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 |
AI‑enabled forecasting supports index insurance and credit markets by reducing information asymmetries and could lower risk premia for smallholders.
Pilot projects and program evaluations of forecasting tools and index insurance cited in the synthesis; conceptual discussion on mechanisms for reduced information asymmetry.
Returns to AI investments are contingent on complementary inputs (credit, irrigation, extension); policy should target bundles of support rather than stand‑alone technology handouts.
Comparative analysis across technology‑led vs hybrid interventions and conceptual frameworks showing complementarities; supporting case studies where bundled support increased effectiveness.
Public investment in digital infrastructure, training, open data, and targeted subsidies or incentives is critical for equitable scaling of ag‑tech among smallholders.
Policy review and examples of public–private partnerships and subsidy models; comparative analysis showing better diffusion where public investments accompanied technology introduction.
Green financial instruments (blended finance, index insurance) and tailored finance products lower barriers to adoption but require appropriate risk assessment and product design for smallholders.
Policy review and program evaluation examples of blended finance and index insurance schemes; synthesis notes conditional success depending on product design and risk modeling.
Climate‑smart and agroecological practices enhance resilience and ecosystem services when combined with technological tools.
Synthesis and comparative analysis of ecology‑led and hybrid interventions; case studies showing improved resilience indicators (soil health, water retention, pest regulation) when ecological practices are used alongside technology.
A technology mix (precision agriculture, AI, IoT) improves input targeting (water, fertilizer, pesticides), yield forecasting, and supply‑chain efficiency.
Compiled evidence from pilot projects, case studies, and program evaluations reporting improved targeting and forecasting using precision sensors, AI models, and IoT monitoring; comparative analysis highlighting technological contributions to supply‑chain data flows.
Integrating advanced technologies (precision agriculture, AI, IoT), ecological practices (climate‑smart agriculture, agroecology), and inclusive finance can substantially raise smallholder productivity, resource efficiency, and environmental sustainability.
Synthesis of findings from empirical studies, pilot projects, case studies, and program evaluations across multiple regions; comparative analysis contrasting technology‑led, ecology‑led, and hybrid interventions. No single long‑run RCT establishes magnitude; evidence comes from multiple types of shorter‑term or context‑specific studies.
AI increases returns to managerial capabilities that supervise and integrate AI systems, making measurement of managerial capital central for assessing firm performance.
Conceptual linkage between managerial capital and AI complementarities, supported by illustrative cases and recommendations for empirical measurement (e.g., managerial-skills proxies), not by new causal estimates.
Organizational value from AI depends on complementary assets — data quality, IT infrastructure, managerial expertise, and organizational routines.
Conceptual complementarities framework drawing on economics of organization and technology adoption literature; illustrated with case vignettes rather than a specific econometric analysis.
Decision-making is shifting from intuition-driven to data- and model-informed processes: managers use predictive models and prescriptive algorithms to inform choices while retaining responsibility for value trade-offs and unmodelled risks.
Theoretical integration and qualitative examples from organizational practice; references to task-level analyses and possible experimental designs rather than new randomized evidence.
Management systems evolve toward continuous monitoring, predictive forecasting, automated workflows, and adaptive control loops that change KPI definitions and performance measurement.
Synthesis of existing management and information-systems literature and illustrative organizational examples; recommendations for measurement and simulation-based investigation.
AI acts as a complement to — not a wholesale replacement for — human managerial skills; effective management in the AI era requires combining algorithmic capabilities with human judgment, ethics, and leadership.
Theoretical argumentation and cross-sector illustrative examples; integration of prior empirical findings from AI and management literatures rather than new causal evidence.
AI is transforming management by augmenting traditional managerial functions (planning, organizing, leading, controlling).
Conceptual synthesis and literature review drawing on prior management theory and illustrative case studies; no single new large-scale empirical dataset reported.
New markets will emerge for verification-as-a-service, provenance tooling, and compliance tools, and firms that embed stronger integrated verification may gain competitive advantage.
Market-structure reasoning and conjecture about firm incentives; illustrative examples but no market-size estimates or empirical validation.
AI-assisted development will increase demand for verification-specialist roles and tools, shifting labor from routine construction toward oversight, validation, and incident response.
Economic reallocation argument and industry forecasting reasoning; no labor market data or trend analysis included in the paper.
Large language models and generative tools dramatically increase the rate at which code, tests, configs, and docs can be produced.
Conceptual claim supported by descriptive argumentation and illustrative examples (thought experiments and plausible developer workflows). No empirical dataset or measured throughput reported in the paper.
