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
Adoption of GenAI is constrained by multiple interrelated challenges.
Cross-paper synthesis from the systematic review of 28 studies identifying recurring barriers and constraints reported in the literature.
Ongoing issues remain such as data access, model transparency, ethical concerns, and the varying relevance across Global North and Global South contexts.
Critical synthesis within the review drawing on discussions and critiques in the literature about barriers and ethical challenges; based on reported limitations and regional comparisons in reviewed studies (no numerical breakdown provided).
Human judgment is constrained by bounded rationality, cognitive biases, and information-processing limitations.
Cited as established findings from prior research across decision sciences and related fields (extensive literature evidence referenced; no new empirical data in this paper's abstract).
There are significantly negative spatial spillover effects between digital–real integration and New Quality Productive Forces (i.e., each variable has negative spillover impacts on the other across regions).
Spatial spillover coefficients estimated in the GS3SLS spatial simultaneous equations model using panel data for 30 provinces (2011–2022) are reported as statistically significant and negative.
AI substitutes many routine tasks, including both manual and cognitive/rule-based activities, disproportionately affecting middle-skill occupations.
Task-based substitution reasoning within SBTC framework and cross-sectoral task analysis. The paper provides conceptual synthesis rather than presenting new microdata or quantified task-level estimates.
Key implementation challenges include data quality and integration, model interpretability, cybersecurity and privacy, regulatory/compliance uncertainty, skills gaps among accounting professionals, and implementation costs.
Identified by the paper through literature review and practitioner reports; these are presented as recurring barriers rather than quantified with a specific sample.
Many studies on serious-game DSTs are small-scale or experimental, and long-term impact data on behavioral change and emissions outcomes are sparse, limiting generalizability.
Review of the literature summarized in the chapter showing predominance of case studies, prototypes, and short-term evaluations rather than longitudinal or large-sample studies.
Ensuring scientific validity of game models, scaling co-design processes, measuring real-world behavioral change, and aligning incentives (policy/subsidies, markets) are remaining challenges to using serious games for DST uptake.
Chapter discussion of limitations and gaps identified in the reviewed literature; absence or sparsity of long-term validation studies and large-scale co-design implementations documented in existing research.
Current uptake of DSTs for net zero remains limited because of issues of trust, usability, lack of evidence linking actions to farm profitability, and poor integration into farmer workflows.
Literature synthesis, qualitative interviews and surveys, case studies documenting low adoption and barriers; multiple practice reports and studies cited in the chapter. Many studies report limited or uneven adoption across contexts.
Regulatory uncertainty around blockchain/DeFi for corporate finance and cross-border data rules is a material risk to adoption.
Paper notes regulatory uncertainty as a risk; no jurisdictional legal analysis or compliance case studies provided in the summary.
Cybersecurity and data-privacy concerns arise from cloud provider centralization versus blockchain transparency.
Paper highlights this trade-off in its challenges section; discussion-based evidence rather than quantified security assessment in the summary.
Integration complexity with legacy ERPs and heterogeneous vendor ecosystems is a significant implementation challenge.
Paper lists this as a challenge/limitation based on pilot experience and analysis. No quantified measure of integration effort is provided in the summary.
EPC projects feature milestone-based payments, complex stakeholder flows, and large working-capital needs that strain traditional on-premise ERPs.
Problem context statement presented in the paper; consistent with commonly reported characteristics of EPC projects. The summary does not cite empirical industry-wide data.
If deployed without mitigation, GenAI CDS risks widening disparities by performing worse on underrepresented groups or being unequally distributed across resource-rich versus resource-poor settings.
Fairness literature, subgroup performance concerns, and distributional risk analysis cited in the paper; direct empirical demonstrations of widened disparities due to GenAI CDS are limited in the literature per the paper.
Limited public datasets and vendor lock-in constrain independent reproducible evaluations and audits of current generative models in healthcare.
Observation and policy analysis in the paper noting scarcity of public clinical datasets for state-of-the-art models and proprietary constraints; no dataset counts provided.
GenAI CDS creates data privacy and security risks because of high-value medical data and use of external cloud services.
Known cybersecurity risks and documented incidents in health IT; the paper cites the general risk context rather than specific breach sample counts tied to GenAI deployments.
GenAI CDS can amplify bias and inequities if training data underrepresent groups or reflect historical disparities.
