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 |
AI-based ESG systems are increasingly applied to extract deeper sustainability signals from corporate disclosures, reports and external data sources.
Descriptive claim supported by cited literature and examples of AI applications in ESG analytics within the paper's background (references to recent AI/ESG studies). The summary does not quantify the rate of adoption.
Regression analysis revealed that AI-derived ESG scores were more strongly associated with excess returns than traditional ESG metrics.
Regression models estimating the association between ESG scores (AI-derived vs traditional) and excess returns. The summary does not specify the regression specification, control variables, sample size, time horizon, or statistical significance measures.
AI-driven high-ESG portfolios demonstrated lower downside-risk exposure and smaller maximum drawdowns during market stress, indicating stronger resilience.
Downside-risk and maximum drawdown metrics computed for AI-driven high-ESG portfolios versus comparator portfolios during periods of market stress (portfolio-level analysis). Specific stress period(s), sample size and statistical tests are not provided in the summary.
AI-enhanced high-ESG portfolios achieved higher mean returns and superior Sharpe ratios than both AI-based low-ESG portfolios and traditionally rated ESG portfolios.
Portfolio-level performance comparison reported in the study (mean returns and Sharpe ratios calculated for portfolios constructed using AI-driven ESG indicators versus portfolios using conventional ESG ratings). The summary does not report sample size, time period, market coverage, rebalancing frequency, or statistical significance levels.
The study recommends multi-stakeholder collaborations (policymakers, financial institutions, entrepreneurs) to design inclusive AI solutions, bridge the digital skills gap, and foster an environment for equitable entrepreneurial growth.
Policy and practice recommendations drawn in the paper's conclusion based on empirical findings and interpretation of barriers.
Firms with high AI adoption reported superior decision-making quality compared to low adopters.
Survey comparisons of decision-making quality measures between AI adoption groups in the questionnaire data (N=400), reported as superior for high adopters.
Firms with high AI adoption reported significantly higher financial literacy scores compared to low adopters.
Comparison of financial literacy scores between high and low AI adoption groups derived from the structured questionnaire responses (sample N=400); described as 'significantly higher' in the paper.
There is a positive correlation between the level of AI adoption and key business outcomes.
Survey-based correlational analysis reported in the paper linking self-reported AI adoption level to business outcome measures across the sample of 400 respondents.
Upstream foundation model providers offering fine-tuning and inference services to downstream firms creates a co-creation dynamic that enhances model quality when downstream firms fine-tune models with proprietary data.
Conceptual claim and theoretical framing in the paper: description of an AI supply-chain interaction where providers supply compute/inference and downstream firms fine-tune with proprietary data; the paper posits this co-creation improves model quality as part of the motivating narrative.
Under pro-price-competitive policies or compute subsidies, the provider and downstream firms can achieve higher profits along with greater consumer surplus (a win-win-win outcome).
Equilibrium profit comparisons in the game-theoretic model showing that, in the parameter regions where these policies raise consumer surplus, both the upstream provider's profit and downstream firms' profits also increase relative to the baseline.
Policies that promote quality competition in downstream markets always improve consumer surplus.
Model outcomes: comparative-static and equilibrium results show that strengthening downstream quality competition monotonically increases consumer surplus across the parameter space considered in the paper.
Pro-price-competitive policies and compute subsidies are complementary: each is effective in different cost regimes and together can cover more cases.
Analytical results from the game-theoretic model showing complementary effectiveness across varying compute/preprocessing cost parameters (comparative statics demonstrating non-overlapping regions of effectiveness).
New employment opportunities are emerging in AI-complementary occupations.
Findings from job-posting analyses and other empirical studies summarized in the paper that identify growth in AI-complementary job listings and roles (specific metrics not provided in excerpt).
Generative AI (GenAI), particularly tools such as ChatGPT and Gemini, has rapidly transformed the global technological landscape.
Qualitative/observational statement in paper citing the rapid public adoption of GenAI tools since late 2022; no specific empirical sample sizes reported in the text provided.
Holistic AI integration across supply chain functions yields greater performance benefits than isolated technological implementations.
Comparative analysis using survey and statistical methods (correlation/regression) on data from supply chain professionals; the summary reports superior outcomes for integrated (ecosystem-level) AI adoption versus isolated implementations, but does not provide the comparative metrics or sample breakdown.
AI-enabled performance management plays a mediating role that strengthens the linkage between strategic planning and operational outcomes.
Mediation analysis conducted on survey data from supply chain professionals (manufacturing and service sectors); the summary indicates a mediating effect of performance management but provides no mediation statistics (indirect effect size, confidence intervals) or sample size.
AI-enabled execution emerged as the strongest direct predictor of supply chain performance.
