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
Autonomous software-engineering agents remain unreliable in realistic development settings.
Assertion in abstract summarizing the observed current state; likely based on prior literature and/or authors' observations (no empirical sample size given in abstract).
The paper identifies five fundamental architectural mismatches between conventional APIs and autonomous agent requirements: exact-identifier dependence, rendering-oriented responses, single-shot interaction assumptions, user-equivalent authorization, and opaque error semantics.
Conceptual analysis and problem-framing presented in the paper (qualitative identification of five mismatch categories).
Current surveys remain fragmented across system optimization, architecture design, and trust, lacking a unified framework to evaluate the fundamental trade-off between output quality and economic cost.
Authors' literature review and critique of existing surveys; based on mapping of prior works into separated strands (qualitative assessment rather than quantified meta-analysis).
Exponential token consumption introduces severe computational, collaborative, and security bottlenecks.
Synthesis presented in the paper arguing that rising token usage causes system-level constraints; based on literature survey and conceptual analysis (no single empirical sample reported).
Producing hardened, production-grade agent workflows may require extra compute and time, and these costs must be amortized through reuse across a broad user community.
Argument in paper reasoning that added rigor entails higher compute/time costs and that reuse across users is needed to amortize these costs; no empirical cost estimates provided.
By focusing on rapid, real-time synthesis, AI agents are effectively delivering users improvised prototypes rather than systems fit for high-stakes scenarios in which users may unwittingly apply them.
Conceptual argument presented in the paper asserting a qualitative mismatch between on-the-fly agents and high-stakes production needs; no empirical validation reported.
The on-the-fly paradigm short-circuits disciplined software engineering processes—iterative design, rigorous testing, adversarial evaluation, staged deployment, and more—that have delivered relatively reliable and secure systems.
Argumentative claim in paper linking the on-the-fly loop to reduced application of standard SE processes; no empirical study, sample, or quantitative evidence provided.
Without parallel investment in digital literacy, organizational culture, and inter-firm networks, AI will reproduce rather than reduce employment inequalities.
Authors' conclusion drawn from thematic analysis of interviews and conceptual framing; predictive statement based on qualitative findings.
AI adoption in peripheral economies is not a purely technological or financial challenge but a social and human capital challenge, embedded in a biocultural environment shaped by brain drain, institutional thinness, and weak civic intermediation.
Synthesis of interview findings using Bitsani's Biocultural City framework; qualitative evidence from 12 interviews supports this argument.
Knowledge deficits and financial constraints emerge as primary barriers [to AI adoption].
Thematic analysis of the twelve semi-structured interviews reporting these themes as primary barriers.
The welfare equivalence property is unique to the Brier score: for every non-Brier strictly proper scoring rule, the welfare gap under smooth C^1 oversight is bounded below by Ω(Var(1/G'') (γ/β)^2).
Mathematical lower-bound result proved in the paper comparing welfare under smooth C^1 oversight for non-Brier scoring rules; the bound is expressed as Ω(Var(1/G'') (γ/β)^2) in the paper.
The impossibility (that non-affine approval undermines truthful reporting) holds for all strictly proper scoring rules, and the paper provides a closed-form perturbation formula.
General theoretical result proved across the class of strictly proper scoring rules, accompanied by a closed-form formula for the perturbation in the paper.
Any non-affine approval makes truthful reporting suboptimal under the combined objective whenever deviation is undetectable — the principal cannot avoid the perturbation that undermines calibration.
Analytical impossibility theorem in the paper's formal model showing that non-affine approvals create incentives for non-truthful reports when deviations are undetectable (mathematical proof).
Current AI tools are not yet mature enough to replace developers.
Conclusion drawn from the controlled experiment and participant feedback comparing AI-assisted vs traditional task-splitting.
Breaking down user stories into actionable tasks is a critical yet time-consuming process in agile software development.
Background/introductory statement in the paper describing the problem motivation; no experimental sample size reported for this claim.
AI adoption deepens the negative indirect effect of CEO–TMT faultlines on green innovation via reduced eco-attention (moderated mediation).
Reported moderated mediation analysis on the panel dataset (35,347 firm-year observations) showing that AI moderates the indirect path from CEO–TMT faultlines to green innovation through eco-attention, making the indirect effect more negative when AI is greater.
