Evidence (7560 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
Filtered →
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 |
Human Ai Collab
Remove filter
We measure the change in the skill premium using a difference-in-differences design on freelance websites worldwide.
Statement of empirical method: difference-in-differences design applied to data from freelance platforms with global coverage; no sample size provided in the abstract.
The present wave of automation targets non-routine cognitive activity such as coding, technical writing, and graphic design, unlike past automation which mainly involved routine manual activity.
Framing/background statement in the paper contrasting historical automation (routine manual tasks) with current AI-driven automation of non-routine cognitive tasks; no sample size or quantitative test reported in the abstract.
There is a similar shift to agentic tooling outside OpenAI, particularly within organizations, although external adoption remains lower and more uneven.
Comparative usage analysis across three populations (external personal-account users, external organizational-account users, and OpenAI workers) from Codex logs.
There exists a six-bit prior for which R_max(μ)/R_0(μ) = 39/31 > 5/4, so no universal 5/4 bound holds.
Constructive counterexample provided in the paper: an explicit six-bit prior is presented and analyzed to compute the ratio. This is a theoretical construction, not empirical data.
If the prior μ is close to the independent product prior with the same marginals in the sense that μ(x) ≥ (1−η) π_μ(x) for every state x, then R_max(μ) ≤ R_0(μ) + η n.
Mathematical derivation/proof in the paper under the stated closeness assumption (formal theorem conditional on parameter η and number of bits n). No empirical/sample data.
For any prior μ, R_max(μ)/R_0(μ) ≤ 3/2.
Mathematical proof (theorem) within the paper's Bayesian persuasion model where the sender is strategic and the receiver guesses bits. The result is presented as a proven upper bound under the model's assumptions (no empirical/sample data).
Some skills generalize broadly across tasks and models, whereas others become specialized to role-specific workflows and lose effectiveness under transfer.
Analyses reported in the paper showing heterogeneous transfer behavior across the 22 procedural skills in the AFTER benchmark, with some skills showing broad cross-task and cross-model generalization and others showing role-specific specialization and reduced transfer performance.
Participants' IAT scores were predictive of the time they spent in human-AI collaboration.
Reported predictive relationship between individual IAT scores and measured time spent interacting with/considering resumes during human-AI collaborative screening tasks (likely from regression or correlation analyses); exact statistics and sample size not provided in the excerpt.
The Simpson's paradox in the pooled result is driven entirely by agent composition: Codex dominates 64.9% of the dataset.
Descriptive composition statistics from the AIDev dataset showing agent shares; explicit statement that Codex comprises 64.9% of dataset.
Better measurement matters, but improved measurement alone will not close the coordination gap between researchers and policymakers.
Authors' analytical conclusion arguing that measurement improvements are necessary but insufficient.
These patterns suggest a commoditization effect of AI on labor, with implications for online labor market design, workers' incentives to invest in human capital, and labor welfare.
Interpretation synthesized from the three empirical findings above (decline in human-capital importance, rise in price importance, decline in demand premium for high-human-capital workers, and reallocation toward lower-priced workers). This is presented as the paper's conceptual/mechanistic conclusion and policy implication rather than a separately tested causal estimate. (Empirical basis: Upwork analysis and difference-in-differences; sample size not reported in abstract.)
Stronger synchronization can increase collective output but may also increase systemic fragility and reduce mobility.
Analytical results and trade-off analysis in the model showing the effects of synchronization on collective output, fragility, and mobility; theoretical deduction without empirical sample.
The guarded engagement loop framework conceptualizes generative AI adoption as a feedback process in which risk perceptions may shape interaction conditions that, in turn, can influence observed performance and subsequent trust calibration.
Central conceptual claim of the paper; framework articulated by the authors and presented as a set of testable propositions (theoretical contribution rather than empirical finding in the abstract).
Risk salience may shape interaction dynamics with LLMs via a multilevel feedback mechanism called the 'guarded engagement loop', in which risk perceptions shape interaction strategies that influence observed performance and, in turn, recalibrate trust in generative AI systems.
Conceptual framework proposed by the authors, integrating theories from trust in automation, privacy calculus, algorithm aversion, and social amplification of risk; presented as a theoretical model rather than an empirical test.
LLM guidance was associated with increased pupil size variability.
Physiological eye-tracking measure (pupil size variability) reported and compared across conditions in the simulated SAR experiment.
Eye-tracking data revealed an attention-guidance trade-off: visual resources shifted to the chat interface when LLM guidance was present.
Eye-tracking measures collected during the experiment showing changes in gaze allocation (increased fixations/dwell time on the chat interface) across LLM-guided vs baseline conditions.
