Evidence (8974 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
10085 claims
Filter claims →
Productivity
8974 claims
Filtered →
Governance
8062 claims
Filter claims →
Human-AI Collaboration
7749 claims
Filter claims →
Org Design
5057 claims
Filter claims →
Innovation
4896 claims
Filter claims →
Labor Markets
4088 claims
Filter claims →
Skills & Training
3372 claims
Filter claims →
Inequality
2377 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 | 882 | 244 | 117 | 1097 | 2424 |
| Governance & Regulation | 1010 | 469 | 229 | 135 | 1875 |
| Organizational Efficiency | 977 | 235 | 149 | 90 | 1462 |
| Technology Adoption Rate | 781 | 299 | 143 | 128 | 1362 |
| Research Productivity | 506 | 155 | 74 | 363 | 1110 |
| Output Quality | 555 | 219 | 71 | 70 | 915 |
| Decision Quality | 395 | 200 | 95 | 54 | 751 |
| Firm Productivity | 523 | 67 | 101 | 27 | 724 |
| AI Safety & Ethics | 262 | 309 | 75 | 36 | 688 |
| Market Structure | 195 | 201 | 135 | 30 | 566 |
| Task Allocation | 248 | 77 | 96 | 38 | 464 |
| Innovation Output | 300 | 34 | 55 | 20 | 411 |
| Skill Acquisition | 207 | 75 | 65 | 21 | 368 |
| Employment Level | 138 | 67 | 119 | 24 | 350 |
| Fiscal & Macroeconomic | 156 | 80 | 53 | 33 | 329 |
| Task Completion Time | 211 | 38 | 13 | 16 | 280 |
| Firm Revenue | 183 | 52 | 29 | 5 | 270 |
| Consumer Welfare | 131 | 77 | 48 | 13 | 269 |
| Inequality Measures | 50 | 141 | 54 | 9 | 254 |
| Worker Satisfaction | 104 | 85 | 25 | 13 | 227 |
| Error Rate | 87 | 112 | 11 | 5 | 215 |
| Automation Exposure | 69 | 69 | 37 | 20 | 198 |
| Wages & Compensation | 102 | 49 | 31 | 11 | 193 |
| Team Performance | 115 | 30 | 30 | 11 | 187 |
| Regulatory Compliance | 88 | 74 | 17 | 7 | 186 |
| Training Effectiveness | 109 | 22 | 14 | 21 | 168 |
| Developer Productivity | 116 | 21 | 15 | 8 | 161 |
| Job Displacement | 12 | 92 | 26 | 1 | 131 |
| Hiring & Recruitment | 57 | 12 | 9 | 5 | 83 |
| Skill Obsolescence | 6 | 59 | 10 | 2 | 77 |
| Social Protection | 43 | 17 | 8 | 2 | 70 |
| Creative Output | 35 | 21 | 9 | 4 | 70 |
| Labor Share of Income | 18 | 23 | 17 | 1 | 59 |
| Worker Turnover | 15 | 16 | — | 4 | 35 |
| Industry | — | — | — | 1 | 1 |
Productivity
Remove filter
The study's synthesis contributes to the Industry 5.0 conversation and provides a blueprint for organizations, educators, and policymakers to help ensure training programs meet the needs of warehouse automation.
Author assertion based on the secondary data review of literature and industry reports from 2022–2026; presented as contribution/implication rather than an empirical measurement; no sample size reported.
Structured reskilling programs, human-centric system design, deliberate role enrichment, and participatory governance are strategic recommendations to address workforce transformation in AI-driven logistics environments.
Conclusions and recommendations from the paper's secondary data review of peer-reviewed research and industry evidence (2022–2026). These are prescriptive recommendations rather than outcomes from a new empirical test; no sample size provided.
Successful warehouse human-robot collaboration (HRC) requires a portfolio of multi-dimensional competencies, including technical skills in robotic systems, cognitive and supervisory skills, communication and teamwork, and adaptive learning.
Secondary data literature review of peer-reviewed research and industry evidence published 2022–2026 (method: secondary data review / synthesis). No primary sample size reported in the paper.
