Evidence (552 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
21267 claims
Filter claims →
Productivity
17978 claims
Filter claims →
Governance
17038 claims
Filter claims →
Human-AI Collaboration
16914 claims
Filter claims →
Org Design
11104 claims
Filter claims →
Innovation
11087 claims
Filter claims →
Labor Markets
6711 claims
Filter claims →
Skills & Training
5616 claims
Filter claims →
Inequality
4343 claims
Filter claims →
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 | 1880 | 496 | 296 | 1854 | 4721 |
| Organizational Efficiency | 2906 | 665 | 438 | 180 | 4210 |
| Governance & Regulation | 2162 | 929 | 480 | 247 | 3866 |
| Technology Adoption Rate | 1533 | 545 | 278 | 210 | 2593 |
| Decision Quality | 1391 | 534 | 321 | 173 | 2429 |
| Output Quality | 1298 | 472 | 231 | 145 | 2153 |
| AI Safety & Ethics | 682 | 821 | 230 | 90 | 1837 |
| Research Productivity | 855 | 253 | 121 | 425 | 1675 |
| Firm Productivity | 1105 | 171 | 175 | 73 | 1531 |
| Task Allocation | 735 | 229 | 361 | 99 | 1433 |
| Market Structure | 457 | 461 | 251 | 47 | 1222 |
| Innovation Output | 673 | 94 | 108 | 36 | 913 |
| Task Completion Time | 499 | 118 | 43 | 38 | 702 |
| Firm Revenue | 458 | 130 | 61 | 26 | 677 |
| Skill Acquisition | 381 | 122 | 113 | 34 | 650 |
| Consumer Welfare | 316 | 176 | 115 | 39 | 648 |
| Employment Level | 223 | 143 | 177 | 53 | 600 |
| Error Rate | 246 | 282 | 44 | 19 | 594 |
| Fiscal & Macroeconomic | 283 | 142 | 78 | 52 | 562 |
| Inequality Measures | 103 | 329 | 106 | 13 | 552 |
| Worker Satisfaction | 225 | 185 | 63 | 30 | 503 |
| Automation Exposure | 158 | 155 | 72 | 37 | 426 |
| Regulatory Compliance | 186 | 126 | 35 | 14 | 362 |
| Team Performance | 193 | 56 | 51 | 24 | 326 |
| Developer Productivity | 224 | 58 | 27 | 13 | 323 |
| Wages & Compensation | 148 | 108 | 50 | 17 | 323 |
| Training Effectiveness | 218 | 44 | 21 | 27 | 313 |
| Job Displacement | 23 | 159 | 53 | 5 | 240 |
| Hiring & Recruitment | 109 | 61 | 32 | 11 | 215 |
| Skill Obsolescence | 16 | 107 | 26 | 6 | 155 |
| Creative Output | 71 | 44 | 28 | 6 | 150 |
| Social Protection | 58 | 31 | 12 | 3 | 104 |
| Labor Share of Income | 29 | 43 | 25 | 2 | 99 |
| Worker Turnover | 45 | 29 | 6 | 4 | 84 |
| Industry | — | — | — | 1 | 1 |
AI-enabled agricultural platforms may alter bargaining power and margin shares across value-chain actors and may either include or exclude smallholders.
Research question and risk identified in a conceptual discussion; no platform data or estimated distributional effects are reported.
An alternative AI-patent proxy reproduces the same U-shaped relationship between AI and the urban-rural income gap.
The paper reports robustness using an AI-patent proxy in addition to the baseline Bartik shift-share measure.
The theoretical model predicts that the urban-rural income gap decreases at low levels of AI task autonomy and increases at high levels of AI task autonomy.
The model derives the marginal effect of AI on the log urban-rural income ratio. At low AI autonomy, rural productivity growth is assumed to exceed urban capital deepening; at high autonomy, rural productivity gains saturate while urban capital deepening continues.
Artificial intelligence has a U-shaped relationship with the urban-rural income ratio: AI initially narrows the gap, but at higher levels of AI exposure it widens the gap, with a unique interior minimum.