The paper formalizes four mechanism theorems explaining the overhead-pressure dynamics: overhead non-additivity, augmentation-saved-time pathways, innovation-premium amplification, and human-AI dyad attribution uncertainty.
Presentation of four mechanism theorems within the paper (theoretical/mathematical exposition rather than direct empirical tests).
The ICH framework predicts three distinct augmentation regimes (determined by combinations of A and C) with distinct policy implications.
Theoretical classification derived from the model; conceptual prediction presented in the paper.
AI-induced changes are displacing existing labor jobs while also creating new jobs that require high technological skills.
Summary claim from the SLR reporting that reviewed empirical studies report both displacement of existing jobs and creation of new, high-skill jobs; no quantified displacement/creation rates provided in the excerpt.
Between 2017 and 2025, studies identified current trends of AI-induced changes affecting both blue-collar and white-collar occupations.
Synthesis statement in the paper reporting that reviewed empirical studies identified trends across blue- and white-collar jobs (timeframe 2017–2025). Specific studies or counts not provided in the excerpt.
AI's rapid evolution has profound effects on the labor market, influencing the levels, skills needed for jobs, and overall jobs content.
Statement from the paper's synthesis/introduction summarizing reviewed empirical studies (systematic literature review covering studies from 2017–2025). Number of underlying studies not reported in the excerpt.
The paper develops the concept of 'bidirectional dynamics' in digital sovereignties, applying a paradoxical view to interpret institutional control objectives and individual autonomy aspirations as persistent organizational tensions in AI adoption.
Theoretical/conceptual development grounded in the empirical single-case study; concept introduced and motivated by observed tensions in the organization (empirical method details and sample size not provided).
Early digital transformation presents tensions but also synergies between digital sovereignty levels in AI adoption.
Empirical observations from the single-case study of a Nordic public transportation organization during early AI adoption; qualitative examples and analysis (specific methods/sample size not stated).
Embodied intelligence is driving the human-machine relationship from a "human-dominated" model toward "collaborative co-creation," which, while boosting productivity, also triggers deep-seated contradictions in production relations.
Conceptual/theoretical argumentation in the paper, drawing on Marx's theory of reproduction; no empirical sample or quantitative data reported.
The near-term value of Agentic AI does not lie in full autonomy or workforce reduction, but in controlled partial autonomy for simple and medium complexity business processes.
Central argumentative claim/recommendation in the paper (theoretical justification; no empirical study or sample size reported).
This shift raises fundamental questions for consumer theory, which has traditionally modeled humans as the primary decision-makers.
Conceptual argument presented in the paper framing the research problem and motivating the new theoretical framework; literature critique rather than empirical test.
The modality gap (weaker penalty for visual vs. textual AI-use disclosure) widens when AI is used in final products but narrows when AI is used in marketing materials.
Interaction analyses across application stages (final product vs. marketing material) within the 41,073 Kickstarter projects, using LLM-assisted classification to label both modality and application stage and entropy balancing for covariate control.
The effectiveness of AI in strategic core functions is contingent upon the human–AI interface.
Stated as a conditional claim in the paper—AI effectiveness depends on the quality of the human–AI interface; no empirical quantification provided in the summary.
Bias transfer from the LLM is asymmetric: agency is suppressed in female-target essays while male-target writing remains largely unaffected.
Comparative analysis within the participant data showing differential effects by target gender (female-target vs male-target essays) in the N = 123 study; reported asymmetry in the paper summary.
Tranquil periods lower subjective risk assessments, raise AI substitution intensity, and compound leverage, generating a cognitive Minsky moment in which subjective risk falls while true systemic fragility rises.
Derived dynamics and comparative statics in the formal model; stated as one of the paper's propositions. No empirical data.
Dominant comments shifted in tone from mockery toward gatekeeping and structural protest.
Speech-act coding of 300 confirmed accusations and sentiment/trajectory analysis showing relative decline in mockery-coded acts and increase in gatekeeping/structural-protest acts over time.
Across compression sweeps, real factor archives, and LLM-SRBench tasks, hybrid gains concentrate in weakly represented but target-bearing directions and vanish as the hypothesis space approaches full rank.
Empirical claim based on experiments over compression sweeps, analyses of real factor archives (A-share factor discovery), and LLM-SRBench tasks; no numerical sample sizes or effect magnitudes provided in the abstract.
Forms of resistance exist, including localisation efforts and Indigenous ethical frameworks, but they remain structurally limited.
Synthesis of examples and themes across the 50 reviewed articles noting reported resistance strategies and their limits.
