2025 |
| AI Safety & Ethics | 634 | 754 | 207 | 83 | 1688 |
| Research Productivity | 826 | 241 | 114 | 422 | 1624 |
| Firm Productivity | 1052 | 154 | 163 | 66 | 1441 |
| Task Allocation | 685 | 211 | 331 | 99 | 1335 |
| Market Structure | 433 | 423 | 242 | 46 | 1150 |
| Innovation Output | 639 | 91 | 105 | 34 | 871 |
| Task Completion Time | 476 | 113 | 43 | 36 | 672 |
| Firm Revenue | 445 | 126 | 58 | 25 | 656 |
| Skill Acquisition | 364 | 119 | 109 | 34 | 626 |
| Consumer Welfare | 288 | 167 | 104 | 31 | 592 |
| Employment Level | 214 | 140 | 174 | 50 | 582 |
| Error Rate | 230 | 251 | 35 | 16 | 535 |
| Fiscal & Macroeconomic | 268 | 136 | 71 | 50 | 532 |
| Inequality Measures | 100 | 307 | 96 | 12 | 515 |
| Worker Satisfaction | 221 | 173 | 60 | 30 | 484 |
| Automation Exposure | 155 | 138 | 65 | 36 | 398 |
| Regulatory Compliance | 171 | 120 | 30 | 13 | 335 |
| Developer Productivity | 222 | 58 | 27 | 13 | 321 |
| Team Performance | 188 | 56 | 50 | 24 | 320 |
| Wages & Compensation | 146 | 104 | 46 | 16 | 312 |
| Training Effectiveness | 207 | 41 | 21 | 26 | 298 |
| Job Displacement | 23 | 153 | 52 | 4 | 232 |
| Hiring & Recruitment | 102 | 57 | 30 | 11 | 202 |
| Skill Obsolescence | 16 | 102 | 24 | 6 | 148 |
| Creative Output | 71 | 42 | 23 | 6 | 143 |
| Social Protection | 57 | 30 | 11 | 3 | 101 |
| Labor Share of Income | 29 | 42 | 24 | 2 | 97 |
| Worker Turnover | 43 | 29 | 6 | 4 | 82 |
| Industry | — | — | — | 1 | 1 |
Inequality
Remove filter
Educational data should be treated as an economic asset with potential negative externalities, including privacy harms and surveillance, requiring data-governance and property-rights analysis.
Normative political-economy implication drawn from the essay's analysis of platformisation and datafication; this is a proposed analytical framework rather than an empirically estimated result.
The platformisation and datafication of education are reshaping educational markets, data governance, and pedagogical practices while introducing new power asymmetries based on surveillance and data monetization.
Conceptual analysis of EdTech and platform actors, drawing on existing scholarship and contemporary developments in technology and education.
Data ownership and algorithmic control influence who receives economic rents from AI-driven value creation.
Qualitative case studies and literature synthesis focused on data rights, data dividends, ownership, and algorithmic governance.
Automation causes employment disruption in some occupations, while the net employment effect varies by sector, skill composition, and institutional context.
Synthesis of empirical studies within the review, with heterogeneity reported across sectors, worker skill compositions, and institutional settings.
Short-run disruption includes job churn and wage compression for affected groups, while long-run outcomes depend on reskilling, capital re-allocation, and institutions.
Asserted in the supplied example contribution; no longitudinal employment, wage, or reskilling evidence is provided.
Regions with higher human capital and adoption capacity capture more productivity gains, while disadvantaged regions face stagnation.
Presented as a regional heterogeneity claim; no regional panel, productivity measure, or comparative estimate is supplied.
AI substitutes for routine cognitive and manual tasks, shifting worker duties toward nonroutinized, interpersonal, and creative tasks.
Presented as a task-based displacement claim; no task-level dataset or estimates are supplied.
Net employment effects are modest short-run losses, with potential long-run gains if complementary skill investment and policy support occur.
Asserted in the supplied example contribution; the text provides no employment panel, identification strategy results, or quantified estimates.
High-skill cognitive tasks and complementary occupations gain earnings, while routine tasks and low-skill occupations face displacement and wage pressure.
Asserted in the supplied example contribution; no occupational employment or wage data are presented.
AI-driven automation accelerates occupational task reallocation, raising productivity but producing uneven wage effects.
Asserted in the supplied example contribution; no underlying paper, dataset, sample, or statistical analysis is provided.
Rapid technological change in diagnostics, pharmaceuticals, and digital health is shifting which countries produce key health inputs.
Qualitative policy analysis based on observed technological diffusion and changes in production capacity.
The sign and magnitude of digitalization's effect on aggregate inequality are not universal because they depend on the dominant channel and the inequality measure selected.
Analytical characterization and comparative-static or simulation experiments in a heterogeneous-agent growth model evaluate multiple economy-wide inequality statistics.
Within-skill-group inequality can move independently of between-skill-group inequality, so skilled–unskilled wage gaps do not fully characterize the distributional effects of digitalization.
