Evidence (13827 claims)
Adoption
8454 claims
Productivity
7544 claims
Governance
6789 claims
Human-AI Collaboration
6327 claims
Org Design
4126 claims
Innovation
4058 claims
Labor Markets
3520 claims
Skills & Training
2924 claims
Inequality
2057 claims
Evidence Matrix
Claim counts by outcome category and direction of finding.
| Outcome | Positive | Negative | Mixed | Null | Total |
|---|---|---|---|---|---|
| Other | 749 | 195 | 97 | 889 | 1979 |
| Governance & Regulation | 815 | 391 | 188 | 121 | 1539 |
| Organizational Efficiency | 771 | 189 | 124 | 83 | 1177 |
| Technology Adoption Rate | 624 | 233 | 123 | 96 | 1084 |
| Research Productivity | 410 | 121 | 56 | 331 | 929 |
| Output Quality | 466 | 177 | 59 | 47 | 749 |
| Decision Quality | 320 | 174 | 75 | 42 | 618 |
| Firm Productivity | 435 | 55 | 88 | 20 | 604 |
| AI Safety & Ethics | 214 | 276 | 65 | 33 | 593 |
| Market Structure | 178 | 166 | 122 | 24 | 495 |
| Task Allocation | 206 | 64 | 70 | 31 | 376 |
| Skill Acquisition | 165 | 57 | 60 | 17 | 299 |
| Innovation Output | 201 | 27 | 41 | 18 | 288 |
| Employment Level | 105 | 51 | 107 | 13 | 278 |
| Fiscal & Macroeconomic | 131 | 69 | 43 | 26 | 276 |
| Consumer Welfare | 116 | 63 | 42 | 11 | 232 |
| Firm Revenue | 149 | 46 | 26 | 3 | 224 |
| Inequality Measures | 44 | 122 | 49 | 6 | 221 |
| Task Completion Time | 169 | 29 | 8 | 12 | 219 |
| Worker Satisfaction | 89 | 61 | 20 | 12 | 182 |
| Error Rate | 69 | 91 | 10 | 2 | 172 |
| Regulatory Compliance | 76 | 68 | 14 | 5 | 163 |
| Training Effectiveness | 92 | 19 | 13 | 19 | 145 |
| Wages & Compensation | 77 | 36 | 25 | 6 | 144 |
| Automation Exposure | 51 | 54 | 22 | 12 | 142 |
| Team Performance | 86 | 17 | 27 | 9 | 140 |
| Developer Productivity | 94 | 17 | 14 | 6 | 132 |
| Job Displacement | 12 | 80 | 20 | 1 | 113 |
| Hiring & Recruitment | 51 | 7 | 8 | 3 | 69 |
| Skill Obsolescence | 5 | 45 | 6 | 1 | 57 |
| Creative Output | 31 | 16 | 7 | 2 | 57 |
| Social Protection | 27 | 16 | 8 | 2 | 53 |
| Labor Share of Income | 17 | 17 | 17 | — | 51 |
| Worker Turnover | 11 | 12 | — | 3 | 26 |
| Industry | — | — | — | 1 | 1 |
The R&D deduction policy has stronger effects on non-state-owned firms.
Heterogeneity analysis contrasting policy effects between state-owned and non-state-owned firms reported in the paper.
The R&D deduction policy has stronger effects on firms with high capital intensity.
Heterogeneity analysis in the paper showing larger estimated policy effects for high capital intensity firms.
The R&D deduction policy has stronger effects on firms characterized by rapid technological obsolescence.
Heterogeneity analysis reported in the paper comparing treatment effects across firms with different rates of technological obsolescence.
The policy effect operates by improving total factor productivity (TFP).
Mechanism analysis showing a positive association between the R&D deduction policy and firms' estimated TFP.
The policy effect operates by boosting firms' innovation capabilities.
Mechanism analysis in the paper linking the R&D deduction policy to measures of innovation capability (e.g., innovation output/indicators).
The policy effect operates by alleviating financing constraints for firms.
Mechanism analysis reported in the paper (mediation/heterogeneity analyses linking policy to reduced financing constraints).
The additional deduction policy for R&D expenses (the R&D policy) significantly enhances the sustainable development outcomes of intelligent manufacturing enterprises.
Panel data from listed manufacturing firms in China analyzed using a quasi-natural experiment design; main empirical specification shows a statistically significant treatment effect (abstract reports significance). Robustness checks reported.
HEWU is designed to become the cited standard before better-resourced players define competing frameworks, establishing measurement infrastructure for the cognitive industrial revolution the way GAAP established it for capital markets.
Aspirational/strategic claim made by the authors about intended role and adoption of HEWU (no empirical support provided).
In that deployment the framework measured approximately $378,000 in annual labor value of machine-equivalent work.
Same empirical manufacturing deployment reported in the paper (single case/example).
In a representative manufacturing deployment, the framework measured 8.4 FTE of machine-equivalent labor.
Empirical example reported in the paper described as a 'representative manufacturing deployment' (appears to be a single deployment/case).