Regression analysis from the quantitative survey of supply chain professionals comparing AI-enabled planning, execution, and performance management as predictors of supply chain performance; specific coefficients, significance levels, and sample size are not reported in the excerpt.
AI integration significantly improved overall supply chain performance.
Quantitative study using data collected from supply chain professionals and analyzed with reliability testing, correlation, and regression methods; the provided text does not include sample size, p-values, or effect magnitudes.
AI integration significantly improved responsiveness (supply chain responsiveness).
Survey data from supply chain professionals across manufacturing and service sectors analyzed via correlation and regression analyses; the summary does not state sample size or numerical results.
AI integration significantly improved operational efficiency.
Quantitative survey of supply chain professionals (manufacturing and service sectors) with statistical analyses including reliability testing, correlation, and regression; specific sample size and effect sizes not provided in the summary.
AI integration significantly improved forecasting accuracy.
Quantitative survey of supply chain professionals (manufacturing and service sectors) analyzed using reliability testing and correlational/regression statistics; exact sample size and effect size not reported in the provided text.
The approach provides a closed-form mapping from information primitives to equilibrium outcomes.
Paper presents explicit formulas relating primitives (noise processes/Brownian shocks, signal-generation parameters, payoff matrices) to equilibrium objects (strategies, beliefs kernels, information wedge, and resulting payoffs).
The characterization yields an explicit information wedge V^i_t — a deterministic Volterra process — that prices the marginal value of shifting opponents' posteriors.
Derived closed-form expression in the paper: defines V^i_t as a deterministic Volterra-type process arising from the fixed-point solution; interprets it as the marginal value (price) of changing opponents' posterior beliefs.
This collapse reduces Nash equilibrium to a deterministic fixed point with no truncation and no large-population limit required.
Analytical reduction presented in the paper: after representing beliefs by deterministic kernels, the equilibrium conditions are expressed as a deterministic fixed-point problem solvable without approximations like truncating the belief hierarchy or taking N→∞.
Conditioning on primitive Brownian shocks (a dynamic analogue of Harsanyi's common-prior construction) collapses the infinite belief hierarchy onto deterministic two-time kernels.
Methodological derivation in the paper: change of conditioning variable from physical state to primitive Brownian shocks yields deterministic two-time kernel representation of agents' beliefs (i.e., belief dynamics become deterministic kernels rather than stochastic hierarchies).
We provide the first exact equilibrium characterization of finite-player continuous-time LQG games with endogenous signals.
Paper's constructive solution: derives an exact equilibrium by conditioning on primitive Brownian shocks and mapping the game to a deterministic fixed point; applies to finite number of players in continuous time with linear-quadratic-Gaussian structure and signals that depend on controls.
AI and Big Data enable proactive risk management strategies that contribute to lowering market uncertainty.
Qualitative case studies and quantitative analysis indicating firms used AI/Big Data for proactive risk management; details on number of cases or measurement of 'proactive risk management' not provided in the summary.
The reduction in market uncertainty occurs through enhanced predictive modeling capabilities enabled by AI and Big Data.
Findings reported in the paper attributing improved predictive modeling (from quantitative analysis and case-study observations) as a mechanism for uncertainty reduction (no specific metrics or effect sizes provided in the summary).
Strategic integration of AI and Big Data can significantly reduce market uncertainty during periods of economic turbulence.
Mixed-methods study combining quantitative analysis of market data and qualitative case studies of firms implementing AI and Big Data solutions (specific sample size and statistical details not provided in the summary).
The expanding use of AI is reshaping agricultural production systems and has emerged as a key driver of high-quality development in the sector.
Synthesis and interpretation of the paper’s empirical findings (significant AI effects on TFP, identified channels, and heterogeneous impacts) based on the listed-firm panel analysis.
Productivity gains from AI are more pronounced in regions facing higher natural risks.
Heterogeneity analysis in the paper that compares regions with differing natural-risk levels and finds stronger AI–TFP effects in higher-risk regions using the 2007–2023 panel of listed agricultural firms.
Productivity gains from AI are more pronounced among firms in their growth stage.
Heterogeneity analysis in the paper that splits the sample by firm life-cycle/stage and reports larger AI-associated TFP effects for firms classified as being in the growth stage.
AI fosters productivity growth by facilitating inter-firm resource sharing.
Mechanism analysis in the paper indicating a significant association between AI adoption and measures of inter-firm resource sharing, which in turn are associated with higher TFP in the panel sample.
AI fosters productivity growth mainly by optimizing labor structures.
Mechanism analysis reported in the paper linking AI adoption to measures of labor-structure optimization and finding that this channel is a significant contributor to TFP gains in the sample of listed agricultural firms.
The adoption of AI improves factor allocation efficiency and constitutes a critical economic foundation for efficiency-driven sustainable growth in agriculture by optimizing resource utilization and strengthening risk-management capacity.
