Evidence (16496 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
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 |
Traditional international marketing theories, constrained by static assumptions and linear logic, struggle to explain intelligent contexts.
Conclusion from the paper's systematic review and content analysis of core literature (2010–2025); no quantitative test or sample size reported in the summary.
Cost and lack of applicable use case are the most cited barriers to AI adoption, followed by expertise.
Survey question(s) on barriers to adoption in the Census Bureau survey in which respondents reported reasons for not adopting AI; ranking provided in the paper (cost, lack of use case, then expertise).
Intensity-weighted adoption is far lower than the 22.8 percent headline rate.
Survey-derived intensity-weighted measure of AI adoption constructed from the same Census Bureau survey (no numeric value reported in the excerpt).
Only 22.8 percent of plants report any AI use as of 2021.
Direct descriptive estimate from the Census Bureau survey of manufacturing establishments; year reported as 2021.
ID-centric ranking models fail to generalize in livestreaming recommendation due to the short-lived nature of live rooms and poorly learned item IDs.
Authors' assertion linking the cold-start item ID problem to poor generalization of ID-centric rankers (motivating claim). No specific experimental metrics or sample sizes cited in the excerpt.
A live room typically broadcasts for only tens of minutes, so its item ID remains poorly learned in a persistent cold-start state.
Authors' observational/operational claim about livestream characteristics stated in the paper (motivating problem statement). No sample size or quantitative backing provided in the excerpt.
There is a negativity asymmetry: negative histories induce 1.62x more bias than positive (paired per item; t = 13.46, p < 10^-39, n = 2,481).
Paired per-item comparison of bias induced by negative versus positive histories; reported multiplicative factor, t-statistic, p-value, and sample size n = 2,481.
Static benchmarks capture only part of how large language models behave in practice.
Argument supported by the paper's experimental design comparing static evaluations with a timed multi-phase Risk environment that includes repeated planning/execution loops and real-system constraints.
The de-coring and skill-demand changes are concentrated among low entry-threshold, small firms.
Abstract statement reporting heterogeneity: concentration of observed patterns among firms characterized as small and with low entry thresholds.
Both displacement and augmentation exposure are associated with a de-coring pattern: a shallower and more dispersed skill portfolio with within-category importance diverging from share movements.
Empirical description in abstract that both forms of exposure correlate with changes in portfolio depth and dispersion, and with divergence between within-category importance and category shares.
Displacement exposure is negatively associated with the routine cognitive skill share.
Empirical result stated in abstract: negative association between displacement exposure and routine cognitive share, identified using within-firm variation and the constructed exposure measures.
In deployed settings, the effects of AI systems on human agency, creativity, and institutional well-being emerge over time, shaped by repeated interaction, reuse, and integration into real-world workflows, and these dynamics are rarely visible through pre-deployment evaluation or isolated prompt–response analysis.
Argumentative observation based on conceptual reasoning; no empirical data or sample size reported.
The most significant barriers to AI adoption reported by entrepreneurs are human-centred—talent scarcity, organisational resistance, and change management—rather than technology or cost alone.
Theme 'Barriers and the Adoption Journey' from thematic analysis of interviews (n=16); interviewees repeatedly cited human-centred barriers (talent scarcity, resistance, change management) over purely technical/cost barriers.
Raw interaction logs are inherently noisy, contain trial-and-error and low information density, and are inefficient for direct model training.
Author assertion describing properties of raw interaction logs; no empirical quantification provided in the excerpt.
Static 'human data' is expensive to scale and bounded by the knowledge of its creators.
Author claim/argument in the paper's introduction; no empirical sample or quantitative test reported in the provided text.
Because contracts are negotiated by legal departments alone, many apparent legal disputes are incentive misalignment problems that only scientists at the table can correctly diagnose.
Argumentative claim presented in the paper (normative/diagnostic); no empirical study or sample provided in the excerpt.
These failures are not for scientific reasons, but because academics must publish while companies must protect models trained on proprietary data, and no standard contract framework resolves this tension.
The paper presents this as the causal explanation (analytical/argumentative claim); no empirical testing or sample reported in the provided text.
Industry-academia ML collaborations routinely fail to launch.