The paper introduces the Machine Labor Index (HEWU-PSI), a time-series economic indicator designed to track aggregate machine labor output at company, sector, and national level, analogous in function to the Purchasing Managers' Index.
Methodological contribution described in the paper (proposal of an index and its intended scope; no empirical time-series dataset reported).
The paper introduces AILU (AI Labor Units) as a software-specific subset metric.
Methodological contribution described in the paper (definition of a software-specific metric subset).
The paper presents the conceptual foundation, mathematical model (HEWU = MO ÷ HB × CF × QF), calibration framework, Baseline Library architecture, and auditability mechanisms underlying the standard.
Paper's methodological content (explicit model formula and supporting frameworks described).
This paper introduces the Human-Equivalent Work Unit (HEWU), a standardized metric that converts AI and automation system output into human labor equivalents, expressed as full-time employee (FTE) equivalents and annual labor value ($).
Methodological contribution described in the paper (definition and proposal of a new metric; no empirical validation sample reported).
Artificial intelligence systems are autonomous agents performing economically meaningful labor at scale across customer service, software engineering, logistics, manufacturing, and knowledge work.
Author's conceptual/empirical assertion in the paper (no specific sample, presented as general observation).
Smart devices adoption is particularly influential (positively associated) for exports to China and to other countries (multivariate probit result).
Multivariate probit model of destination-specific export decisions showing significant positive associations for smart devices with exports to China and 'other countries' (sample size not reported in prompt).
Robotics adoption is a key factor (positively associated) for exports to all destination regions examined (multivariate probit result).
Multivariate probit analysis of destination-specific export decisions indicating significant positive associations between robotics adoption and exports across all destinations (sample size not reported in prompt).
Cloud computing adoption is significantly associated with exports to countries outside the European Union and China (multivariate probit model result).
Multivariate probit analysis of destination-specific export decisions indicating significant effects of cloud computing for exports to non-EU, non-China countries (sample size not reported in prompt).
Adopting smart devices significantly increases the likelihood that a firm exports (probit model result).
Probit regression analysis of firms' export probability using smart devices adoption as an explanatory variable (sample size not reported in prompt).
Adopting robotics significantly increases the likelihood that a firm exports (probit model result).
Probit regression analysis of firms' export probability using robotics adoption as an explanatory variable (sample size not reported in prompt).
Adopting cloud computing significantly increases the likelihood that a firm exports (probit model result).
Probit regression analysis of firms' export probability using cloud computing adoption as an explanatory variable (sample size not reported in prompt).
Adopting artificial intelligence (AI) significantly increases the likelihood that a firm exports (probit model result).
Probit regression analysis of firms' export probability using AI adoption as an explanatory variable (sample size not reported in prompt).
The analysis identifies seventeen emerging occupational categories benefiting from reinstatement effects, concentrated in human-AI collaboration, AI governance, and domain-specific AI operations roles.
Modeling/taxonomy result reported in the paper listing 17 emerging occupational categories characterized as benefiting from reinstatement effects (human-AI collaboration, governance, operations).
Our findings indicate an increasing agent activity in open-source projects.
Trend analysis reported in the paper showing growth in agent-originated activity within the assembled dataset of PRs and associated metadata.
TAI introduces recursive feedback loops between technology, knowledge, and output that redefine long-term growth trajectories and the equilibrium conditions of economies.
Derived from the paper's dynamic model: analytical results showing feedback mechanisms between technology, knowledge stock, and output; presented as theoretical model implications rather than validated empirical findings.
The model integrates AI as both a productivity amplifier and an autonomous driver of capital accumulation.
Stated methodological contribution: the authors extend Solow (1956) and Romer (1990) frameworks to build a dynamic model in which AI enters production as an amplifier of productivity and as an autonomous engine for capital accumulation; evidence is theoretical/model construction rather than empirical.
Transformative artificial intelligence (TAI) is capable of driving structural economic change comparable to the industrial revolution.
The paper asserts this claim by analogy and conceptual argument in the introduction; it frames TAI as 'capable of driving structural economic change comparable to the industrial revolution' without reporting empirical data — supported by theoretical reasoning and historical analogy.
Governments should create an enabling environment that aligns AI innovation with inclusive financial systems to stimulate entrepreneurship, including strengthening entrepreneurship support, enhancing R&D incentives and STEM capacity, sustaining targeted innovation funding, and reforming financial regulations to improve start-up financing and reduce early-stage capital constraints.
Policy recommendations given in the abstract, presented as implications of the empirical findings from the analysis of 23 countries (2002–2023).
AI significantly stimulates entrepreneurship only in financially advanced environments (i.e., above a threshold of financial development), where robust financial institutions and capital investment unlock its transformative potential.
Threshold results from dynamic panel threshold regression reported in the abstract for a sample of 23 countries (2002–2023) showing the AI effect on entrepreneurship is significant only in higher financial development regimes.
Financial development has a positive moderating effect on the AI–entrepreneurship nexus, suggesting complementarities between technological innovation and financial systems.
