Evidence (4892 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
Filtered →
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 |
Org Design
Remove filter
Intelligent centralized orchestration fundamentally improves multimodal AI deployment economics.
Authors generalize from the reported empirical results (reductions in time-to-answer, conversational rework, and cost on their 2,847-query evaluation) to claim broader economic benefits of centralized orchestration.
Critical thinking development and ethical reasoning cultivation retain 70-75% human centrality.
Authors provide a numerical estimate (70-75% human centrality) in their functional analysis; the paper does not report empirical methods or sample evidence for this figure.
Mentorship and social development remain largely human-dependent with only 25-30% substitutability by AI.
Paper's estimated substitutability range (25-30%) for mentorship and social development; the estimate is not accompanied by empirical data or described methodology.
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.
The presented framework contributes to the responsible use of AI, productivity, and long-term economic competitiveness in the United States.
Forward-looking claim rooted in conceptual reasoning and literature synthesis; no longitudinal data, economic modeling, or empirical evidence is provided to demonstrate the claimed macroeconomic effects.
Deterministic verifiers and benchmarks like SkillsBench are important for certification and procurement decisions because they enable verifiable, repeatable gains.
Normative implication in the paper based on the use of deterministic verifiers to measure Skill impact reproducibly; this is an interpretive claim about downstream decision-making rather than an experiment-derived metric.
Focused, modular Skill design favors modular pricing and bundling strategies (i.e., narrow high-impact Skills premium; broad libraries lower margin).
Policy/market implication derived from the experimental finding that focused 2–3-module Skills outperform comprehensive documentation; the pricing/bundling claim is an economic inference, not empirically tested in the paper.
Because curated Skills yield large average gains, human curation of high-quality procedural knowledge has economic value and could be a high-return activity.
Paper's economic implication drawn from the empirical +16.2 pp average pass-rate improvement for curated Skills. This is an interpretation/inference rather than a direct empirical economic measurement.
To establish causal links between price, perceived value, and outcomes, researchers should use field experiments, A/B tests, instrumental variables, and natural experiments.
Methodological recommendations in the paper's implications section, grounded in authors' assessment of current methodological gaps.
AI economics research should build hybrid behavioral–machine learning models that predict perceived value at scale and integrate them into pricing optimization frameworks.
Implications and research agenda provided by the authors based on gaps identified in the SLR; recommended modeling approach rather than empirical finding.
Future research should incorporate ethics, fairness, and transparency into pricing algorithms and leverage predictive technologies to estimate and operationalize perceived value in real time.
Authors' explicit future-research recommendations derived from gaps identified in the SLR.
Organizational capabilities (data, analytics, governance, cross-functional alignment) are critical enablers of successful digital VBP.
Repeated identification of organizational capability factors across the 30 reviewed studies and synthesis into a thematic cluster by the authors.
Platform and market designers should not assume human-like conversational properties and may need protocols (e.g., provenance tagging, limits on template replies) to preserve information quality.
Synthesis of observed structural features on Moltbook (high formulaicity, low alignment, introspection bias, coherence decay) and recommended interventions; this is a prescriptive implication derived from empirical patterns.
When pipelines are hierarchical (trees or series-parallel), decentralised pricing converges to stable equilibria, optimal allocations can be found efficiently, and agents have no incentive to misreport values within an epoch under the paper's mechanism.
Combination of theoretical model/analysis (mechanism design under quasilinear utilities and discrete slice items) and simulation results from the ablation study showing convergence and high allocation quality on hierarchical topologies; experiments used multiple random seeds per configuration within the 1,620-run suite.
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.
The framework formalizes complementarities between AI and managerial/human capital (e.g., exception handling, trust-driven adoption), suggesting empirical work should measure task reallocation rather than simple displacement.
Conceptual claim and research agenda recommendations in the paper (no empirical measurement provided).
Staged, practice-oriented workflows lower upfront adoption costs and implementation risk for SMEs, increasing marginal adoption likelihood when organizational readiness and governance are explicit.
Theoretical/economic implication derived from the framework and pilot rationale; not directly validated by large-scale empirical evidence in the paper (asserted implication).
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).
Overall economic aim: lowering the hidden costs and power imbalances introduced by opaque AI systems so that data‑intensive research remains ethically accountable, competitively efficient, and equitably beneficial across jurisdictions.
Authors' stated conclusion and framing of implications for AI economics; normative goal rather than an empirically tested outcome.
