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
AI outputs commonly contain errors and hallucinations: generated code can be incorrect, incomplete, or misleading.
Practitioner reports and observed interactions with AI tools documented in the Netlight qualitative study; specific instances and practitioner concerns described in the paper; no quantitative error rates provided.
Adaptive RL-driven campaigns complicate attribution and causal inference, so rigorous experimental designs (multi-armed trials, off-policy evaluation) are required for valid measurement.
Methodological claim in the implications section; supported by discussion of policy adaptivity and the need for specific evaluation techniques. No empirical demonstration provided.
The system raises privacy, fairness, and safety risks including data leakage, demographic bias in generated content, manipulative targeting, and potential regulatory non-compliance.
Risk assessment and red-team / audit practices described; paper cites known classes of ML deployment risks and recommends logs/audits. This is a conceptual identification rather than a quantified empirical finding.
Integration and engineering complexity (legacy systems, privacy/compliance pipelines, multi-channel platforms) is a persistent barrier to deployment.
Industry case studies and practitioner reports synthesized in the review documenting integration challenges; no systematic cost accounting or sample sizes presented.
Hallucinations and factual errors from generative AI can damage service quality and customer trust.
Documented failure cases and empirical reports from the literature aggregated by the review; no novel incident count or experimental data in this paper.
Generative AI is susceptible to social and representational biases and to factual errors or hallucinations; it lacks tacit, contextual domain expertise.
Documented examples in the literature of biased outputs and hallucinations; controlled evaluations and audits of model outputs; qualitative reports highlighting lack of tacit knowledge in domain-specific tasks.
The quality of AI-generated outputs is highly variable; models frequently produce mediocre but plausible-sounding content that requires human filtering.
Multiple user studies and qualitative reports documenting variability in output quality and the need for human curation; outcome measures include error rates, user-rated quality, and time spent vetting.
Factual errors and 'hallucinations' create misinformation risks and can produce costly service failures.
Model evaluation studies, incident case reports from deployments, and academic/industry analyses documenting hallucination rates and concrete failure examples.
Resource, compute, privacy, and deployment costs associated with CRAEA were not fully quantified in the paper.
Authors note that resource, compute, privacy, and deployment costs were not fully quantified; no cost analyses or benchmarks provided in the summary.
Evaluation was performed in an artificial/simulated home environment; therefore real-world transfer, robustness to noisy perception, and hardware constraints remain open questions.
Authors explicitly state evaluations occurred in a simulated home environment and acknowledge limits on real-world transfer and robustness. This is a stated limitation rather than an experimental finding.
High linguistic diversity in Africa makes building and evaluating multilingual language technologies more difficult and is a barrier to inclusive AI.
Synthesis of technical literature on NLP and multilingual model development and policy/NGO reports highlighting missing language resources; no original model evaluation reported.
Structural constraints—limited digital infrastructure, scarce and skewed data, and high linguistic diversity—complicate AI development, deployment and evaluation in African contexts.
Desk review of infrastructure and data availability reports and scholarly literature demonstrating gaps and their effects; no new measurement in this paper.
Privacy concerns, regulatory/compliance issues, biased or opaque models, and the need for change management and HR analytics capability building are significant risks constraining adoption.
Recurring risks and constraints reported by multiple included studies; summarized in the review's 'risks and constraints' theme.
Implementation of data-driven HRM faces recurring challenges: data quality, privacy and ethics, algorithmic bias, and deficiencies in skills and organizational readiness.
Commonly reported implementation issues across the 47 reviewed studies; extracted as a central theme in the review's thematic analysis.
Rapid skill obsolescence in AI necessitates frequent curriculum updates and responsive governance.
Identified as a risk: the paper notes AI skill change rates and recommends frequent updates and governance mechanisms. This aligns with general domain knowledge; the paper does not provide empirical measurement of obsolescence rates.
Aligning multiple standards is complex, posing a disadvantage and implementation risk.
Stated explicitly in Disadvantages/Risks: complexity of aligning multiple standards is listed. This is a reasoned observation in the paper rather than empirically demonstrated.
Implementing this framework requires significant resources and continuous updating.
Stated explicitly under Main Finding and Disadvantages/Risks; paper lists cost/time metrics to track (cost-per-curriculum, time-to-update) and highlights resource intensity. Support is descriptive/analytic rather than empirical.
Constraints and risks include model risk (overfitting, drift), algorithmic bias, privacy and data-sharing limits, legacy ERP complexity, interoperability challenges, and limited organizational readiness and skills.
