Evidence (8807 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
Filtered →
Governance
7870 claims
Filter claims →
Human-AI Collaboration
7560 claims
Filter claims →
Org Design
4892 claims
Filter claims →
Innovation
4781 claims
Filter claims →
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 |
Productivity
Remove filter
Failures in engineering reasoning by AI systems may produce physically invalid yet superficially plausible solutions, posing risks for engineering education, scientific assistance, and technical decision-making.
Argumentative claim in the paper highlighting potential risks of reasoning failures in high-stakes engineering contexts (motivational/background statement).
Datasets are rarely standardized or shared.
Review synthesis and commentary across included studies and supplementary documents indicating limited data standardization and sharing.
Agents performed more weakly on a task requiring novel bioinformatics reasoning.
Reported ABC-Bench results indicating relatively lower agent scores on the task characterized by novel bioinformatics reasoning (authors' summary in the abstract).
This regulatory pressure creates a direct conflict between multi-stakeholder transparency and corporate data privacy.
Paper's conceptual argument describing a tension between transparency requirements and proprietary data protection; no empirical study provided.
Regulatory compliance demands have surpassed the capacity of manual corporate reporting.
Assertion in paper (conceptual observation about reporting capacity); no empirical measurement or sample size reported.
The convergence of the 2026 European Union Safe and Sustainable by Design (SSbD) framework, Corporate Sustainability Due Diligence Directive (CSDDD), and Carbon Border Adjustment Mechanism (CBAM) introduce a severe governance bottleneck for advanced semiconductor manufacturing facilities ("Smart Fabs").
Declarative claim in paper based on policy convergence analysis; no empirical dataset or sample size reported (conceptual/analytical argument).
Learning specialized simulator input languages can cost domain scientists hours to days.
Stated motivating claim in the paper (no experimental sample size or formal measurement reported in abstract).
In hyperscale cloud network infrastructure, traditional human-driven incident response cannot keep pace with the volume, velocity, and complexity of failures.
Stated as background/motivation in the paper; no quantitative data, sample size, or empirical comparison provided in the abstract.
Agents frequently overlook subtle yet critical details that are obvious to real human researchers.
Reported as a qualitative result/observation from the authors' experiments on AARRI-Bench; no numeric frequency or sample size provided in the excerpt.
Extensive experiments across frontier models and agentic systems reveal that even the best-performing configuration (Mini-SWE-Agent with Claude Opus 4.7) achieves only a 68.3% success rate on AARRI-Bench.
Empirical evaluation reported in the paper: experiments across multiple models/agentic systems; the excerpt reports the top configuration and its success rate. The excerpt does not state the number of tasks or sample size.
Despite their evolution from research assistants into autonomous research agents, these systems still exhibit significant limitations in field sensitivity, research ethics, and nuanced scientific judgment, and consequently remain unable to fully replace human researchers.
Asserted in the paper as a high-level observation and motivation; the excerpt does not provide quantified evidence or sample sizes for these limitations.
Current research on AI-supported conflict techniques has focused predominantly on Devil's Advocate (DA) and has neglected Dialectical Inquiry (DI).
Literature review / gap statement in the paper pointing to relative emphasis on DA in prior research and lack of work on DI.
Other methods, such as variants of prediction-powered inference, do not have the 'do no harm' guarantee.
Comparative methodological claim in the paper (abstract)—likely supported by theoretical discussion and comparisons in the main text.
Even a perfect non-proprietary-data report would be capped at 3.83 by B's coverage (i.e., B imposes an upper bound on non-proprietary informed decision-quality).
Analytic upper-bound calculation based on B's measured coverage on the curated gold record (exact derivation not provided in abstract).
GenAI usage significantly decreased creativity-relevant skills.
Experiment with 82 participants reported in the paper; authors report a statistically significant decrease in measures of creativity-relevant skills for participants using GenAI.
GenAI usage significantly decreased domain-relevant skills.
Experiment with 82 participants reported in the paper; authors report a statistically significant reduction in measures of domain-relevant skills for the GenAI condition.