Conceptual framing supported by the paper's empirical findings (panel data on agricultural firms listed on Shanghai and Shenzhen A-share markets, 2007–2023) that show AI raises total factor productivity (TFP) and stronger effects in higher natural-risk regions (interpreted as improved risk management).
A set of emerging methodological approaches—prompt-based experiments, synthetic population sampling, comparative-historical modeling, and ablation studies—map onto familiar social-scientific designs while operating at unprecedented scale.
Survey and mapping of methodological techniques presented in the paper; claim is a conceptual synthesis rather than a report of a particular dataset or experiment in the provided text.
Instruct-only and modular adaptation regimes constitute pragmatic compromises for behavioral research because they can preserve pretrained cultural regularities while allowing researchers to elicit targeted behaviors.
Methodological recommendation derived from comparing adaptation regimes (conceptual argument / review of adaptation strategies); no empirical comparison or sample sizes provided in the excerpt.
Trained on unprecedented volumes of human-produced text, LLMs encode large-scale regularities in how people argue, justify, narrate, and negotiate norms across social domains.
Inference based on known pretraining procedures for LLMs and the paper's theoretical account; no specific corpus size or empirical validation reported in the provided text.
There is a third, emerging ambition in AI research: using large language models (LLMs) as scientific instruments for studying human behavior, culture, and moral reasoning.
Argumentative proposal grounded in the paper's conceptual analysis and review of existing methodological work; framed as an emerging research program rather than demonstrated empirical fact.
Vocational graduates who undergo strong work-based training demonstrate competitive and sometimes superior long-term employment trajectories compared with other pathways.
Comparative empirical studies and secondary analyses referenced in the paper that link work-based vocational training to favorable long-term outcomes (the summary does not provide exact studies, effect sizes, or sample sizes).
Higher education graduates generally experience favorable employment outcomes.
Synthesis of prior empirical studies and secondary labor-market indicators cited in the paper indicating better employment prospects for higher education graduates (no specific effect sizes or sample n given in the summary).
There has been substantial growth in higher education attainment across the countries examined.
Descriptive results drawn from secondary data and comparative empirical studies documenting trends in higher education enrollment and attainment (paper does not report specific country list or sample sizes in the summary).
The findings provide practical guidance for entrepreneurs on building adaptive, AI-integrated organizations by redefining hiring, decision processes, and learning practices.
Prescriptive recommendations derived from the interview analysis and observed patterns in the sample of entrepreneurs (qualitative grounding; specific examples or measured impacts not provided in the excerpt).
Hybrid decision architectures have emerged: startup-specific configurations where algorithmic reasoning and human judgment recursively interact to shape decisions, roles and routines.
Thematic synthesis of interview data identifying recurring patterns of human–AI recursive interaction in decision-related practices across the studied startups (qualitative evidence; no quantitative counts reported).
Entrepreneurs who founded startups after ChatGPT's release integrated AI into their post-release ventures.
Direct accounts from the subset of interviewees who founded startups after ChatGPT's release describing AI incorporation in those ventures (qualitative interview evidence; sample details not given).
AI is becoming embedded in the architecture of startups rather than serving only as a task-automation tool.
Interview data and qualitative analysis identifying patterns of AI integration across startup roles, routines and structures (derived from the same semi-structured interview sample; exact N not provided).
Facilitated access to AI following the release of ChatGPT is transforming how startups organize and make decisions.
Qualitative study using semi-structured interviews with entrepreneurs who founded startups both before and after ChatGPT's release and who integrated AI into their post-release ventures; thematic/qualitative analysis of interview data. (Sample size not reported in the provided excerpt.)
Perceived autonomy enhances the positive effect of perceived algorithmic standardized guidance on riders' outcomes (i.e., strengthens the beneficial impact on mental health and reduction in risky riding via work pressure).
Interaction/moderation effects tested via SEM on 466 Chinese food delivery riders; results reported that perceived autonomy amplifies the beneficial pathways from standardized guidance.
Perceived autonomy mitigates (buffers) the negative effect of perceived algorithmic tracking evaluation on risky riding behavior (i.e., reduces the tendency toward risky riding driven by tracking evaluation via work pressure).
Moderation analysis within SEM using sample of 466 Chinese delivery riders with bootstrapped tests for interaction effects between tracking evaluation and perceived autonomy.
Perceived autonomy mitigates (buffers) the negative effect of perceived algorithmic tracking evaluation on riders' outcomes (i.e., reduces the adverse impact on mental health and risky riding via work pressure).
Moderation tested in SEM on data from 466 Chinese food delivery riders; interaction effects reported indicating perceived autonomy weakens the negative pathways from tracking evaluation.