Evidence (14922 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
9047 claims
Filter claims →
Productivity
8066 claims
Filter claims →
Governance
7278 claims
Filter claims →
Human-AI Collaboration
6912 claims
Filter claims →
Org Design
4439 claims
Filter claims →
Innovation
4359 claims
Filter claims →
Labor Markets
3652 claims
Filter claims →
Skills & Training
3018 claims
Filter claims →
Inequality
2160 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 | 795 | 210 | 105 | 955 | 2131 |
| Governance & Regulation | 886 | 414 | 197 | 126 | 1654 |
| Organizational Efficiency | 826 | 204 | 129 | 87 | 1257 |
| Technology Adoption Rate | 681 | 259 | 128 | 110 | 1189 |
| Research Productivity | 464 | 138 | 65 | 349 | 1028 |
| Output Quality | 503 | 196 | 61 | 53 | 813 |
| Decision Quality | 351 | 180 | 84 | 51 | 673 |
| AI Safety & Ethics | 238 | 288 | 71 | 34 | 637 |
| Firm Productivity | 455 | 58 | 92 | 20 | 631 |
| Market Structure | 186 | 172 | 123 | 25 | 511 |
| Task Allocation | 222 | 70 | 76 | 34 | 407 |
| Innovation Output | 238 | 28 | 48 | 18 | 334 |
| Skill Acquisition | 177 | 62 | 62 | 17 | 318 |
| Employment Level | 107 | 57 | 108 | 13 | 287 |
| Fiscal & Macroeconomic | 135 | 72 | 44 | 26 | 284 |
| Firm Revenue | 172 | 50 | 28 | 5 | 256 |
| Consumer Welfare | 121 | 68 | 45 | 12 | 246 |
| Task Completion Time | 183 | 33 | 10 | 13 | 240 |
| Inequality Measures | 45 | 126 | 50 | 6 | 227 |
| Worker Satisfaction | 95 | 74 | 23 | 12 | 204 |
| Error Rate | 77 | 98 | 11 | 4 | 190 |
| Regulatory Compliance | 84 | 73 | 17 | 7 | 181 |
| Automation Exposure | 61 | 61 | 27 | 14 | 166 |
| Training Effectiveness | 98 | 21 | 14 | 19 | 154 |
| Wages & Compensation | 78 | 37 | 25 | 6 | 146 |
| Developer Productivity | 105 | 18 | 14 | 6 | 144 |
| Team Performance | 87 | 17 | 28 | 10 | 143 |
| Job Displacement | 12 | 83 | 23 | 1 | 119 |
| Hiring & Recruitment | 53 | 8 | 8 | 3 | 72 |
| Social Protection | 39 | 17 | 8 | 2 | 66 |
| Creative Output | 32 | 20 | 8 | 3 | 64 |
| Skill Obsolescence | 5 | 50 | 6 | 1 | 62 |
| Labor Share of Income | 17 | 20 | 17 | — | 54 |
| Worker Turnover | 15 | 15 | — | 3 | 33 |
| Industry | — | — | — | 1 | 1 |
The reviewed studies rely on feature-engineered sentiment indices derived from lexicons or sentence-level classification.
Review synthesis noting frequent use of lexicon-based sentiment scoring and sentence-level classification to produce engineered sentiment features across the sampled studies.
Most studies focus on the U.S. stock market.
Findings from the review of 42 studies indicating a majority of the reviewed works concentrate on U.S. markets (geographic coding/synthesis across studies reported by the authors).
Machine learning methods have been widely used to predict stock prices using technical indicators and sentiment features, mostly extracted from social media and news.
Systematic review of the literature summarized in the paper (corpus of 42 studies published 2014–2025) reporting that many reviewed studies use ML to predict stock prices and that sentiment inputs commonly come from social media and news sources.
Visa recapture would reclaim approximately 339,000 unused visas from prior years, delivering immediate backlog relief under existing statutory authority.
Authors' calculation/estimate of cumulative unused employment-based visas available for recapture (presumably based on historical visa usage statistics from the Department of State); the excerpt does not show the year-by-year accounting or the assumptions used to reach 339,000.