The model distinguishes worker ability levels and separately analyzes between-group and within-group inequality measures.
Digitalization-induced changes in labor-time supply—including flexibility, intensity, and hours—can affect earned income and contribute to distributional differences.
The model explicitly includes a supply-side labor-time modulation mechanism and evaluates its implications for earned income and inequality.
Digital tools that raise labor productivity can change relative wages across workers or tasks and thereby affect wage inequality.
The theoretical model incorporates a demand-side labor-productivity channel that raises productivity for tasks or workers and examines its effects on relative wages.
Automation changes the composition of labor demand by substituting for some tasks, with consequences for economic growth and wage inequality.
The model explicitly includes automation as one of three digitalization mechanisms and traces its effects on labor demand, growth, and distributional outcomes.
Digitalization can either increase or decrease wage inequality, depending on the inequality metric used and the relative strength of its underlying channels.
Theoretical dynamic growth/general-equilibrium model with heterogeneous workers, endogenous occupational choice, and digitalization operating through automation, productivity, and labor-time modulation; comparative-static or simulation analysis considers multiple inequality measures.
The reviewed evidence is heterogeneous in its measures, contexts, and outcomes, and many underlying studies provide limited causal identification.
Limitations reported by the narrative review concerning the heterogeneity and methodological quality of the secondary literature.
Corporate narratives portray digital twins and industrial AI as enabling synchronized, centrally managed, and predictive factory operations, but observed shop-floor practice departs from this framing.
Discourse analysis of promotional texts, vendor claims, and managerial presentations contrasted with ethnographic observation and worker interviews.
Industrial AI and digital-twin systems in the Vietnamese electric-vehicle welding plant do not fully automate or simply replace human labor; their operation depends on workers’ routine desynchronization of machine-directed processes.
Qualitative ethnographic fieldwork combining shop-floor observation and semi-structured interviews at a single high-tech electric-vehicle welding factory in Vietnam.
The paper argues that digitalization increases the organic composition of capital and expands a reserve army of labor while contributing to the emergence of a new middle class composed mainly of medium-skilled workers.
Marxist political-economy interpretation of the empirical pattern, linking reduced low-skill employment and increased medium- and high-skill demand to capital deepening, surplus labor, and middle-class formation.
The three organizations studied—Testworks, Kakao, and Microsoft—use AI to widen labor-market access, but none integrates older adults and persons with disabilities within a single sociotechnical system.
This finding comes from the paper's multiple-case cross-case analysis of three theoretically distinct organizations.
The employment effect of AI for persons with disabilities changes from negative to positive beyond critical levels of AI adoption.
The paper summarizes Abid et al. (2024), which examined linear and nonlinear effects; the underlying sample size and numerical thresholds are not reported in the supplied text.
Disclosure requirements may increase firms' investment in documentation and fairness-aware development while also creating entry barriers for smaller firms unless compliance support is provided.
Conceptual analysis of firm incentives and market-structure implications; no measured compliance costs or entry effects are reported.
Workforce preparedness differs significantly across gender, age, job level, and years of experience, while differences across sectors are not statistically significant at the 5% level.
Independent-samples t-tests and one-way ANOVA applied to survey data from 600 workers.
GPU count showed more consistent positive associations with citation and award outcomes than newer hardware generation; GPU count was positively associated with awards, whereas aggregate capability and hardware generation showed no comparably robust award evidence.
Complementary citation analyses and award models, including linear-probability and Firth models; the award analysis used 5,357 papers.
Reported GPU capability is neither necessary nor sufficient for high citation impact: 85.5% of high-capability papers were not highly cited, and most highly cited papers were outside the high-capability group.
Overlap analysis between the top 20% of papers by reported GPU capability and the top 10% by citations within publication-year-by-venue groups.
The annual top 20% of papers ranked by reported GPU capability accounted for 83.9%–89.9% of reported GPU capability during 2020–2023, but only 27%–32% of citations and 20%–33% of paper awards.
Annual concentration analysis comparing the share of total reported GPU capability, citations, and awards attributable to the top 20% of GPU-quantifiable papers.
The increase in reported GPU capacity was driven mainly by adoption of newer hardware generations and expansion of medium-scale multi-GPU configurations, rather than by a field-wide shift to very large GPU clusters.
Year-over-year descriptive comparisons of reported GPU counts and hardware generations; configurations using nine or more GPUs accounted for only 11.7% of papers with reported GPU counts in 2025.
In the second preregistered experiment, gender feedback substantially increased selection of Western women but did not significantly increase selection of Eastern women.
Preregistered randomized experiment recruiting participants from China, South Korea, Italy, and Germany; 1,443 participants completed the task.
Gender feedback causally increased selection of Western women for a promotional campaign, but did not significantly increase selection of Eastern women.
Preregistered randomized experiment with 750 participants from the United States. Participants received either gender-composition feedback or control feedback before making one additional scholar selection.
Papers with citation diversity statements cite more authors with English-origin names and fewer authors with Chinese, Japanese, and Arab names.