A dual-sector general-equilibrium model predicts a sign reversal in the marginal effect of AI on the urban-rural wage ratio. Empirically, the paper uses panel data from 288 Chinese prefecture-level cities from 2012-2023 and a Bartik shift-share AI-exposure measure; the Lind-Mehlum U-test rejects monotonicity at the 5% level.
The distribution of economic benefits from the evaluated pricing mechanisms is heterogeneous across households.
Household-level allocation and evaluation of user costs, prosumer revenues, and shares of the community benefit in the simulated 20-household community.
The benefits and harms of digital tools are unevenly distributed according to socioeconomic status, digital infrastructure, parental support, and school capacity.
The article synthesizes evidence on inequality and access, while noting underrepresentation of low- and middle-income contexts and marginalized groups.
AI adoption can simultaneously widen growth gaps across countries and inequality within countries.
The commentary summarizes results from Che, Xin, and Yoshida's calibrated small open economy overlapping-generations model of endogenous AI adoption covering 15 Asia-Pacific economies.
AI regulatory tightening is expected to increase dispersion in firm outcomes because firms differ in their SLC.
Theoretical prediction concerning heterogeneous firm responses to privacy, transparency, algorithmic-audit, and related AI regulation; no event-study or difference-in-differences results are reported.
Digital transformation promotes economic growth and convergence but also creates risks of digital inequality and technological dependence.
The study combines cross-country comparative analysis with econometric and scenario modeling, and explicitly evaluates digital-development asymmetries and technological dependence.
Unequal access to need crafting across firms or worker skill groups could widen wage gaps, while broad adoption could help diffuse AI-complementary skills and mitigate displacement risks.
Distributional hypothesis based on differential adoption and access; no wage, inequality, or displacement data are presented.
AI enforcement may generate heterogeneous distributional effects across firms, individuals, large businesses, small businesses, and the informal sector.
Policy implication concerning differential exposure to AI-enabled tax enforcement; no distributional estimates are reported.
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.
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.
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.
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.
Correcting decisions flagged by AIBF substantially improves the disparate impact ratio but does not achieve legal parity because merit features retain residual proxy correlation.
The paper reports a post-correction disparate-impact analysis and identifies residual proxy correlation in features classified as merit.
On the Adult dataset, the mean counterfactual shift was +7.5 score points for the privileged group and −8.0 score points for the disadvantaged group.
Evaluation on the held-out Adult test dataset using the AIBF counterfactual shift.
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.
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 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.
Since the 1970s, life expectancy has stalled for people without a college degree while continuing to rise rapidly for college-educated Americans.
Descriptive population-level evidence synthesized from empirical studies of life expectancy by educational attainment.
AI benefits are expected to be uneven across researchers, teams, and fields, concentrating where AI fills workflow bottlenecks and where complementary human skills are strong.
Theoretical argument that differences in production plans, workflows, and human capital generate heterogeneous returns to AI adoption.
AI-driven displacement pressure is present across the full occupational distribution, while augmentation gains and associated economic returns are disproportionately concentrated in occupations requiring higher levels of formal education.
The study compares modeled substitution, facilitation, and net labor-capacity measures across the five O*NET Job Zones, which are used as occupational proxies for educational attainment.
The paper argues that imprecision or ranking noise can preserve a nonzero probability that structurally disadvantaged group members are selected, functioning as an unintentional equity buffer; increasing precision erodes this buffer.
Mechanistic interpretation of the Gaussian ranking model: noise randomizes selection near the threshold, while higher fidelity concentrates selection according to the group separation.
The predictive accuracy of the saliency models is systematically higher for younger, White, and politically moderate viewers than for older, Black, and ideologically extreme viewers.
Demographic subgroup comparisons using 3,023 US adults recruited to national quotas; comparisons used center-corrected scores with recruitment-wave and device fixed effects and participant-clustered standard errors.
Improving average simulation accuracy or the accuracy of a target group does not necessarily improve representational equality.
Comparison of contextual-adaptation and parametric-modification interventions using both average accuracy and cross-country equality metrics.
Demographic-parity-constrained AI equalised access to high-opportunity recommendations across socioeconomic groups but did not improve mean simulated feasibility-adjusted fit or reduce the socioeconomic FAF gap.
Study 1 simulation comparing demographic-parity-constrained AI with accuracy-optimised AI across socioeconomic quintiles.