Dependent exemption (excluding spouses and minor children from counting toward the annual cap) would ensure that all 140,000 visas are allocated to independently qualified principal workers rather than divided among family members.
Policy design claim; premise depends on current family-derivative usage of the cap and would require counting statistics (number of visas currently used by dependents) to quantify effect—those counts are not provided in the excerpt.
Increasing the annual employment-based visa ceiling would alleviate the overall shortage that persists regardless of allocation methods.
Logical/policy claim that raising the statutory cap increases supply; the excerpt does not include a quantitative elasticity, model, or simulation showing the required increase or magnitude of backlog reduction.
Phasing out the seven-percent per-country cap would gradually transition visa allocation from nationality-based limits to a demand-driven system, allowing applicants from high-demand countries to advance in the backlog without causing abrupt increases in wait times for those from low-demand countries.
Policy proposal with implied simulation/modeling rationale (demand-driven allocation); the excerpt does not provide a formal model, simulation parameters, or empirical test showing the gradual, non-disruptive transition.
Closing the gender gap in digital skill use at work will require more than increasing women’s participation in STEM education or occupations; workplace organisation, task allocation, progression pathways, and organisational practices also need attention.
Policy inference drawn from empirical finding that education, field of study and occupational controls explain only a minority of the gender gap in advanced digital task use in ESJS decompositions.
Partial adoption of artificial agents can still improve aggregate outcomes.
Mixed-population analysis and simulation results reported in the paper showing aggregate welfare improvements under partial adoption scenarios.
Unilateral entry of artificial-agent technology is feasible: adopters are not structurally penalized.
Analysis of mixed populations of adopters and non-adopters presented in the paper (mixed-population evolutionary analysis and simulations); exact parameter sweeps and sample sizes are not provided in the abstract.
Artificial agents can shift the learning dynamics to favour coordination outcomes.
Findings from evolutionary dynamics analysis and reinforcement learning experiments demonstrating changes in learning trajectories and equilibrium selection when artificial agents are present.
Introducing artificial agents that use globally observable signals increases coordination among agents.
Experimental results reported in the paper using reinforcement learning experiments and evolutionary-dynamics simulations with artificial agents that observe global signals (details of experimental setup and sample sizes are not specified in the abstract).
AI adoption raises ethical controversies that require public policy action to promote social equity and economic opportunity.
Synthesis of debates on AI ethics and policy from the literature; the paper provides normative recommendations rather than empirical measurement of policy impact.
Labor market regulatory frameworks should be updated in response to AI adoption.
Narrative review of regulatory issues and recommendations drawn from existing literature and policy debates; no empirical testing of specific regulatory interventions included.
Social safety net programs need changes to respond to AI-related labor market disruption.
Policy analysis and synthesis of prior proposals in the literature; the review presents arguments rather than new program evaluation data.
There is an urgent need for education and training policy to address AI-driven changes in the labor market.
Policy-focused literature review and the authors' policy recommendations based on synthesis of studies on skill demand shifts; no primary policy evaluation or randomized trial reported.
AI generates employment opportunities emerging from new technologies and innovation.
Narrative review of studies and examples in the literature cited by the paper; no new empirical measurement or sample provided in this review itself.
This study extends the technology–organisation–environment (TOE) theory by providing comprehensive empirical evidence of internal and external factors affecting BT adoption.
Use of the TOE framework to structure empirical analysis on 27,400 firm-year observations (2013–2021) linking technology (AI), organisation (corporate culture), and environment (market competition, government support, digital financial development) variables to BT adoption outcomes.
Environmental factors—market competition, government support, and the level of digital financial development across provinces—positively affect BT adoption.
Empirical tests using the 27,400 firm-year sample (2013–2021) incorporating provincial- and market-level environmental variables (market competition, measures of government support, and provincial digital financial development indices) alongside firm-level data and BT adoption coding from annual reports.
Externally oriented corporate cultures, specifically competition-oriented and creation-oriented cultures, positively affect BT adoption.
Same sample of 27,400 firm-year observations (2013–2021). Corporate culture indicators (competition- and creation-orientation) collected via Python web crawler from the management discussion & analysis (MD&A) sections of annual reports; BT adoption measured by manual annual report keyword search and content validation.