Fairness and robustness audit literature and subgroup performance analyses referenced in the paper; specific empirical demonstrations for contemporary GenAI CDS are limited and sample sizes not given.
GenAI CDS systems hallucinate and can produce incorrect but plausible recommendations, which can cause patient harm if trusted unchecked.
Documented failure modes of generative models and examples from controlled evaluations; the paper references known hallucination behavior from model audits and case reports, though it does not quantify incidence rates or provide large-scale observational harm data.
There is limited long-term impact evidence and few system-level assessments of AI in developing-country agriculture.
Authors' methodological caveat based on the temporal scope and types of studies available in the >60-study review.
The evidence base is skewed toward pilots and high‑performer contexts; there is a lack of long‑panel, multi‑project longitudinal studies to validate typical returns and scalability.
Authors' assessment of evidence types in the 160 studies: mix of conceptual papers, case studies, pilots, and only limited larger empirical evaluations.
Substantial compute and resource requirements for training and inference concentrate capabilities among well‑resourced labs and firms.
Paper discusses large compute budgets for training/inference and states that performance scales with data, model size, and compute; it infers concentration of capabilities but provides no empirical market concentration measures.
Structure predictors depend on training data and exhibit biases; experimental validation remains necessary.
Paper notes dependence on training data biases and the need for experimental validation; references data sources (PDB, UniRef, metagenomic catalogs) but does not quantify bias magnitudes.
Current limitations include inaccurate prediction of multi‑chain complexes, flexible or rare conformational states, and limited prediction of dynamic ensembles.
Paper explicitly enumerates these limitations in the 'Ongoing limitations' section; no quantitative failure rates are given.
Traditional computational methods struggle without homologous templates or with complex folding/dynamics.
Paper discusses limitations of traditional computational methods, emphasizing dependence on homologous templates and difficulty with complex folding/dynamics; specific method comparisons or sample sizes are not provided.
Inequities in climate-AI systems appear across three development phases—Inputs, Process, and Outputs—creating multiple failure points where Global North advantages propagate into final products.
Conceptual framework developed from cross-disciplinary synthesis, literature review, and illustrative examples (Inputs → Process → Outputs mapping).
Foundation-model development and high-performance computing (HPC) capacity are overwhelmingly located in the Global North.
Descriptive mapping of global HPC infrastructure and foundation-model authorship described in the paper (infrastructure mapping and authorship analysis). No single quantitative sample size reported; evidence based on spatial mapping and documented locations of compute centers and model-development institutions.
Performance degrades when forecasted features are removed from the downstream regression model.
Ablation study results reported in the paper which compare full FutureBoosting against variants without TSFM-generated forecasted features using the same evaluation protocols.
When pipelines have cross-cutting ties, prices oscillate, allocation quality drops, and management becomes difficult.
Empirical simulation results from the ablation study: configurations with non-hierarchical, cross-cutting graph structures produced larger price volatility, frequent oscillations in price updates, and lower allocation value/throughput compared to hierarchical graphs (measured across many runs and random seeds within the 1,620-run experimental set).
On the 22 postdating (contamination-free) incidents, no agent achieved end-to-end exploitation success across all 110 agent–incident pairs evaluated.
Empirical evaluation of 110 agent–incident pairs reported in the study (end-to-end exploit attempts on the 22 incidents).
The original EVMbench had a data contamination risk because it relied on audit-contest data published before every evaluated model's release, which could have been seen during model training.
Timing relationship between the audit-contest dataset used by EVMbench and the release dates of evaluated models (dataset predated model releases).
The original EVMbench evaluation was narrow: it evaluated 14 agent configurations and most models were tested only with their vendor-provided scaffold.
Description of the original EVMbench experimental setup (number of agent configurations and scaffold usage) cited in this study.
There is a risk that NFD will overfit to individual practices and lead to privacy/IP leakage if crystallization is not carefully governed.
Limitations and risk analysis in the paper; conceptual argument and case study discussion raising privacy/IP concerns. No empirical incidence rates provided.
NFD requires sustained practitioner engagement and incentive alignment to be effective.
Limitations and discussion sections of the paper explicitly state this requirement; logical inference from method (human-in-the-loop commercialization and continual crystallization).
Limitations of the study include reliance on self-reported perceptions (subject to response and survivorship bias), lack of experimental/causal identification, potential non-representative sample, and cross-sectional design limiting inference about long-term productivity effects.