GenAI usage significantly decreased intrinsic task motivation.
Randomized experiment reported in the paper with 82 participants; authors report a statistically significant decrease in intrinsic task motivation for participants using GenAI.
AI cannot yet refute economic theory on its own.
Main conclusion: based on the experiments (models failed to autonomously find true errors) and caveats about data contamination, the author concludes models are not yet capable of independently refuting economic-theory papers.
No model located a true error without substantial human guidance.
Author reports that in the experiments none of the models identified a real error autonomously; successful identifications required substantial human guidance.
Other models (Gemini, Refine, Claude) fared worse than ChatGPT Pro at these tasks.
Reported qualitative performance differences across the four models on the 4 papers; other models did not match ChatGPT Pro's performance.
Algeria lags behind peer countries on key indicators of digital infrastructure, human capital, and institutional frameworks as evidenced by World Bank (2022) and Oxford Insights indices.
Specific comparative claim based on the paper's use of World Bank (2022) indicators and Oxford Insights Government AI Readiness Index scores; the summary does not report numeric index values or sample sizes.
Findings reveal that Algeria exhibits significant lag in digital infrastructure, human capital, and institutional frameworks compared to peers (Morocco, Egypt, Turkey).
Result reported from the paper's comparative analysis using World Bank indicators, the Oxford Insights Government AI Readiness Index, and sector-specific studies comparing Algeria to Morocco, Egypt, and Turkey; specific quantitative comparisons not provided in the summary.
Existing research has significant shortcomings in terms of local empirical evidence, micro task mechanisms, and the impact of cutting-edge AI.
Critical appraisal in the paper's discussion of gaps identified through the systematic literature review; no single-study sample size.
Skill mismatch constitutes the core contradiction of labor force transformation.
Interpretive conclusion from the literature review asserting that mismatches between worker skills and job/task requirements are central to the labor-market effects of AI.
Despite the growing prevalence of human-AI decision making, the human-AI team’s decision performance often remains suboptimal, partially due to insufficient examination of humans’ own reasoning.
Motivating claim stated in the paper's introduction/abstract (appears to be based on broader literature and motivation rather than a new empirical test in this paper).
AACT also triggers higher cognitive load.
Reported measurement of cognitive load in the same house price prediction case study comparing AACT to traditional AI support (details and sample size not provided in abstract).
Current results show that the hardest tier remains far from saturated: across mainstream harness and backbone configurations, the average full pass rate is 2.6%.
Empirical evaluation results reported by the authors summarizing ALE benchmark performance across mainstream harness and backbone configurations (no further detail on exact configurations or task/sample counts in excerpt).
The gap is largely an evaluation problem: widely used benchmarks lack sustained performance measurement on real and economically valuable workflows.
Author argument presented in the paper; motivated by benchmarking limitations rather than an empirical test in the excerpt.
These gains have not translated into economically meaningful deployment across many professional domains.
Assertion in paper arguing a deployment gap between benchmark performance and real-world economic adoption; no quantitative deployment data provided in the excerpt.
Manual processing of these documents is time-consuming, inconsistent across reviewers, and unscalable.
Author claim / background motivation; no quantitative time or consistency metrics reported in the statement.
Actuaries rely primarily on structured numerical data for reserving and ratemaking, while valuable predictive information in unstructured text including medical records, adjuster notes, and call transcripts remains largely unused.
Author statement/observation in paper introduction; no empirical data or sample size provided to support prevalence claim.
The determining barrier to adoption observed in the two studied public-service units was not technological but training-related.
Qualitative analysis and intervention observations across two auditable case studies (SES/CONT in 2024 and UCI/SEDET in 2025); author-developed intervention and outcome changes used to support inference.
The adoption of generative artificial intelligence in the public sector has been treated predominantly as a technological problem, with the expectation that productivity gains would follow from more capable models.
Author statement / literature-positioning in paper (assertion about prevailing treatment); no quantitative data provided in text to support prevalence.
Government subsidies exert a negative moderating influence on the relationship between fintech development and corporate total factor productivity.