AI technology positively affects blockchain technology (BT) adoption.
Empirical analysis of 27,400 firm-year observations of Chinese A-share listed firms (2013–2021). AI technology measured using AI patent data collected via a Python web crawler from annual report MD&A sections and China National Knowledge Infrastructure (CNKI). BT adoption identified by manual search of annual reports for the keyword 'blockchain technology' and content assessment to confirm adoption status.
To alleviate adverse spatial spillovers, it is necessary to strengthen interactive development between digital–real integration and New Quality Productive Forces, foster interregional cooperation, and optimize resource allocation.
Policy recommendations derived from the paper's empirical findings (bidirectional positive relationship and negative spatial spillovers) — normative conclusion based on observed results.
The promotional effect of digital–real integration on New Quality Productive Forces is slightly stronger than the reverse effect (New Quality Productive Forces on digital–real integration).
Comparison of estimated coefficients from the GS3SLS spatial simultaneous equations model (paper reports the coefficient for integration→productive-forces is marginally larger than productive-forces→integration).
Workers who reported clear career pathways, internal mobility, and opportunities to apply newly acquired skills demonstrated higher optimism and stronger retention intentions.
Subgroup analyses within the 5,000-worker survey showing that respondents reporting clear career pathways, internal mobility, and opportunities to apply new skills had higher career optimism scores and greater self-reported retention intentions.
Career optimism is strongly associated with perceptions of AI-related competencies.
Survey measures of respondents' perceptions of their AI-related competencies were analyzed against career optimism scores in the national sample; paper reports a strong association.
Career optimism is strongly associated with financial stability.
Reported associations in the cross-sectional survey linking respondents' financial stability indicators with their career optimism measures (national sample of 5,000 workers).
Career optimism is strongly associated with organizational support for skill development.
Survey analyses correlating measures of perceived organizational support for skill development with respondents' career optimism scores in the 5,000-worker sample.
Career optimism is strongly associated with access to advancement opportunities.
Cross-sectional analyses of the nationally representative survey (5,000 workers) examining organizational factors associated with career optimism; reported strong association between self-reported access to advancement opportunities and measured career optimism.
The RL-FRB/US model achieved substantially lower debt (reported as debt-to-GDP ratios) by 2024: RL-FRB/US model: 26,535 trillion $ vs. FRB/US model: 30,186 trillion $, attributed to more strategic debt management during expansionary periods.
Reported simulation outputs of aggregate federal debt (paper labels these as debt-to-GDP ratios though reported as absolute trillion-dollar figures) comparing RL-FRB/US and baseline FRB/US, with interpretation attributing improvements to model fiscal policy behavior during expansions.
During recessions the RL-FRB/US model delivered superior counter-cyclical responses, with unemployment peaks significantly reduced—for example, during the 1982 recession peak unemployment reached 9.9% in the RL-FRB/US simulation versus 10.9% in traditional simulations.
Counterfactual/historical-recession simulation comparisons reported in the paper showing peak unemployment during specified recessions (example given: 1982); based on model-simulated recession scenarios versus baseline FRB/US simulation.
By 2024Q2 the RL-FRB/US model produced lower unemployment: 3.23% versus FRB/US model: 3.96%.
Reported simulation outputs for unemployment rate at 2024Q2 comparing RL-FRB/US and baseline FRB/US in the paper; derived from the model simulations.
By 2024Q2 the RL-FRB/US model achieved higher real GDP: 23,407 trillion $ versus FRB/US model: 23,218 trillion $.
Reported point-simulation outputs for 2024Q2 from the paper comparing RL-FRB/US and baseline FRB/US; based on the model simulations rather than observed national accounts data.
The RL-FRB/US model demonstrates significant performance improvements over baseline FRB/US simulations in the period 2000–2024.
Comparative simulation experiments reported in the paper, contrasting RL-FRB/US outputs vs. baseline FRB/US simulations over the 2000–2024 period (presumably quarterly simulation runs); exact number of runs or statistical tests not specified in the statement.
Generative AI (GenAI) systems have assumed increasingly crucial roles in selection processes, personnel recruitment and analysis of candidates' profiles.