The true energy bottleneck in edge VLM inference is not what the model sees (visual input), but how much it says (output length).
Synthesis of profiling results: power invariance, much higher per-token cost for outputs, and large energy variation driven by output length across models and images.
The variation in energy with image complexity arises not from increased visual processing cost, but from differences in output length.
Analysis linking image complexity to longer model outputs (more output tokens) and showing power invariance; hence increased energy follows increased decode time/output token count rather than increased per-token visual processing power.
Image complexity, measured by the number of objects in an image, induces up to 4.1x energy differences at identical resolution.
Profiling across images varying in object count at the same spatial resolution; reported energy differences up to 4.1× correlated with image complexity.
Each output token costs 11 to 39x more wall-clock time than each input token due to the compute-bound and memory-bound asymmetry between prefill and decode, making output token count the dominant driver of both latency and energy.
Measurement of per-token wall-clock time for prefill (input tokens) vs. decode (output tokens) across the profiled models and hardware platforms; observed ratios reported as 11–39×.
All energy variation across inputs must arise from variation in inference time, not from variation in power draw.
Inference power shown to be invariant (<5% variation) across inputs in the same profiling study, implying energy differences (energy = power × time) are due to time differences rather than power differences.
Small open economies should not maximise AI adoption as an isolated target; they should build institutional absorptive capacity that converts AI exposure into productivity, worker mobility, and shared prosperity.
Policy implication directly drawn from the DIAC theoretical framework and its derived propositions (analytical/recommendation).
Framing organizations as systems built on accumulated experience provides practical guidance for responsible AI integration.
Conceptual argument in the paper proposing that organizational experience should guide AI design and deployment decisions; illustrative examples provided.
A five-part Human–AI Collaboration Framework can help organizations gain efficiency from AI while keeping human judgment active and accountable in key HR decisions.
Authors' proposed framework and prescriptive analysis; theoretical argumentation rather than direct experimental validation.
People use prior experience to interpret context, notice subtle cues, and make sense of ambiguous situations—capabilities that differ fundamentally from how large language models process data.
Synthesis of research from cognitive science and neuroscience showing mechanisms of expertise and contextual interpretation; compared conceptually to how large language models operate.
By offering a structured typology and a processual conceptual framework, this study provides a foundation for theory-advancing research and for more explicit and actionable managerial practice.
Authors' asserted contribution based on their synthesis (N=68) and the conceptual outputs (typology and framework); positioned as implication and contribution to research and practice.
We develop a processual framework that specifies how and when GenAI is embedded within organizational decision-making processes, delineating how generative applications support, augment, or co-perform decision-making activities.
Conceptual framework produced by authors based on synthesis of reviewed studies (N=68); framework distinguishes embedding modes (support, augment, co-perform).
Building on the harmonization and translation of these tasks, we propose a typology comprising six active GenAI roles and one collaborative human-AI role.
The authors' conceptual typology derived from synthesis of the literature (N=68) identifying role types; counts reported as six active roles plus one collaborative role.
The study maps these tasks and categories onto six recursive decision-making components: attention, intelligence, design, choice, implementation, and feedback.
Conceptual mapping developed by the authors from the synthesized tasks and categories identified across the 68 reviewed publications.
Those 53 tasks are aggregated into 18 task categories.
Author synthesis and categorization based on the set of identified tasks from the literature review (68 publications).
The study identifies 53 tasks performed by generative applications.
Synthesis of the 68 reviewed publications; authors enumerate 53 distinct tasks attributed to GenAI applications.
This study conducts a systematic literature review that identifies 68 relevant publications.
Systematic literature review conducted by the authors; explicit count of included publications reported as 68.
Successful AI adoption depends on organizational readiness, leadership commitment, technological infrastructure, and governance frameworks.
Synthesis of determinants reported across the 22 studies in the systematic review (PRISMA protocol; sources: Scopus, ScienceDirect, Google Scholar; 2017–2026); thematic analysis identified these recurring enablers.