Moderation analysis reported in the paper on Chinese A-share listed manufacturing firms (2015–2023); paper states government subsidies weaken the positive fintech–TFP relationship (no numeric interaction estimates provided in the excerpt).
Neither setup speaks to the operationally most relevant case for marketing practice: building detailed individual twins from the pre-existing heterogeneous panel data that firms already accumulate through CRM systems, loyalty programs, and repeat surveys.
Author's argument / positioning (identifying a gap between existing published twins and practical marketing use cases).
In binary classification, no internal local composition can achieve complementarity under endpoint-monotone losses (including standard Bregman and many finite Bernoulli f-divergence losses); an analogous obstruction holds for multiclass aggregation under cross-entropy.
Impossibility results proved in the paper for binary classification under endpoint-monotone losses and for multiclass cross-entropy (formal mathematical proofs; no empirical sample).
Selector-based HAIs, including self- or AI-reliance, cannot achieve complementarity regardless of task, loss, or prediction quality.
Formal impossibility theorem proved within the paper's tree-based HAI formalism (mathematical proof; no empirical sample).
Reliable deployment faces three obstacles: (1) no large-scale evidence on how today's strongest model-and-harness combinations behave on end-to-end legal matters; (2) no agent architecture adapted to the legal vertical, only general-purpose harnesses; and (3) no mechanism for systems to learn from their own outcomes in a changing setting.
Authors' diagnosis / framing of gaps in the literature and practice motivating the study and system design (stated in the paper's introduction/abstract).
Strict matter completion stalls (does not improve) despite stronger models.
Harvey LAB empirical results (12,510 agent trajectories) report that while per-criterion accuracy increases, strict matter completion does not show corresponding improvement.
Even frontier agents remain far from completing matters in a single pass.
Results reported from the Harvey LAB empirical study (12,510 agent trajectories) comparing end-to-end matter completion across agent runs.
GPU utilization surged from 57% to 94% following the mining software's public release, displacing legitimate research workloads.
Measurement of GPU utilization levels before (57%) and after (94%) the public release of mining software; authors attribute displacement of research workloads to the utilization surge.
Budget GPU rental prices rose 38% following the mining software's public release.
Market measurements of budget GPU rental prices before and after the public release of the mining software, reporting a 38% increase.
The mining computation is commodity integer arithmetic portable to any hardware platform, offering no vendor lock-in.
Analysis of the computation showing it relies on basic integer arithmetic operations and is implementable across diverse hardware architectures.
Mining is unprofitable at current PRL prices ($0.21) across all GPU tiers (-54% to -72% ROI).
Profitability analysis/calculation across GPU tiers using current token price of $0.21; reported ROI range of -54% to -72%.
Statistical distribution checks are trivially defeated by adversarial Gaussian sampling.
Demonstration that adversarial Gaussian-sampled outputs pass the system's statistical distribution checks; experimental or analytic demonstration reported.
The verification protocol accepts random matrices by design, confirmed by 44 pool-accepted shares from our open-source miner across NVIDIA, AMD, CPU, and Apple Silicon hardware.
Protocol analysis showing acceptance criteria; empirical confirmation via 44 pool-accepted shares generated by an open-source miner run on multiple hardware architectures (44 accepted shares observed).
The dominant mining software contains no inference code.
Static/dynamic analysis of the dominant mining software deployed on the network showing absence of AI inference routines.
Pearl's 24 EH/s network -- representing approximately 320,000 GPU-equivalents consuming an estimated 112 MW -- produces zero useful AI computation.
Empirical measurement of Pearl network hashrate (24 EH/s) and mapping to GPU-equivalents and power consumption; analysis of miner code and verification showing no useful AI inference performed.
Across most risks, experts identified information, finance, and national security as the most vulnerable sectors.
Sector vulnerability ratings from the Delphi study (n=272); paper reports that information, finance, and national security sectors were most frequently judged vulnerable across risks.
AI users and the general public were judged the most vulnerable to these risks.
Delphi panel rated actor vulnerability; results reported in paper indicate AI users and general public received highest vulnerability ratings (n=272).