Paper-level regression analysis comparing 216 papers with citation diversity statements to 432 matched papers without them, with controls for publication year, field, region, team size, and citation impact.
Citation diversity statements are associated with a shift away from citations to gender-blind names, while citations to men’s names do not significantly differ between papers with and without the statements.
Paper-level regressions across the matched sample; the coefficient for men’s names was not statistically significant, whereas the coefficient for gender-blind names was negative and statistically significant.
Avoidance of binding duties trades off lower short-term political and administrative costs for platforms against potential long-term welfare losses from unmitigated harms to children.
Normative welfare analysis of platform compliance costs, enforcement design, and child-harm externalities; the paper provides no quantified welfare estimate.
Age-based bans and administrative rules impose different costs and compliance strategies from enforceable duties of care, influencing firms’ regulatory optimization and innovation paths.
Comparative economic analysis of blanket platform restrictions, reporting requirements, and enforceable duties, including their implications for user bases, advertising revenues, monitoring, and algorithmic-safety investment.
The Albanese government presented the Online Safety Amendment Bill 2024 as applying a duty of care while structuring it to avoid open-ended, binding legal obligations and liabilities.
Analysis of the bill’s legal text, explanatory memorandum, government statements, press releases, and parliamentary debate concerning the under-16 social-media restriction.
The paper examines platform labour participation in relation to three household-division outcomes: domestic-service expenditure, spousal labour supply, and labour-force exit due to household care duties.
The authors construct proxy dependent variables from the CSS2023 survey and estimate logistic and ordinary least squares regression models with demographic, socioeconomic, household, and geographic controls.
Platform workers observed in the CSS2023 are younger, more educated, and more likely to belong to higher-income dual-earner households than the precarious-gig-worker profile commonly portrayed in qualitative research.
Descriptive comparison reported by the authors based on the CSS2023 sample; the supplied text does not report numerical differences or statistical tests for these demographic characteristics.
Treating frontier AI as scientific infrastructure implies that AI-access decisions shape the long-run distribution of scientific capability and power.
Conceptual interpretation of the model's endogenous feedback between AI access, credit, resources, future access, and laboratory participation.
Providing baseline AI tools to all laboratories narrows inequality, but can reduce aggregate scientific output; the output loss is bounded in the model.
Policy extension of the theoretical model that gives all laboratories baseline AI capability and incorporates an intervention-cost parameter; welfare and comparative-statics analysis evaluate the trade-off.
The strongest predictors of AI job growth in US counties were the share of STEM degrees, local labor-market tightness, and patenting activity, while manufacturing intensity was negatively associated with AI job growth.
Andreadis et al. used US county-level job-posting data from 2014 to 2023 and examined predictors of AI job growth.
AI-related job shares are highly unequal across US counties: Slope County, North Dakota, had an AI job share of 10%, Santa Clara County, California, had a share of 8.2%, and many rural counties had virtually none.
Andreadis et al. analyzed US county-level job-posting data from 2014 to 2023.
The IMF estimates that almost 40% of global employment is exposed to AI, with exposure of 60% in advanced economies, 40% in emerging economies, and 26% in low-income countries.
The paper cites IMF cross-country estimates of employment exposure to AI.
The urban skill upgrading described for Italian cities occurred primarily through low-skill workers leaving and high-skill workers moving in, rather than through existing workers acquiring new skills.
The paper's interpretation of the Capello and Lenzi evidence on urban labor-market adjustment.
In Italian cities, low-skill employment fell while high-skill employment expanded and overall labor-force participation remained steady.
The paper summarizes Capello and Lenzi's urban findings from Italian NUTS-3 data for 2009–2019.
The labor-market value of multilingual competencies varies by occupational sector: multilingualism is a clearer professional asset in client-facing, international, and language-services occupations, but its recognition is more inconsistent in technical and administrative fields with entrenched standardized, monolingual communication norms.
Synthesis of sector-specific literature on multilingualism and labor-market recognition, citing Roberts (2013) and Wilczewski and Alon (2023).
Algorithmic recruitment tools are not inherently more biased than the human-mediated processes they replace; their effects on equity depend substantially on system design, training data, and auditing practices.
Review of countervailing evidence comparing structured or algorithmic screening with unstructured human interviews, citing Bogen and Rieke (2018) and Stone et al. (2024).
Which long-run ownership regime occurs is determined by law and initial conditions rather than by technology alone.
The paper’s theoretical ownership analysis treats legal ownership arrangements and initial ownership as determinants of ε_t and the terminal regime.
The model permits three terminal regimes: a rentier regime with a positive human ownership share, a fully decoupled regime in which the human share converges to zero, and a socialized regime in which states or public funds hold ownership on behalf of citizens.
Theoretical characterization of terminal regimes based on ownership dynamics, retained earnings, buybacks, cross-holding, and public ownership.
In the model, an arbitrarily large GDP is welfare-relevant to humans only through their financial ownership share of the corporate machine network.
Theoretical welfare analysis sets human production and consumption to zero and summarizes the human stake using the ownership-share state variable ε_t.