Evidence (4781 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
9875 claims
Filter claims →
Productivity
8807 claims
Filter claims →
Governance
7870 claims
Filter claims →
Human-AI Collaboration
7560 claims
Filter claims →
Org Design
4892 claims
Filter claims →
Innovation
4781 claims
Filtered →
Labor Markets
4004 claims
Filter claims →
Skills & Training
3308 claims
Filter claims →
Inequality
2332 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 | 870 | 233 | 116 | 1066 | 2363 |
| Governance & Regulation | 976 | 451 | 218 | 133 | 1809 |
| Organizational Efficiency | 949 | 224 | 144 | 88 | 1416 |
| Technology Adoption Rate | 764 | 287 | 141 | 122 | 1325 |
| Research Productivity | 501 | 152 | 74 | 362 | 1101 |
| Output Quality | 542 | 216 | 69 | 69 | 896 |
| Decision Quality | 387 | 198 | 94 | 54 | 740 |
| Firm Productivity | 513 | 67 | 101 | 27 | 714 |
| AI Safety & Ethics | 249 | 303 | 73 | 36 | 667 |
| Market Structure | 190 | 192 | 134 | 27 | 548 |
| Task Allocation | 243 | 77 | 91 | 36 | 452 |
| Innovation Output | 291 | 33 | 55 | 20 | 401 |
| Skill Acquisition | 206 | 72 | 65 | 21 | 364 |
| Employment Level | 133 | 63 | 115 | 22 | 335 |
| Fiscal & Macroeconomic | 153 | 79 | 52 | 32 | 323 |
| Task Completion Time | 206 | 37 | 12 | 15 | 272 |
| Firm Revenue | 179 | 52 | 29 | 5 | 266 |
| Consumer Welfare | 130 | 76 | 47 | 13 | 266 |
| Inequality Measures | 48 | 137 | 51 | 6 | 242 |
| Worker Satisfaction | 101 | 81 | 25 | 13 | 220 |
| Error Rate | 84 | 110 | 11 | 5 | 210 |
| Wages & Compensation | 98 | 47 | 30 | 10 | 185 |
| Regulatory Compliance | 88 | 73 | 17 | 7 | 185 |
| Automation Exposure | 66 | 64 | 33 | 16 | 182 |
| Team Performance | 105 | 29 | 30 | 11 | 176 |
| Training Effectiveness | 109 | 22 | 14 | 21 | 168 |
| Developer Productivity | 114 | 21 | 14 | 8 | 158 |
| Job Displacement | 12 | 90 | 24 | 1 | 127 |
| Hiring & Recruitment | 57 | 9 | 9 | 5 | 80 |
| Skill Obsolescence | 6 | 56 | 9 | 1 | 72 |
| Social Protection | 43 | 17 | 8 | 2 | 70 |
| Creative Output | 35 | 21 | 9 | 4 | 70 |
| Labor Share of Income | 18 | 21 | 17 | 1 | 57 |
| Worker Turnover | 15 | 16 | — | 4 | 35 |
| Industry | — | — | — | 1 | 1 |
Innovation
Remove filter
There is an evolution from task‑specific automation toward systems that incorporate ecological domain knowledge, robustness to ecological heterogeneity, and evaluation on applied conservation objectives.
Evolution-of-approach observation based on trends reported across the papers in the collection (comparative description of earlier vs newer works).
AI-adopting firms exhibit higher productivity and higher market value after adoption.
Estimates showing increases in productivity (e.g., TFP measures) and market-value measures (e.g., market capitalization or Tobin's Q) for adopters relative to nonadopters using the stacked diff-in-diff design.
Post-adoption patents include more claims (i.e., are broader/more detailed) for AI-adopting firms.
Patent-level analysis using number of claims per patent as outcome in the stacked diff-in-diff framework.
Peer-driven digitalization matters not only for firm-level resilience but also for long-term sustainable competitiveness in manufacturing ecosystems.