Generative AI and large language models are emerging as transformative tools for market intelligence and strategic insight generation.
Observation from the systematic review's thematic analysis of the 22 included studies and discussion of recent technological trends (search period 2017–2026 across Scopus, ScienceDirect, Google Scholar).
Business Intelligence (BI) and Decision Support Systems (DSS) enhance managerial decision speed and accuracy.
Thematic synthesis from the 22 studies identified and reviewed under PRISMA across Scopus, ScienceDirect, Google Scholar (2017–2026).
AI-powered CRM and predictive analytics systems improve marketing effectiveness and customer engagement.
Synthesis of findings from the 22 studies included in the systematic review (PRISMA-guided search across Scopus, ScienceDirect, Google Scholar, 2017–2026) using thematic analysis.
AI-aided Strategic Information System (SIS) tools significantly enhance organizational competitiveness by enabling data-driven decision-making, improving customer intelligence, optimizing supply chain performance, and strengthening strategic agility.
Systematic review following PRISMA of 22 empirical/theoretical studies identified via searches on Scopus, ScienceDirect, and Google Scholar for 2017–2026; data were synthesized thematically.
Equity-by-design, explainable architecture, federated infrastructure, and modernized statutory framework are necessary to achieve responsible AI adoption in tax compliance systems.
Recommendations synthesized from reviewed literature and governance analysis across the 37 included sources advocating technical, infrastructural, and legal interventions (equity-by-design, XAI, federated approaches, statutory modernization).
Governance alignment with NIST AI RMF 1.0, EO 14110, and 26 U.S.C. § 6103 is critical for responsible AI adoption in the U.S. tax system.
Policy-analysis synthesis in the literature review; the paper argues alignment with existing U.S. frameworks and statute (NIST AI Risk Management Framework, Executive Order 14110, and tax confidentiality statute 26 U.S.C. § 6103) is necessary for compliance, privacy, and risk management.
Machine learning models reduce forecasting MAPE by 15–30 percent compared to legacy systems.
Summary result from the literature review synthesizing studies on revenue/collection forecasting performance where ML/AI models report lower Mean Absolute Percentage Error (MAPE) relative to legacy forecasting approaches.
Machine learning models decrease audit no-change rates by an estimated 15–20 percentage points compared to legacy systems.
Aggregated/summary finding from the reviewed literature (2020–2026) comparing ML-based audit selection to legacy audit selection systems; specific studies and metrics synthesized in the review.
AI and predictive analytics have a transformational opportunity to enhance compliance risk scoring, audit selection, revenue forecasting, and fraud detection at the IRS.
Synthesis from the organized literature review of 37 sources (searches on SSRN, Google Scholar, Web of Science, Scopus, and government repositories covering 2020–2026) reporting applications and performance gains from ML/AI in tax/compliance contexts.
This study is among the first to empirically integrate UTAUT, DOI, and RBV into a unified AI-adoption framework in hospitality and tourism, demonstrating how adoption extends beyond consumer intentions to generate strategic organizational outcomes.
Authors' stated contribution; paper presents conceptual integration and empirical tests using 8 expert interviews and a 499-respondent survey.
Managers should enhance personalization, simplify user experiences, invest in employee training, strengthen cybersecurity, and leverage data analytics to improve performance, reduce costs, and support sustainable operations.
Recommendations derived from study findings (qualitative interviews + survey of 499 AI-aware consumers) linking strategic actions to higher adoption and better organizational outcomes.
The study integrates UTAUT, DOI, and RBV into a holistic framework linking micro-level acceptance drivers to macro-level strategic outcomes, extending traditional technology acceptance models toward a process-oriented understanding of AI adoption in hospitality and tourism.
Conceptual integration described in the paper and empirically tested using the mixed-methods design (8 expert interviews; survey of 499 consumers).
AI adoption significantly enhances organizational outcomes, including technology management, sustainability, and cost efficiency (operational efficiency).
Quantitative links reported between consumer acceptance/AI adoption and organizational outcome measures in the survey of 499 AI-aware consumers; qualitative interview data support interpretation.