Frontier proprietary models achieve near-zero success under GUI-based interaction, whereas COM-based execution yields substantial immediate gains.
Experimental comparison reported in the paper on ComCADBench between GUI-based interaction by proprietary models and COM-based execution (authors report success rates and comparative performance).
Environment engineering can amplify productive behaviors (e.g., open-ended exploration, systematic artifact management, inter-agent collaboration) while suppressing harmful behaviors (e.g., reward hacking and high-friction human oversight).
Framing and argument in the paper describing expected effects of environment design (conceptual; no quantification provided in the excerpt).
Trust is conceptualized as network-mediated expectation stabilization in the embodied finance framework.
Theoretical claim in the framework articulating trust as stabilized through network interactions among humans, machines, and platforms; no empirical data.
The proposed framework—the machine–platform–crowd triangle—reframes agency, trust, and value as emergent properties rather than institutional attributes.
Conceptual framing and argumentation within the paper; synthesis of theory to reconceptualize agency, trust, and value; no empirical testing reported.
Architectural smell density (ASD) declines by 6.7% (p = 0.004), but this decline is a denominator effect resulting from lines-of-code growth rather than an actual architectural improvement.
Observed ASD change computed from estimated smell counts and LOC changes in the 151-repository panel and interpreted by decomposing density into numerator (smells) and denominator (LOC).
The same observation is seen with the amount of changes (e.g., code churn, number of modified files) and with the efforts to merge an agentic PR (e.g., merge time and number of comments).
Reported that pre/post comparison across projects shows mixed/no consistent improvement patterns for code churn, modified files, merge time, and comment counts after instruction-file creation (analysis over 15,549 PRs in 148 projects).
Specifying instructions for AI-agents does not necessarily lead to better results.
Project-level before/after comparison of performance metrics for projects that created instruction files (pre/post comparison across 148 projects using the 15,549 PRs).
Du et al. (2026) find that information-based team faultlines can enhance proactive behavior via deep information processing, while AI adoption moderates and mitigates the negative effects of social-based faultlines on team cooperation.
Information-processing theoretical framing and empirical analysis reported in the paper (study type and sample size not specified in the excerpt).
Liao et al. (2026) identify multiple equifinal pathways to high performance in digit-oriented spin-offs (parent-oriented, independent-oriented, ambidextrous-oriented configurations) using fuzzy-set qualitative comparative analysis (fsQCA).
fsQCA analysis reported in the paper (methodological approach described; sample not specified in excerpt).
A store-level policy learned from logged marketplace data selects a discrete multiplier that shifts the dispatch optimizer's tradeoff between delivery quality and batching efficiency.
Methodological description: store-level policy trained from logged data that outputs a discrete multiplier to alter optimizer objective weights; stated design and training approach in paper (no numerical evaluation details provided in the excerpt).
The article develops a conceptual framework linking GenAI use in higher education to knowledge transformation, critical thinking, ethical judgment, digital capability, managerial decision-making, business ethics, workforce readiness, and organizational readiness.
Presentation of a conceptual framework by the authors as part of the review (theoretical/conceptual work; no empirical validation reported).
GenAI should be understood as more than an educational technology: it affects the development of managerial decision-making, business ethics, and workforce readiness for future managers, entrepreneurs, administrators, policymakers, and business professionals.
Conceptual argument and literature synthesis presented in the review article (no primary empirical sample).
Generative AI (GenAI) is reshaping higher education by changing learning practices, academic writing, knowledge access, assessment preparation, research support, and student engagement.
Narrative literature review / synthesis (review article). No primary empirical sample reported — claim drawn from cited literature and conceptual synthesis.
AI agents can rival or exceed human methodological diversity at the design layer while remaining vulnerable at the verdict layer.
Synthesis of above experimental findings: Claude Code and Codex matched/exceeded human methodological diversity measures (20 runs) but exhibited vulnerability to prompt-induced changes in verdict behavior (especially Claude Code).
Implementation success depends heavily on data quality, workflow redesign, interpretability, governance, and procurement alignment.
Synthesis of factors identified across included studies and supporting regulatory/industry documents as important determinants of successful deployment.
However, evidence is uneven: many studies are simulation-based.
Review observation from the synthesis of the 35 included studies noting study designs (simulation prevalence noted but not numerically specified).
Perkembangan AI mengotomatisasi tugas rutin sekaligus menciptakan peluang pekerjaan baru berbasis digital.
Sistematis studi literatur yang menelaah 33 sumber ilmiah, laporan lembaga internasional, dan kebijakan terkait (n=33).