Adoption of AI in research strengthens institutional research performance and enhances global academic competitiveness.
Stated in Key Points and Implications. Presented as an implication of observed productivity gains; likely supported by case studies, institutional reports, and correlational analyses (usage logs correlated with productivity metrics) referenced in the literature synthesis, but no causal identification or sample details given in the abstract.
AI tools reduce cognitive and technical workload, enabling researchers to work more efficiently and produce higher-quality outputs.
Stated in Key Points and Main Finding. Basis appears to be aggregated empirical and experiential reports (surveys/interviews, case studies, and some task-based experiments in the literature). The paper's abstract does not provide explicit measurement or sample details.
AI tools assist across the full research lifecycle: idea generation, study design, literature review and synthesis, data management and analysis, writing/editing, publishing, communication, and compliance.
Key point asserted in the paper. Implied support comes from aggregated reports and studies of tool functionality and user reports (literature review, surveys, case studies). No specific sample or usage statistics provided in the abstract.
AI is becoming an integrated research productivity layer in universities that speeds and improves the entire scholarly workflow — from idea generation through analysis to dissemination — by lowering cognitive and technical burdens, which boosts research quality and institutional research performance.
Statement presented as the paper's main finding. Abstract summarizes "recent evidence" but does not specify original data or methods; likely based on literature synthesis (empirical studies, survey/interview work, case reports) rather than a single original dataset. No sample size, measurement definitions, or identification strategy provided in the abstract.
First‑mover adoption and superior governance can create persistent competitive advantages for firms deploying generative AI effectively.
Theoretical reasoning and case examples from industry reports included in the synthesis; absence of broad causal evidence noted.
Scale and data advantages associated with generative AI adoption may reinforce winner‑take‑all dynamics, favoring large firms that can exploit data and integration economies.
Conceptual argument and industry observations synthesized in the review; no comprehensive market concentration empirical analysis presented.
Realizing sustainable economic value from generative AI requires robust governance, AI literacy, and human‑centric augmentation strategies (AI as assistant, not replacement).
Normative conclusion based on conceptual synthesis of empirical patterns and theoretical arguments in the review.
Generative AI has potential to improve the quality of information processing and the speed of decision‑making.
Conceptual arguments plus early case examples and small empirical studies reported in the literature synthesis; no broad causal estimates provided.
Short‑term deployments of generative AI produce efficiency gains such as time savings and faster turnaround.
Early empirical studies and industry reports summarized in the review; reported case examples of tool deployments (no unified sample size reported).
Generative AI produces measurable gains in operational efficiency and strategic insight.
Synthesized findings and illustrative case examples from early empirical studies and industry reports; authors note lack of large-scale causal evidence.
Generative AI enables scalable personalized communication with customers, employees, and partners.
Aggregation of industry use cases and early empirical reports discussed in the conceptual synthesis (no large-scale causal studies reported).
Generative AI enhances decision support by synthesizing information, surfacing options, and generating explanations for decision‑makers.
Critical literature synthesis and early case examples from industry reports and small studies cited in the review; theoretical evaluation of decision workflows.
Generative AI automates routine administrative workflows and parts of analytical pipelines.
Nano review / conceptual synthesis aggregating early empirical studies, industry reports, and case examples; no original primary dataset reported.
Short-run: measurable productivity gains for many coding tasks imply higher effective output per developer.
Controlled experiments and benchmark tasks that report time savings and/or increased task throughput with LLM assistance; studies often in lab/microtask settings with varying sample sizes.
Organizations will need to build processes and tools (automated testing, static analysis, code review augmented for AI outputs) to realize net benefits safely.
Qualitative case studies and practitioner reports documenting emerging organizational practices and recommendations; derived from observed failure modes and security/IP risks.
The highest value arises when human developers verify, adapt, and integrate AI suggestions—human–AI complementarity.
User studies and controlled experiments showing improved outcomes when humans validate and edit AI outputs; qualitative interviews and case studies reporting effective human-in-the-loop workflows.
These tools lower initial barriers for novices by giving example code, explanations, and templates, potentially accelerating onboarding.
User studies, observational analyses, and qualitative interviews reporting that novices use LLM outputs as examples and templates; evidence primarily short-term and context-dependent.
LLMs are most effective when used interactively as assistants rather than as autonomous code authors.
User studies, observational analyses, and controlled comparisons showing better outcomes for interactive, iterative prompting and verification versus one-shot autonomous code generation; heterogeneous study designs (mostly short-term lab or microtask settings).