Abstract states moderation/interaction evidence from dynamic panel threshold regression applied to the panel of 23 countries (2002–2023) showing financial development strengthens the AI–entrepreneurship relationship.
Capital formation, human development, and financial development also play essential roles in driving entrepreneurial growth.
Reported as significant predictors in the dynamic fixed-effects panel analysis on 23 countries (2002–2023) described in the abstract.
AI promotes entrepreneurship by fostering innovation and efficiency.
Estimated with dynamic fixed-effects and dynamic panel threshold regressions on a panel of 23 developed and developing countries covering 2002–2023; abstract reports a positive association between AI technology innovation and entrepreneurship.
Effective collaboration with AI for software engineering (SE) tasks may benefit from functional design rather than replicating human SEI traits, thereby redefining collaboration as functional alignment.
Authors' conclusion and recommendation derived from qualitative interview evidence (10 practitioners) and the proposed concept of functional equivalents.
The authors introduce the concept of 'functional equivalents': technical capabilities (internal cognition, contextual intelligence, adaptive learning, and collaborative intelligence) that achieve collaborative outcomes comparable to human SEI attributes.
Conceptual contribution proposed by the authors based on interview findings and theoretical argumentation (no quantitative validation reported).
Socio-emotional intelligence (SEI) enhances collaboration among human teammates.
Stated as background in the paper (no primary data from this study provided to support the claim).
The binding constraint on human–AI complementarity in the Global South is not technology access but labor market institutions (formality).
Interpretation of empirical findings (formality interactions, triple interaction result) from the augmented Mincer regressions on Colombian data (N = 105,517).
These results provide the first developing-country evidence of cognitive factor decomposition in AI-augmented labor markets.
Claim based on the empirical results from the study using Colombian data and comparison to literature (author statement).
The augmentation premium is strongest in the health and education sectors.
Heterogeneity analysis / sectoral estimates in the augmented Mincer regression using the merged dataset (N = 105,517); reported strongest effects in health and education.
The augmentation premium (return to H^A with AI) is strongest for experienced workers (ages 46-65).
Heterogeneity analysis / subgroup estimates by age in the augmented Mincer regression using the merged dataset (N = 105,517); reported finding that ages 46–65 show the largest augmentation premium.
A triple interaction confirms formality as the binding mechanism: beta_{AHC x D x Formal} = +0.272 (p < 0.001).
Coefficient on triple interaction term in augmented Mincer regression estimated on merged dataset (N = 105,517); reported estimate +0.272, p < 0.001.
In the estimated augmented Mincer equation, the wage return to augmentable-cognitive capital (H^A) increases with AI adoption in the formal sector (beta_2 = +0.051, p < 0.001).
Econometric estimate from augmented Mincer regression using merged data (household survey N = 105,517; LLM-based occupational augmentability measures); reported coefficient beta_2 = +0.051 with p < 0.001.
The empirical analysis uses LLM-generated measures of occupational augmentability for 18,796 O*NET task statements mapped to 440 Colombian occupations, merged with household survey microdata (N = 105,517 workers).
Data construction described in the paper: LLM scoring of O*NET tasks (18,796 tasks), mapping to 440 occupations, merged with household survey microdata (sample N = 105,517).
I derive a corrected Mincerian wage equation and show that the standard specification is misspecified in AI-augmented economies.
Analytical derivation in the paper (theoretical correction to Mincerian wage equation).
AI capital interacts asymmetrically with those components: it substitutes for routine cognitive work (H^C) while complementing augmentable cognitive work (H^A) through an amplification function phi(D).
Theoretical production-function model and derivation in the paper (analytical result).
The paper proposes a decomposition of human capital into three orthogonal components: physical-manual (H^P), routine-cognitive (H^C), and augmentable-cognitive (H^A).
Theoretical proposal in the paper (modeling framework).
Implicit budget constraints from BCR circumvent adversarial gradients and catastrophic optimization collapse that occur with explicit length penalties, providing a highly stable, constraint-based alternative for length control.
Empirical comparison between BCR (implicit budget constraints) and methods using explicit length penalties reported in the paper; claim of improved stability and avoidance of catastrophic optimization collapse.
Qualitative analyses reveal emergent self-regulated efficiency: models autonomously eliminate redundant metacognitive loops without explicit length supervision.
Qualitative analysis of model behavior reported in the paper (no quantitative effect sizes provided in the excerpt).
BCR challenges the traditional accuracy-efficiency trade-off by demonstrating a 'free lunch' phenomenon at standard single-problem inference (i.e., reduced token usage with maintained or improved accuracy even at N=1).
Reported experimental results on 1.5B and 4B model families showing token reductions and maintained/improved accuracy at standard single-problem inference.
As N increases, accuracy degrades far more gracefully than baselines, establishing N as a controllable throughput dimension.
Comparative experiments versus baselines varying concurrent-problem count N; qualitative claim that accuracy degradation is 'far more graceful' than baselines.
As the number of concurrent problems N increases during inference, per-problem token usage decreases monotonically.
Reported experimental finding described as a novel task-scaling law observed when varying N at inference time; no numeric effect sizes provided in the excerpt.