Contextual/introductory claim in the paper; supported by cited literature and domain observation rather than primary data from this study (no sample size required).
Focused, small Skills (2–3 modules) are more effective than comprehensive documentation-style Skills.
Experimental analysis comparing Skill granularity: authors report higher pass-rate gains for Skills composed of 2–3 focused modules versus larger, comprehensive documentation-style Skills within the SkillsBench experiments. (Details on exact sample counts per granularity condition are reported in the paper's Skill-design analyses.)
Complementary occupations that support, deploy, and regulate AI will be created.
Qualitative sectoral analysis and theoretical reasoning about complementarities; no explicit empirical enumeration or occupational survey sample presented.
Productivity-induced demand expansion (cheaper goods/services) will generate additional employment and new services.
Standard macroeconomic/consumer-demand theory applied to productivity gains from AI; argument provided by theoretical synthesis, without reported empirical elasticity estimates or sample-based quantification.
Indirect employment effects will arise from new industries and platform ecosystems enabled by AI.
Theoretical/qualitative argument and sectoral examples (synthesis); the paper does not report empirical measurement of the magnitude or sample-based evidence of such industry creation.
AI complements labor by raising productivity and increasing demand for high-skill, technology-intensive roles (developers, data scientists, AI specialists, etc.).
Complementarity arguments within labor economics theory and sectoral analysis; no new empirical counts or representative labor market sample described in the paper.
Policy interventions (lifelong learning, reskilling programs, active labor-market policies, social protection) are necessary to manage transitional unemployment and distributional effects.
Policy prescriptions based on theoretical framework and synthesis of prior policy evaluations; the paper recommends these approaches but does not present new impact estimates.
AI indirectly creates employment via platform ecosystems, new industries, and productivity-induced demand expansion.
Economic theory on demand-driven employment effects and literature synthesis of platform and productivity spillovers; cross-sectoral discussion rather than a new empirical estimate.
AI directly creates new occupations and tasks related to AI development, deployment, maintenance, and oversight.
Empirical and conceptual synthesis noting observed emergence of AI-specific roles in labor markets and task-based theory of job creation; no single quantified sample provided.
AI complements high-skill, technology-intensive roles, increasing demand for advanced cognitive, creative, and supervisory skills.
Task-complementarity argument from theory and empirical patterns in literature where technology raises demand for skilled workers; cross-sectoral examples cited conceptually.
Adoption of AI in accounting can raise firm-level productivity via faster close cycles, better control, and improved forecasting, potentially affecting profitability and investment decisions.
Theoretical and literature-based claim; the paper suggests mechanisms but does not present a specified empirical estimation in the abstract.
The paper advocates a complementary (augmenting) view of AI in accounting instead of a pure substitution view.
Argumentative conclusion based on synthesis of reviewed studies and theoretical considerations presented in the paper.
AI adoption changes accountants' roles from data entry and routine processing to analysis, interpretation, and strategic decision support.
Inferred from qualitative literature, surveys, and case studies discussed in the paper rather than from a specified empirical identification strategy.
Documented benefits of AI in accounting include increased efficiency, fewer manual errors, faster close cycles, improved report accuracy, and better fraud/irregularity detection.
Reported from literature and industry reports/case examples cited by the paper; the paper does not provide detailed sample sizes or econometric estimates in the abstract.
AI complements accountants rather than substituting them, raising productivity and shifting accountants' focus toward strategic financial management.
Argument based on literature review and qualitative interpretation of workflow changes (surveys/case studies likely); no randomized or quasi-experimental evidence reported in the abstract.
AI technologies (machine learning, robotic process automation, and advanced analytics) are materially improving accounting by automating repetitive tasks, reducing errors, detecting fraud, and providing predictive insights.
Stated as the paper's main finding and supported by cited literature and industry/case examples; the abstract does not specify an empirical design or sample for causal estimation.
Serious-game DSTs can reduce informational frictions by making model outputs (including AI-based recommendations) more interpretable and actionable, lowering barriers to adoption and improving translation of technical advice into economic behavior.
Conceptual synthesis and illustrative practice examples where visualization and interactivity improved understanding; empirical evidence is limited to qualitative user reports and small demonstrations.