AI technology strengthens the negative relationship between CEO–TMT faultlines and eco-attention (AI exacerbates the adverse effect of faultlines on eco-attention).
Moderation/interaction analysis reported in the paper using the same panel dataset (35,347 firm-year observations) indicating a significant interaction between AI adoption and CEO–TMT faultlines on eco-attention.
CEO–TMT faultlines reduce eco-attention (organizational attention to environmental issues).
Direct association reported in the paper from regression/mediation models using the panel dataset (35,347 firm-year observations) showing a negative relationship between CEO–TMT faultlines and eco-attention.
CEO–TMT faultlines negatively affect green innovation through reduced eco-attention.
Empirical mediation analysis on the panel dataset (35,347 firm-year observations, 2010–2023) testing CEO–TMT faultlines -> eco-attention -> green innovation.
AGI (Artificial General Intelligence) is problematic both conceptually and definitionally.
Authorial assertion in the paper stating AGI is problematic as a concept and definition; framed as a conditioning assumption that shapes the subsequent analysis.
The paper argues we should avoid assuming the inevitability of the current situation relating to AI (i.e., the current commercial AI development trajectory is not inevitable).
Authorial methodological claim in the paper's framing/introductory text; presented as a normative methodological stance rather than empirical evidence.
Existing coordination approaches often occupy two extremes: highly structured methods that rely on fixed roles/pipelines assigned a priori, and fully unstructured teams that enable adaptability but suffer inefficiencies like error propagation, inter-agent conflicts, and wasted resources.
Framing/background claim made in the paper (conceptual argument motivating LATTE).
We contribute a non-additive harm decomposition (welfare loss W, coverage loss C) that exposes how attrition shifts harm from the regulator-accountable surface to a regulator-invisible one.
Methodological contribution in the paper: definition of welfare loss W and coverage loss C and analysis showing attrition reallocates observable vs. unobservable harm; supported by theoretical exposition and simulation examples.
An audit-aware OffAuditDrift strategy that exploits Stackelberg commitment defeats both (Periodic-with-floor and history-conditioned suspicion-escalation) auditor extensions.
Construction of the OffAuditDrift auditee strategy in the paper and simulation/theoretical demonstration that it can evade both proposed auditor policies by exploiting auditor commitment.
We identify a structural feature of any noise-aware static-auditor design: a cover regime in which coverage gaps and granularity gaps cannot be closed simultaneously (formalized as Observation 1).
Theoretical observation/proposition in the paper (Observation 1) derived from the formal model of continuous auditing under noise-aware static auditing rules.
Regulated systems can delay outcome reporting, drift their reports within plausible noise envelopes, exploit longitudinal sample attrition, and cherry-pick among ambiguous metric definitions.
Specification and enumeration of auditee strategies in the paper (Delay, Drift, Cherry-pick, Attrition, OffAuditDrift); conceptual examples and inclusion in simulator.
Continuous post-deployment compliance audits, mandated by emerging regulations such as the EU AI Act and Digital Services Act, create a class of strategic gaming distinct from the one-shot input/output gaming studied in prior work.
Conceptual and theoretical argument in the paper, motivated by regulatory context; formalization of continuous auditing as a multi-round interaction (T-round Stackelberg game).
DePAI entails risks including security, centralization, incentive failure, legal exposure, and the crowding-out of intrinsic motivation, requiring value-sensitive design and continuously adaptive governance.
Risk analysis and conceptual argument in the paper identifying possible failure modes and recommended design/governance responses; no empirical incidence data provided.
The cultural and technical misalignment of the data center and electric power sectors makes coordination difficult.
Analytic claim in the paper describing differing design principles, operational philosophies, and economic incentives as sources of misalignment; presented as conceptual analysis without empirical measurement in the excerpt.
A single hyperscale training campus can draw power comparable to a mid-sized city, driven by one tightly synchronized job whose demand swings by hundreds of megawatts in seconds.
Concrete illustrative assertion in the paper about facility-level power draw and rapid demand swings; no numeric source, dataset, or case-study details provided in the excerpt.
AI training data centers break that assumption (load diversity).
Argumentative claim in the paper asserting that characteristics of AI training workloads violate the load-diversity assumption; no quantitative study included in the excerpt.
Responsible AI research typically focuses on examining the use and impacts of deployed AI systems, and there is currently limited visibility into the pre-deployment decisions to pursue building such systems.