Reviewed literature (empirical studies, technical evaluations, and standards) documenting technical and organizational failures, risk incidents, and common barriers to implementation.
Key audit/control weaknesses with respect to prompt fraud include lack of provenance for inputs/prompts and model outputs, inadequate access controls, and missing or ineffective monitoring and anomaly detection for AI outputs.
Qualitative control analysis and adaptation of established auditing principles to GenAI workflows; recommendations based on threat modeling rather than field data.
GenAI outputs can be tailored to mimic corporate styles, templates, and evidence artifacts (e.g., summaries, memos, audit trails), which increases their credibility to auditors, managers, or customers.
Illustrative examples and scenario mapping demonstrating templated output mimicry; no controlled experiments or corpus analysis provided.
Large language models produce fluent, human-like outputs that can mask falsehoods (hallucinations) as facts, making prompt fraud effective.
Well-established LLM behavior cited conceptually and supported in the paper by illustrative examples; no new empirical measurement in this article.
Prompt fraud does not require system intrusion, credential theft, or software exploits; it operates at the reasoning/language layer of large language models and therefore can be executed without technical breaches.
Logical/technical argumentation built from properties of LLMs and illustrative hypothetical attack chains; threat modeling rather than empirical attack logs.
Prompt fraud is a new, distinct fraud modality in which adversaries intentionally craft natural-language prompts (or manipulate prompt inputs) to steer generative AI outputs into producing misleading, fabricated, or compliance-evading artifacts that bypass traditional internal controls.
Conceptual definition presented by the paper based on threat taxonomy and scenario mapping; illustrated with case-style examples. No empirical incident dataset or prevalence statistics provided.
Potential limitations include limited methodological detail on case selection and measurement, possible selection and reporting bias from practitioner-sourced examples, and variable generalizability to small firms or highly regulated industries.
Authors' self-reported limitations in the Methods/Limitations section (qualitative assessment).
Prompt fraud exploits the natural-language interface of large language models (LLMs) to produce outputs that appear authoritative (reports, audit trails, explanations) without system intrusion, credential theft, or software exploitation.
Definition and threat-model description using conceptual examples and case vignettes; literature/regulatory review to position the threat relative to traditional fraud vectors.
Data privacy and cross-border compliance issues arise from using cloud and SECaaS, complicating legal compliance for firms.
Regulatory analyses and compliance reports; documented examples in case studies and industry guidance on cross-border data flows.
The cloud shared responsibility model creates potential ambiguities in liability between providers and customers.
Regulatory guidance, legal analyses, and documented post-incident case studies showing confusion over responsibilities.
China manages the openness–security trade-off through a centralized, developmentalist, techno‑sovereignty approach that privileges coordinated state direction and control.
Qualitative content analysis of national‑level policy texts: 18 Chinese policy documents coded across four analytical dimensions (coordination objectives, institutional actors, governance mechanisms, stakeholder legitimacy).
Automation and LLM-driven orchestration add opacity; errors in instrument control or analysis could propagate quickly, raising liability, insurance, and reproducibility concerns.
Analytical discussion of risks and analogies to automated systems in other domains; no incident-level empirical data from microscopy given.
Ethical and governance issues related to LLM-driven microscopy include accountability, reproducibility, access inequities, data privacy, and concentration of capabilities in large providers.
Policy-oriented synthesis and analogies to governance challenges observed in other AI deployments; no new empirical measurement in microscopy contexts.
Integration of LLMs with microscopes faces challenges including safety and reliability of instrument control, verification of scientific outputs, data provenance, and alignment with experimental constraints.
Analytical discussion based on known reliability and safety issues in automated systems and AI tool use; no empirical incident data from microscopy provided.
There is substantial uncertainty in economic forecasts due to possible scale-up failures, regulatory constraints, feedstock price volatility, and path‑dependent lock‑in effects.
Synthesis of technical failure modes, regulatory uncertainty, and sensitivity analyses reported in TEA/LCA literature and economic modeling sections of the review.
Regulatory and biosafety concerns (including environmental release risks and dual‑use issues) increase fixed costs and create entry barriers that shape industry structure and diffusion.
Policy and governance literature reviewed alongside technical case studies; citations of regulatory requirements, biosafety frameworks, and examples of compliance costs affecting project viability.
Engineering and economic challenges—scale‑up hurdles, process robustness, feedstock cost, and downstream purification—limit industrial deployment of many bio-based processes.