Synthesis and implication drawn from empirical results (peer effects, mediators, and heterogeneity) using Chinese manufacturing A-share firm data from 2013–2022.
The adoption of AI technologies offers a scalable, resilient strategy for modernizing water management and promoting agricultural sustainability in Iraq.
Authors' conclusion based on single-site field experiments, economic and sustainability analyses, and reported robustness in sensitivity analyses; scalability claim is inferential and extends beyond the experimental site.
Information Systems (IS) research is critical for achieving joint optimization of technical capabilities and social systems in the context of GenAI.
Authors' argumentative positioning based on the socio-technical interpretation of the review; proposed role for IS scholarship rather than empirical test within the review.
Policy tools such as bans on sale of certain sensitive data, fiduciary duties for data holders, privacy-by-default, and collective data governance (data trusts, regulated commons) are appropriate levers to limit harms from data commodification.
Prescriptive policy argument based on normative analysis and literature on governance alternatives; recommendations are not evaluated using empirical policy impact studies within the paper.
Policy-relevant implication (extrapolated): diffusion of AI tools among small firms will likely follow social-network channels and be shaped by peer benchmarking, so aggregate incentives may underperform unless they leverage local networks and trusted intermediaries.
Inference and policy implication drawn from main empirical findings on the primacy of social networks and peer effects for entrepreneurial behavior; not directly measured in the dataset for AI-specific adoption.
China exhibits strong long-run integration between core AI and AI-enhanced robotics and a significant contribution from universities and the public sector to patenting.
Country-level decomposition showing (a) a stronger statistical long-run relationship between Chinese core AI and AI-enhanced robotics patent series and (b) actor-type decomposition of Chinese patent filings indicating relatively high shares from universities/public-sector actors (patents 1980–2019). Exact counts/shares not provided in the summary.
Policymakers should combine competition policy, data governance, retraining/redistribution measures, and targeted R&D/green-AI incentives to manage the transition and preserve broad-based demand.
Normative policy recommendation derived from the integrated theoretical framework and literature synthesis; not empirically validated in the paper.
Economically, there will be demand for 'temporal-quality' products: neurotech and AI services that explicitly measure, preserve, or enhance experienced temporality (presence, flow, meaning), representing a distinct market segment.
Speculative market implication derived from conceptual argument and literature on consumer preferences; no market data or empirical demand studies provided.
Industrial automation (industrial robots) can be an effective component of green development strategies when paired with finance and policy instruments.
Inference drawn from core empirical results: (1) IR reduces IWE; (2) effects are stronger with greater financial depth and policy support; combined evidence suggests complementarity between automation, finance, and policy.
Regulators must balance innovation with consumer protection by mandating model auditability, fairness testing, and interoperable data standards to prevent systemic and algorithmic risks.
Policy recommendation derived from synthesis of algorithmic risk, model opacity, and fintech market dynamics; based on normative analysis and best‑practice proposals rather than empirical testing.
Policymakers and firms should prioritize upskilling, standards for model provenance and IP, liability frameworks for AI-generated code, and improved measurement to track AI-driven productivity changes.
Policy recommendations derived from identified risks, barriers, and implications in the literature review and practitioner survey; not an empirically tested intervention.
DPS gives organizations with limited compute budgets a cost advantage for RL finetuning, potentially democratizing access to effective finetuning or shifting demand across cloud compute products.
Economic implications discussed qualitatively by the authors based on reduced rollout requirements; this is a projection rather than an experimental result.
AI-enabled analytics can increase firm-level decision value and productivity—improving capital allocation, speeding risk mitigation, and raising profitability in affected firms and sectors.
Economic implication argued by the paper using theoretical reasoning; no firm-level empirical estimates, sample sizes, or causal identification strategies are reported (paper suggests methods like A/B tests or causal inference for future study).
Policy interventions such as taxes, subsidies, regulation, coordination mechanisms, or credit-market policies can mitigate the inefficient arms race and align private incentives with social welfare.