Organizational training, managerial support, learning opportunities, digital confidence, and organizational culture can either widen or narrow AI-literacy-related inequality gaps.
Moderating variables proposed in the authors' conceptual framework; no empirical moderation analysis is reported.
AI literacy may account for a larger share of differences in individual workplace outcomes than AI access when access to AI is widespread.
Conceptual inference from digital-divide theory and cited literature on unequal digital skills, uses, and outcomes; no original empirical estimate is provided.
Twenty-five percent of the studies analyzed in the review originated from India, while Africa and Latin America were underrepresented.
Descriptive geographical characterization of the 56 studies included in the systematic review.
The effects of technological innovation, economic development, industrial upgrading, environmental regulation, urbanization, and government intervention on AHCC vary across regions and over time, with some effects changing sign.
Regional and temporal subsample analyses assessing spatiotemporal heterogeneity in panel and spatial regression results.
Only a small number of provinces achieved coordinated AI–HED development, with most of these provinces located in eastern China and two located in western China; most provinces remained imbalanced.
Provincial AHCC classification and spatial distribution analysis.
The interaction between perceived job replaceability and task codifiability produces a polarization in perceived income-change outcomes across task types and skill groups.
Interaction analyses by task codifiability and heterogeneity analyses across skill groups and occupation types; the reported contrasts show a stronger positive association for low-codifiability occupations and a stronger exposure-related downside for routine and medium-skill workers.
Coding-agent adoption and productivity gains were concentrated among developers who were already more active and well connected.
The simulation used an active-developer-led diffusion process and compared per-capita planned and completed tasks for developers aware versus unaware of coding agents.
The sequence in which weak links are automated will shape labor demand, sectoral transitions, and inequality.
Conceptual distributional analysis linking the order of task automation to labor-market adjustment; no labor-market data or quantitative distributional estimates are provided.
Remote and hybrid work arrangements are disproportionately used by women, people of color, and workers with disabilities.
The paper cites empirical evidence and surveys comparing remote and hybrid work participation across demographic groups.
Algorithmic governance can alter the distribution of public goods and services, with implications for inequality and social welfare.
Conceptual synthesis of distributional and welfare implications; the paper explicitly notes that it does not provide new causal estimates or quantification.
MADDPG achieves slightly lower profit but the fairest profit distribution among agents.
Same empirical benchmark in the simulated marketplace; reported comparison shows MADDPG yields marginally lower average profit but the most equitable profit distribution across agents.
Leadership, not technology alone, determines whether innovation creates inclusive progress or concentrated power.
Theoretical argument illustrated with cases across industries in the chapter; no randomized or large-n causal identification reported.
The risks and gains from AI adoption are unevenly distributed across sectors, skill levels and regions.
Cross-study synthesis in the review pointing to heterogeneity in impacts by sector, worker skill level, and geography.
Trade, offshorability, educational attainment, employment rates and mark-ups play secondary, period-specific roles in driving wage distributional changes.
Auxiliary/robustness analyses in the empirical framework applied to Spanish micro-data (2000-2019); these factors appear as secondary and vary by period in their association with wage distributional outcomes.
Significant wage shares would shift from the top 10% towards middle and bottom groups (absent task displacement).
Distributional counterfactual exercises on Spanish micro-data (2000-2019) with task-sensitive automation index and instrumental variables.
As AI forecasting ability improves, the distribution of wealth and impact will increasingly depend on the quality of human judgment.
Theoretical argument/analytical claim in the paper projecting how improved predictive power shifts importance to judgment quality; no empirical sample reported.
AI leads to redistribution of economic opportunities with differential effects across sectors and categories of workers.
Synthesis of specialized literature documenting heterogeneous sectoral and worker-level impacts of AI; no original empirical sample provided.
A model of the distributive effects of AI applications is constructed with a particular focus on the manufacturing and service sectors.
Paper reports a focused distributive-effects model applied to manufacturing and service sectors (modeling exercise; no empirical sample size reported).
The paper analyzes distributive effects of AI across different industries and demographic groups.
Explicit statement in the paper: a model of distributive effects is constructed to evaluate impacts across industries and demographic groups. No empirical sample size reported (theoretical/modeling work).