LLMs can speed up many programming tasks (boilerplate, code completion, documentation, simple debugging) and change how developers iterate.
Synthesis of controlled experiments and benchmark tasks comparing developer speed/accuracy with and without LLM assistance, supplemented by user studies and observational analyses; sample sizes and tasks vary across studies (typically lab/microtask settings, often tens to low hundreds of participants).
Token taxes incentivize more efficient model designs (fewer tokens per task) and may shift competition toward lightweight models or on-device solutions.
Mechanism-based economic reasoning about price incentives included in the paper; no empirical or simulation evidence provided.
Agent-based models (ABMs) are needed to simulate micro-to-macro dynamics of token taxes because standard representative-agent or DSGE models may miss heterogeneity, network effects, and path dependence.
Methodological argument in the paper advocating ABMs; no ABM results included (proposal only).
Black-box token verification (tamper-evident consumption tokens or receipts tied to API calls) can prove taxable consumption without full model inspection.
Technical proposal for cryptographic/ledgered receipts described in the paper; no prototype, security analysis, or empirical tests provided.
A staged audit pipeline—black-box token verification, norm-based tax rates, then white-box audits—provides a feasible path to design and evaluate token taxes.
Proposed enforcement architecture described in the paper (conceptual design); no deployment or simulation results presented.
Token taxes can be enforced using existing compute-governance and commercial billing infrastructure (API billing, cloud metering, hardware telemetry, attestation).
Technical architecture discussion proposing use of existing billing and telemetry systems; no implementation or pilot data provided.
Compared with robot- or FLOP-based taxes, token taxes better capture where AI-generated value is realized.
Analytic comparison in the paper arguing tokens map to user-facing consumption while FLOP/robot taxes map to inputs; conceptual reasoning rather than empirical test.
The framework enables scenario testing for policies and shocks (e.g., lockdowns, targeted interventions, information campaigns) where human judgment and adaptation matter.
Paper reports experiments across policy regimes and discusses use cases for testing timing, targeting, and communication strategies; however, concrete policy evaluation examples and quantitative policy results are not detailed in the summary.
Experiments run with multiple LLM backends (proprietary and open-source) show qualitatively consistent dynamics, indicating framework stability to model choice.
Cross-backend comparisons and robustness checks reported in the paper; several LLMs used though the exact models and counts are not specified in the summary.
Behavioral changes in the simulation emerge endogenously from cognitive reasoning rather than from parameterized switches, producing context-sensitive, heterogeneous responses.
Description of agent heterogeneity (differences in perceptions, priorities, and local conditions) and use of CoT reasoning per agent; reported emergent, diverse responses in experiments. (Degree of heterogeneity and quantitative heterogeneity metrics not provided in summary.)
LLM-driven agents embedded in a Perception–Deliberation–Action (PDA) loop produce endogenous, human-like adaptive behaviors via Chain-of-Thought reasoning.
Multi-agent simulation where each agent is implemented as an LLM-driven cognitive unit running the PDA loop each timestep; agents use Chain-of-Thought (CoT) prompts/internal reasoning to make decisions. (Exact simulation sample size / population not specified in summary.)
Task‑based, dynamic exposure measures and real‑time data enable earlier detection of displacement risks and reallocation needs than static, occupation‑level extrapolations.
Conceptual argument and proposed architecture; no empirical timing comparison or lead-time statistics provided.
LLMs can be used to score task automation/augmentation plausibility and to detect emergent tasks.
Methodological proposal describing use of LLMs for semantic mapping/scoring of tasks; no empirical validation or accuracy metrics for LLM task scoring provided in the paper.
Modeling nonlinearity (threshold adoption, network spillovers, complementarities) and path dependence in adoption dynamics is necessary rather than relying on linear extrapolation.
Theoretical argument and model suggestions (S‑curve diffusion, agent-based models) in the paper; no empirical comparison demonstrating superior performance provided.
Applying causal inference methods (difference‑in‑differences, synthetic controls, instrumental variables, structural counterfactuals) can distinguish automation (task substitution) from augmentation (productivity/role change) and estimate net employment effects.
Methodological recommendation with examples of applicable identification strategies; no specific empirical applications or results reported in the paper.
Integrating multiple data streams (CPS, LEHD/LODES, UI wage records, administrative microdata, job ads, occupational manuals, enterprise adoption surveys) yields richer gross‑flows and skills measurement than using single data sources.
Proposed data-integration strategy and references to candidate datasets; no empirical demonstration or quantified improvement in measurement presented.