Authors' stated limitations in the paper summary.
A mathematical analysis bounds or relates expected performance loss of the surrogate to measurable distribution mismatch between the training parameter distribution (samples) and the target parameter distribution.
Theoretical derivations presented in the paper that relate performance loss to distribution mismatch; the summary states the analysis provides a measurable diagnostic for when retraining or reweighting is needed.
Current bottlenecks are disparate quantum and classical resources operating in isolation, causing manual job orchestration, inefficient scheduling, data-movement overheads, and slow iteration that limit productivity and algorithmic exploration.
Use-case-driven analysis and observations from early hybrid deployments and literature; systems design decomposition highlighting latency and data-staging requirements; no quantitative benchmark data.
If deployment value is the time-average for one agent, optimizing the usual expected-value objective can lead to poor real-world outcomes.
Reasoning plus the paper's illustrative example demonstrating policies with high expected reward but poor or highly variable realized time-average outcomes; theoretical exposition, no empirical dataset.
Optimizing the expected cumulative reward (ensemble average across trajectories) can be misleading when reward-generating dynamics are non-ergodic because the ensemble expectation does not generally equal the time-average experienced by a single deployed agent.
Theoretical argumentation and a constructive illustrative example in the paper showing divergence between ensemble expectation and single-trajectory time-average; no empirical sample; analysis-based evidence.
The evidence base presented is limited to a single SME pilot, so generalizability across sectors, firm sizes, and data regimes is untested and requires further research.
Explicit limitation noted in the paper and the fact that the pilot illustrated is a single case study (sample size = 1 SME pilot).
Tasks that are routine, repetitive, or pattern‑based (e.g., boilerplate coding, refactoring, unit test generation, some accessibility fixes) will be increasingly automated by AI.
Task‑level decomposition and examples of current automation capabilities (code generation, test suggestion tools); conceptual projection rather than empirical measurement.
Common barriers to effective RM implementation include siloed functions/weak coordination, limited resources or expertise, poor data quality/lack of metrics, and cultural resistance driven by short-term incentives.
Frequent identification of these barriers across the reviewed literature and practitioner sources synthesized via thematic analysis over the last ten years.
Hierarchy compresses: fewer organizational layers are needed for a given firm output as coordination costs fall.
Analytical proposition in the theoretical model and simulation results showing reduced number of layers under coordination compression.
Global median post-harvest losses are around 19.8% (FAO & Kaggle datasets).
Descriptive statistics cited from FAO and Kaggle datasets referenced in the paper for global context.
Higher measured GDP need not imply higher aggregate welfare: the private costs of the arms race can outweigh the market gains from increased output.
Welfare comparisons performed in the model showing parameter regions where private equilibrium raises GDP but reduces aggregate welfare once investment costs are included.
Because private incentives push agents toward tail outcomes, aggregate overinvestment occurs relative to the social optimum (the arms race is inefficient).
Welfare calculations and comparison of private vs social optima within the model; the paper shows private equilibrium investment exceeds the socially optimal investment given the externalities of the arms race.
Upfront costs for AI adoption are substantial: development, clinical validation, regulatory compliance, EHR integration, and ongoing monitoring.
Implementation and regulatory literature synthesized in the review documenting typical cost categories and reported expenditures for clinical AI projects.
Large language models (LLMs) suffer from hallucinations (fabricated facts), overconfidence, and unpredictable failure modes in open-ended tasks.
Technical papers and benchmarks on LLM factuality, calibration, and failure modes summarized in the review; empirical evaluations showing instances of fabricated outputs and calibration issues.
Contemporary AI systems have no capacity for physical examination, sensorimotor procedures, or direct patient-contact diagnostics.
Technical limitations of CNNs and LLMs described in literature (lack of embodiment, no sensorimotor capabilities) and absence of credible empirical demonstrations of safe autonomous physical clinical procedures in reviewed studies.
Current models exhibit poor out-of-distribution (OOD) generalization: performance degrades when inputs differ from training distributions.
Technical literature and robustness/domain-shift research reviewed in the paper documenting declines in model accuracy under domain shift and dataset changes.
High upfront costs and lack of tailored financing instruments are significant financial constraints on SME AI adoption.
Case studies, finance sector reports, and SME surveys cited in the review showing cost barriers and financing gaps; evidence descriptive rather than causal.