Case study TEA/LCA summaries and process reports in the review highlighting scale-up failures or increased costs at larger scales, purification complexity for low‑concentration products, and sensitivity to feedstock prices.
Technical biological limitations—metabolic burden, pathway crosstalk, byproduct formation, and genetic instability—remain major constraints on strain performance and scalability.
Multiple experimental reports and method papers cited in the review documenting decreased growth/productivity due to engineered pathway burden, unintended interactions between pathways, accumulation of byproducts, and genetic mutations during production runs.
The described pipeline is cross-sectional as presented and should be extended to dynamic models (temporal embeddings, change-point detection) for trend or causal analyses.
Method description in summary indicates cross-sectional pipeline; recommendation to extend for temporal/dynamic modeling when analyzing trends or causal effects.
LLMs and corpora may reflect disciplinary, geographic, or language biases; analyses should adjust or stratify accordingly.
Caveat explicitly stated in summary noting potential biases in LLMs and corpora; recommendation to adjust/stratify analyses.
Cluster reliability should be validated (e.g., bootstrap, perturbations) and automatic labels complemented with expert human validation for critical analyses.
Caveat and recommended validation steps provided in summary; suggests bootstrap/perturbation and manual validation as best practices. No empirical stability metrics provided in summary.
Results are sensitive to model and prompt choice; researchers should perform robustness checks across LLMs, soft prompts, and embedding models.
Caveat explicitly stated in the paper summary noting model and prompt sensitivity; recommended validation steps include robustness checks across models and prompts.
Measurement issues (task-based output measurement, attributing output changes to AI) and selection into early adoption bias estimated productivity gains upward.
Methodological robustness checks reported in the paper: task-based measures, bounding exercises, placebo tests, and analysis of pre-trends; discussions of selection on unobservables and potential upward bias.
Implementing the governed hyperautomation pattern raises upfront costs (governance tooling, monitoring, validation, compliance processes).
Economic and cost-structure discussion in the paper, based on qualitative reasoning and industry experience; no quantified cost estimates or sample-based cost analysis provided.
VIS inherits the limitations of input–output assumptions (fixed coefficients, no price feedbacks); AI-driven structural change may violate those assumptions, so dynamic extensions or calibration are needed.
Paper explicitly cautions about input–output model limitations and the need for dynamic extensions/calibration under structural/technological change.
Increases in K_T reduce employment levels in affected firms and industries even when aggregate productivity rises.
Panel econometric estimates at firm and industry levels relating K_T intensity to employment outcomes, controlling for demand, input prices, and firm characteristics; difference-in-differences specifications and instrumental-variable robustness checks; corroborated by sectoral case studies.
Rising technological capital (K_T) — proxied by robot/automation density, software and intangible capital accumulation, AI adoption surveys, and AI-related patenting — leads to a decline in labor’s share of output.
Firm- and industry-level panel regressions linking constructed K_T intensity measures to labor shares, supported by macro growth-accounting decompositions; robustness checks include difference-in-differences and instrumenting adoption with plausibly exogenous shocks (e.g., cross-border technology diffusion, trade shocks); validated with cross-country comparisons and case studies.
Fuel subsidy reform imposed an enormous fiscal burden that peaked at 2.8% of GDP in 2022, limiting the macroeconomic leverage of AI-driven efficiency gains.
Reported fiscal statistic in the paper (2.8% of GDP in 2022) and its role in analysis of why AI savings do not translate into large macro gains.
The oil and gas trade balance remained in deficit at -1.55 billion USD in May 2025 and -1.58 billion USD in July 2025 despite an overall national trade surplus.
Reported trade-balance figures in the paper (monthly trade statistics for May and July 2025).
The user study reported in the paper involved 34 participants.
Paper explicitly states 'In a 34-participant user study' in the abstract/summary.
We developed Pocket MonstARs, a controlled gamified abstraction of HRC warehouse inventory picking in which virtual monsters serve as proxies for pick targets, while labeled and object-marked boxes preserve the real-world identification demands of the picking task.
Methods section / system description in the paper describing the experimental testbed used for the user study (Pocket MonstARs).
We conducted a small-scale randomized experiment to measure uplift from human-agent collaboration on real-world computational reproducibility tasks.
Randomized experiment described in the paper (authors report it was small-scale; details in methods section).
The paper introduces a predictability phase diagram that organizes tasks into three regimes: Static-Sufficient, Dynamic-Critical, and Noise-Dominant.
Conceptual/theoretical taxonomy presented in the paper (phase diagram and regime definitions; later tested experimentally per the text).