Argument and literature framing presented in the paper based on a scoping review of academic literature, civil society resources, and grey literature.
This concentration can diffuse responsibility and raise the probability of irreversible system-level loss even when local per-action error rates remain low.
Theoretical result/argument from the model linking concentrated decision-energy to increased systemic risk despite low local error rates.
Efficiency pressure, path dependence, scale feedback, and weak boundary constraints concentrate decision-energy in the most efficient node.
Derived from the paper's formal model and argumentation about system dynamics (efficiency and feedback mechanisms); theoretical rather than empirical evidence.
Declining deployment friction changes the safety problem at its root: safety is not only local output correctness or preference alignment, but the control of irreversibility under rising decision density.
Main theoretical argument of the paper; supported by conceptual framing and a formal model that introduces decision-density considerations.
Recent AI systems compress the distance between capability growth and capability deployment.
Conceptual and descriptive claim in the paper's introduction; supported by theoretical argumentation and illustrative examples rather than empirical measurement.
Of these four, integration capacity is the least developed for scientific institutions and the most binding: no improvement in AI tooling can buy it.
Normative/diagnostic claim in the paper about relative scarcity and irreducibility of integration capacity; no empirical measures or sample provided in the excerpt.
Four complements then become scarce and load-bearing for AI-augmented science: verified signal, legitimacy, authentic provenance, and integration capacity (the community's tolerance for delegated cognition).
Theoretical framework proposed by the paper; list of four complements presented as an argument without empirical quantification in the excerpt.
Frontier software engineering agents have saturated short-horizon benchmarks while regressing on the work that constitutes senior engineering: long-horizon, multi-engineer, ambiguous-specification deliverables.
Position asserted in the paper based on literature/benchmark trends and authors' field observations; no original empirical dataset or quantified analysis provided in the paper text excerpt.
Specification discipline, not model capability, is the binding constraint on AI-assisted software dependability.
Synthesis conclusion by the authors based on the multivocal literature review, telemetry findings, conceptual modeling (PRP/SGM), and the four-month pilot evaluation.
These conflicting findings constitute the Productivity-Reliability Paradox (PRP): a systematic phenomenon emerging from non-deterministic code generators and insufficient specification discipline.
Conceptual synthesis and interpretation by the paper's authors, based on the multivocal literature review, telemetry, and experimental evidence summarized above.
Telemetry across 10,000+ developers shows 91% longer code review times.
Observational telemetry data aggregated across >10,000 developers reported in the paper; metric reported is percent increase in review time.
The most rigorous randomized controlled trial (RCT) documents a 19% slowdown for experienced developers.
A single RCT cited in the paper described as the most rigorous trial; result reported as a 19% slowdown for experienced developers. Sample size for the RCT is not provided in the summary statement.
Making LLMs themselves explicitly Bayesian belief-updating engines remains computationally intensive and conceptually nontrivial as a general modeling target
Stated as a limitation in the paper (conceptual and computational argument); no benchmarks or computational cost measurements reported.
Keeping humans in the loop can sometimes make the decision worse.
Argumentative/diagnostic statement in the paper (theoretical assertion; no experimental or observational effect sizes reported in the excerpt).
Leaders may believe oversight remains meaningful when it has become ceremonial.
Conceptual warning in the paper about erosion of meaningful oversight (no empirical validation provided in the excerpt).
The central risk is misrecognition: leaders may keep a human-centered story in place after decision-shaping authority has shifted elsewhere (e.g., to AI).
Analytic/diagnostic claim in the paper (conceptual warning; no empirical sample or measured incidence provided).
Reactive approaches paired with automation or creation produced breakdowns (reduced effectiveness).
Thematic evidence from interviewees describing instances where reactive leadership combined with high automation-or-creation use led to coordination or accountability breakdowns across the 34 cases.
Suppression bias is the systematic suppression of correct-but-difficult recommendations when clinician capability falls below the execution threshold.
Definition and characterization of a proposed failure mode provided in the paper (conceptual/theoretical).
Existing approaches, runtime guardrails, training-time alignment, and post-hoc auditing treat governance as an external constraint rather than an internalized behavioral principle, leaving agents vulnerable to unsafe and irreversible actions.
Author's conceptual/literature critique presented in the paper (argumentative claim, no empirical sample or experiment reported for this statement).