Normative policy discussion based on the model's identified externalities; the paper outlines candidate interventions (Pigovian taxes, subsidies, caps, coordination) but does not present empirical evaluation of policy efficacy.
The paper proposes user rights to opt out of nonessential generative-AI integration and to choose environmentally optimized models.
Policy design section and candidate legislative amendments recommending consumer opt-out and choice rights.
The paper proposes mandatory model-level transparency requirements covering inference energy consumption, standardized benchmarks, and disclosure of compute locations.
Policy design section: normative proposal and drafted candidate legislative amendments (paper authors’ recommendations).
Demand for AI tools, data infrastructure, and related services will grow; markets for research-focused AI products and scholarly-data platforms may expand.
Market implication noted in the paper. Based on projected trends and market signals rather than empirical market-sizing within the paper's abstract.
AI acts as a productivity multiplier that could raise the marginal returns to research inputs (time, funding), altering cost–benefit calculations for universities and funders.
Presented as an implication in the Implications for AI Economics section. This is a theoretical/economic projection rather than an empirically tested claim within the abstract; no empirical estimates or sample-based tests are provided.
Qualified digital endpoints and validated in silico markers create new markets and assets (digital biomarkers, validation services, certified datasets) with potential commercial value.
Market and policy implications discussed in the review; forward-looking argument based on regulatory pathways and observed demand for validation services (speculative, narrative).
Policy and firm responses should emphasize human-in-the-loop governance, training in evaluative/domain skills, data stewardship, and regulatory attention to IP, liability, competition, and robustness standards.
Normative recommendations drawn from the review's synthesis of empirical benefits and limitations; based on identified failure modes (bias, hallucination, variable quality) and economic risks (concentration, mismeasurement).
Cluster assignments can be used to define treatments in quasi-experimental designs (event-study or diff-in-diff) to estimate causal impacts of funding, regulation, or technology shocks on research direction and economic outcomes.
Recommended analytic approach in implications; described as a methodological possibility. No implemented causal analyses or empirical validation reported in summary.
Cluster assignments can be linked to downstream outcomes (patents, product introductions, industry adoption, labor demand) to study knowledge diffusion and productivity effects.
Suggested research direction in implications; described as a use-case for linking clusters to economic outcomes. No empirical demonstration in the paper summary.
Cluster assignments can be aggregated into topic-level growth indicators (counts, share of publications, citation-weighted output) to measure pace and direction of technological change.
Suggested use-case in implications for AI economics; described as a recommended practical step. No empirical implementation or validation in the provided summary.
The pipeline can be used to generate high-resolution topic maps and time series for AI research areas (emergence, growth, decline).
Proposed application described under implications for AI economics; no empirical demonstration of temporal time-series construction provided in the summary (pipeline described as cross-sectional in original methods).
More advanced NLP models (transformer-based encoders, finance-specific topic models, supervised sentiment classifiers) could improve signal quality over LDA and VADER.
Methodological discussion recommends more advanced models to potentially improve signals; this is presented as a likely improvement rather than empirically tested in the study.
Policy implication (inference from results): prioritizing digital infrastructure investment to pass critical thresholds will unlock stronger productivity and environmental gains than focusing solely on advanced digital services.
Inference drawn from panel threshold findings (infrastructure threshold) and observed complementarities; this is a policy recommendation rather than a direct empirical test.
The positive AGTFP gains from digital rural development are geographically heterogeneous and are concentrated in eastern provinces.
Regional heterogeneity analysis / sub-sample regressions across provinces showing larger estimated digitalization effects in eastern provinces compared with other regions.
Digital infrastructure exhibits a threshold effect: its positive impact on AGTFP becomes stronger once digital infrastructure passes a critical level.
Panel threshold model applied to the provincial panel (2012–2022) that identifies a statistically significant threshold in the infrastructure sub-index where marginal effects increase above that value.