Intervening factors, including cybersecurity concerns and financial barriers, also show positive associations with adoption, indicating consumers perceive these risks as manageable trade-offs.
Survey analysis of 499 AI-aware consumers reporting positive associations between intervening factors (cybersecurity/financial concerns) and AI adoption; qualitative interviews informed interpretations.
Strategic actions — such as training, empowerment, and data analytics — further strengthen AI adoption.
Survey of 499 AI-aware consumers, informed by grounded-theory coding of 8 expert interviews; results indicate strategic actions positively moderate/augment adoption.
Contextual and causal conditions positively influence AI adoption.
Quantitative survey analysis of 499 AI-aware consumers (informed by 8 expert interviews); authors report positive associations between contextual/causal conditions and adoption.
This study develops and validates a multilevel framework explaining how contextual, causal, intervening, and strategic conditions shape AI adoption in hospitality and tourism.
Sequential mixed-methods design (Qual → Quan): eight expert interviews (grounded-theory coding) informed survey design; survey of 499 AI-aware consumers used to test/validate the framework.
As a secondary contribution, the authors offer the underlying LLM-assisted, grey-literature theory-building method as a scalable template for software-engineering research, with a public implementation.
Paper reports method and claims a public implementation; stated in abstract.
The theory makes competing positions explicit and turns 'AI is changing code review' into falsifiable propositions with named constructs and moderators.
Construction of a causal model from coded practitioner discourse and explicit naming of constructs and moderators; claimed in abstract.
Agent-authored pull requests are merged several times faster than human-authored ones.
Observational analysis of public GitHub activity reported in the paper (no sample size reported in abstract); reported comparison of time-to-merge between agent-authored and human-authored PRs.
The decisive lever against token maxing is the harness: the orchestration layer that assembles context, exposes tools, sequences turns, delegates work, and carries enterprise observability and governance.
Argument/claim presented in the paper based on empirical results from the controlled swap and conceptual analysis (formalization of token economics and description of mechanism families); not a directly quantified experimental outcome.
Task-completions per million tokens rise from 54.9 to 92.0 when using the harness.
Reported throughput metric (completions per million tokens) before and after harness in the controlled swap experiment.
Quality per dollar rises 82% with the harness.
Reported aggregate metric from the controlled swap comparing quality-per-dollar before and after using the Writer Agent Harness across the evaluated tasks/models.
A model's quality gain from the harness correlates almost perfectly with its baseline strength (r=0.99, n=6) — a phenomenon we term 'harness leverage'.
Correlation reported across six foundation models comparing baseline model strength to measured quality gain under the harness (r=0.99, n=6).
Efficiency is model-invariant — every model gets cheaper (33-61%).
Observed cost reductions for each of six foundation models (listed: Claude Sonnet 4.6, Gemini 3.1, Gemini Flash 3.5, Qwen 3.6, GLM 5.1, Palmyra X6) in the controlled swap; reported per-model percent reductions range 33%–61%.
Holding models constant, the harness reduces tokens per task 38% (14.2k->8.8k).
Same controlled swap experiment with 22 tasks and six models; reported tokens per task 14.2k vs 8.8k.
Holding models constant, the harness reduces median wall-clock time per task 44% (48s->27s).
Same controlled swap experiment over 22 locked evaluation tasks and six foundation models; reported median wall-clock times 48s vs 27s.
Holding models constant, the harness cuts blended cost per task 41% ($0.21->$0.12).
Controlled swap experiment changing only the orchestration layer (Writer Agent Harness vs frozen conventional production loop) across 22 locked evaluation tasks and six foundation models; reported blended cost per task values $0.21 and $0.12.
In some cases the digital sector leads in productivity growth rates.
Observation based on BEA–BLS industry-level data for 63 U.S. industries from 1997–2023, as reported in the paper.
The digital sector contributes twice as much labor input to output growth (relative to the physical sector).
Empirical statement based on BEA–BLS data covering 63 U.S. industries (1997–2023) as reported in the paper.