Policy levers could include harmonizing cross‑border data governance standards, procurement and funding conditionality for data‑sovereignty guarantees, supporting public/community‑owned infrastructures, mandating disclosures from AI service providers, and subsidizing open‑source alternatives and capacity building.
Policy prescriptions synthesized from the paper's analysis of problems (opacity, fragmentation, unequal infrastructure); presented as recommended interventions, not empirically evaluated within the study.
To maintain autonomy and ethical standards, universities and research funders may need to invest in local infrastructure (on‑premise compute, vetted open tools) — a public good with implications for funding priorities and inequality across countries.
Policy recommendation derived from the case study’s identification of infrastructural inequalities and limited mitigation options; not empirically tested in the paper.
Policy recommendations implied include: reinforce worker voice via required worker representation in AI impact assessments and protection of collective bargaining around technology use; mandate disclosure and standardized impact reporting of AI systems used for hiring/monitoring/promotion/termination; and implement targeted sector- or task-specific enforceable regulations.
Normative policy prescriptions derived from the commentary’s analysis of governance gaps and risks; not empirically tested within the paper.
Adoption of generative neural-network audiovisual tools is effectively inevitable.
Narrative synthesis of technological trends and literature in the review; no original longitudinal adoption model or empirical adoption rates provided (qualitative projection based on cited trends).
Policymakers may need to mandate minimum verification standards or standardize audit trails/provenance metadata in safety-critical domains to reduce information asymmetries and monitoring costs.
Policy recommendation derived from risk- and externality-focused analysis; no policy impact evaluation or legal analysis presented.
Cognitive interlocks (e.g., mandatory proof artifacts, enforced testing gates, provenance/audit trails, verification quotas) make the verification burden explicit and non-bypassable, restoring the appropriate burden of proof.
Architectural design proposal with illustrative usage scenarios; no implementation, field trials, or quantitative evaluation in the paper.
The Overton Framework — an architectural model embedding 'cognitive interlocks' into development environments — can align throughput and verification by enforcing verification boundaries and restore system integrity.
Framework proposed and described conceptually; includes design principles and example interlocks but no empirical prototypes, experiments, or effectiveness evaluations reported.
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.
Policy responses (standards for verification, disclosure rules, worker‑training subsidies) could mitigate negative labor and consumer outcomes while preserving productivity benefits.
Authors' policy recommendations based on interpretive analysis of risks and benefits reported by practitioners; normative suggestion, not empirically tested within the study.
The AR-MLLM prompt/design framework is adaptable to other industrial machine-operation scenarios.
Authors state generalizability as an argument based on the architecture and iterative prompt design; the empirical evaluation in the paper is limited to the CMM case study (no cross-domain experiments reported in the provided summary).
The Reversal Register is an auditable institutional artifact that records for each decision the prevailing authority state, trigger conditions causing transitions, and justificatory explanations, thereby supporting auditability and research.
Design specification and instrumentation proposal in the paper; description of required metadata fields and intended uses. No implemented dataset presented.
Firms that build effective orchestration layers and integrate AI across pipelines may capture outsized gains, increasing winner-take-all dynamics and concentration.
Authors' argument extrapolated from observed coordination benefits/frictions at Netlight and theory about returns to scale in platformized toolchains; no empirical market concentration analysis provided.
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).
Policy and regulation should emphasize transparency, auditability, and model-validation standards in finance to reduce systemic risks from misplaced trust or opaque algorithms.
Authors' normative recommendation based on empirical identification of risks (misplaced trust, overreliance) from survey/interview/operational data; recommendation is prescriptive and not an empirical test within the study.
Firms with large, integrated datasets and standardized processes can gain disproportionate returns, creating potential scale economies and winner-take-most dynamics.
Resource-based theoretical interpretation and illustrative patterns in the reviewed literature; the paper notes empirical evidence is limited and calls for further study.
Better-governed automations can reduce firms’ systemic operational risk and may lower insurance premiums or capital charges; insurers and lenders will value documented governance when pricing risk.
Hypothesized consequence grounded in risk-transfer logic and suggested interaction with insurance/lending markets; presented as implication rather than demonstrated outcome; no insurer data provided.
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.
Authors recommend promoting a shift from single-link outsourcing (PAPM) toward whole-process integrated service provision (WAPM) as a policy implication of the findings.
Discussion/policy-implication section of the paper drawing on empirical results (TWFE and robustness checks) from the CLDS 2014–2018 analysis.