Asserted in the paper as an empirical observation/statement; no empirical methods, data, or sample size reported in the provided text (argument/anecdote).
People exhibit self-estimate miscalibration: on average they believe they are using AI less than they actually are.
Same three pre-registered user studies (combined N = 2691) comparing participants' self-reported AI use against observed/recorded AI use during tasks.
Low-information AI neither improves immediate performance nor preserves performance after AI assistance is removed, and is linked to weaker learning overall.
Within-study comparison of low-information AI assistance versus other conditions in the controlled logical reasoning task; immediate and post-AI performance measured (sample size not reported in abstract).
Greater AI usage is associated with weaker skill development: heavy AI users underperform relative to comparable peers, whereas light AI users perform similarly to matched users who do not use AI.
Controlled experiment using a logical reasoning task with on-demand AI assistance; comparison between heavy users, light users, and matched non-users reported in the study (sample size not stated in abstract).
Regulatory uncertainty and the absence of explicit legislation on digital data and artificial intelligence may leave the economic potential of these technologies unexplored while increasing market concentration, inequality, and the risk of personal information misuse.
Argued implications from the paper's theoretical model and comparative legal discussion; no empirical testing or quantified analysis provided.
The measurement bias understates substitution effects more than it understates augmentation effects.
Analytical argument and empirical evidence showing directional bias from measurement error that causes estimated substitution (labor displacement) effects to be more severely understated than augmentation (complementarity) effects.
Reweighting platform-based exposure measures to Bureau of Labor Statistics workforce shares attenuates estimates by 42 to 93 percent.
Reweighting exercise where exposure scores built from platform logs are reweighted to match BLS workforce shares and resulting employment estimates are compared; reported attenuation range of 42–93%.
Studies finding true synergy are scarce.
Authors' literature synthesis / meta-analytic overview claiming that few studies report combined human-AI performance exceeding both parties alone (no numerical count provided).
Genuine human-AI synergy—combined performance that exceeds what either party achieves alone—is uncommon.
Authors' synthesis of the literature and meta-analytic findings referenced in the paper indicating scarcity of studies showing combined performance > either alone (no specific counts or sample sizes given in the excerpt).
Current regulatory frameworks—designed for human-intermediated payments—are ill-equipped to address the dynamic and decentralised nature of agent-led transactions.
Regulatory and legal analysis asserted in the abstract (argument that existing frameworks are mismatched to agent-led payments).
The article identifies and categorises a range of technical, legal and societal risks, including cybersecurity vulnerabilities, liability gaps, regulatory non-compliance, and potential economic disruption.
Risk identification and categorisation presented in the paper (qualitative analysis and case studies referenced in the abstract). No quantitative risk measurement reported in the abstract.
Agentic systems show persistent failures in repository setup, dependency handling, permission gating, and hardware verification.
Issue-resolution benchmarks and hardware/RTL verification research synthesized in the paper (specific failure rates or sample sizes not provided in abstract).
Controlled studies report slowdowns in mature open-source work when using agentic/code-generation systems.
Controlled studies and trials cited in the paper (no sample sizes given in abstract).
Using a frontier model's system prompt to supply the procedure exposes proprietary procedures to third-party providers.
Author statement describing privacy/proprietary risk as a cost of the system-prompt approach (qualitative claim).
Using a frontier model's system prompt to supply the procedure requires a frontier model for every conversation.
Author statement describing operational/cost trade-offs associated with the system-prompt approach (qualitative claim).
Using a frontier model's system prompt to supply the procedure has costs: it consumes the context window.
Author statement referencing trade-offs identified alongside the Dennis et al. result; cost described qualitatively (context window consumption).
Low-wage workers on platforms perform supporting tasks—such as data annotation and content moderation—that underpin technological infrastructures.
Empirical grounding drawn from cited ethnographic, sociological and anthropological studies and mapping exercises discussed in the paper documenting the kinds of work performed on microtask platforms.
Artificial intelligence (AI) systems depend on invisible labor performed on microtask platforms.
Claim based on synthesis of sociological and anthropological studies cited in the paper mapping production networks and documenting microtask platform work (e.g., data labeling, content moderation) that supports AI.
Socio-technical imaginaries that forecast the displacement of humans from production accompany the technological developments of the Fourth Industrial Revolution.
Conceptual claim supported by literature review and theoretical framing in the paper describing historical and contemporary narratives around automation and the Fourth Industrial Revolution.
Emerging evidence indicates that algorithms often inherit and amplify the historical biases present in training data.
Literature claim in paper referencing 'emerging evidence' and empirical studies (2024–2026) — specific studies, methods, and sample sizes not included in excerpt.
Single-threshold scoring at conventional cutoffs misses the upper-tail cost; tail-inclusive scoring reverses the sign of the capability--accuracy relationship on the same outputs.
Empirical comparison in the paper between single-threshold scoring and tail-inclusive (continuous/unbounded) scoring on identical forecast outputs, showing sign reversal of the capability–accuracy relationship (numerical details not provided in excerpt).
A within-family study of Llama-3.1 shows that both model scale and post-training independently contribute to this effect.
Within-family empirical comparisons using Llama-3.1 variants examining effects of model scale and post-training (fine-tuning) on forecasting calibration (details and sample sizes not provided in excerpt).
A per-quantile decomposition shows the failure concentrates at the upper tail, which more capable models shift upward to track aggressive extrapolations of growth, while the lower tail stays put.
Per-quantile decomposition analyses of model predictive distributions reported in the paper, showing quantile-specific changes (specific quantitative results not given in excerpt).
The pattern replicates in real-world datasets on COVID-19, measles, housing markets, and hyperinflation.
Empirical replication reported on multiple real-world datasets (COVID-19, measles, housing markets, hyperinflation) presented in the paper (dataset sizes not provided in excerpt).
The pattern appears on ForecastBench-Sim (FBSim), a contamination-free, simulated-world benchmark we release, in forecasting synthetic SIR epidemics with a matched linear control.
Results on the authors' released simulated benchmark (ForecastBench-Sim) using synthetic SIR epidemic simulations and a matched linear-control experiment reported in the paper (specific number of simulations or runs not stated in excerpt).
We document inverse scaling in LLMs on forecasting problems whose underlying time series exhibit superlinear growth and tail risk of regime change ... more capable models produce worse distributional forecasts.
Empirical experiments reported in the paper comparing LLMs of varying capability on forecasting tasks with superlinear growth and regime-change tail risk; uses distributional forecast evaluation across models (no sample size reported in excerpt).
Regulatory frameworks that address only downstream applications leave the upstream concentration of infrastructural power largely intact.
Policy analysis and theoretical critique of regulatory approaches; argument based on the distinction between upstream infrastructure and downstream applications (qualitative).
Authority in AI systems is exercised not through formal jurisdiction but through infrastructural chokepoints and dependency pathways that precede and condition law.
Genealogical and infrastructural analysis; theoretical argument emphasizing chokepoints and dependency relations (qualitative).
Digital colonialism is distinct from surveillance capitalism: AI extends historical patterns of dispossession and epistemic domination beyond the commodification of individual behavior by embedding extractive and classificatory logics within data architectures, models, and standards.
Conceptual distinction developed via literature review, political-theoretical argumentation, and genealogical analysis (qualitative).
Contemporary biometric and algorithmic systems show continuities with colonial identification infrastructures.
Genealogical analysis and engagement with decolonial scholarship tracing historical continuities (qualitative, no quantitative sample).
AI systems deployed for identification, classification, and governance are the domains where sovereignty is most visibly reconfigured.
Analytic focus and genealogical tracing within the paper; literature review and conceptual examples of identification/classification systems (no quantitative sample reported).
AI constitutes a historically continuous yet technologically novel form of colonial power, shifting sovereignty from territorial authority toward infrastructural and algorithmic control (termed "infrastructural sovereignty").
Theoretical argument and genealogical analysis drawing on political theory and decolonial scholarship; conceptual synthesis presented in the paper (no empirical sample size reported).
The lack of prediction stability and predictability can lead to advertiser-perceivable problems such as repeatability issues, cold start, and under-exploration.
Stated as an intuitive/motivational claim in the paper linking instability to advertiser-facing problems; no empirical quantification provided in the excerpt.