Critique queue · as of 2026-06-17
Which papers to critique next
Selection is an editorial act, so it is made auditable. This queue ranks recently-published, top-tier social-science papers on AI and AGI by a score that re-derives in-app from each candidate’s signals — weighted toward impact (is the claim load-bearing) and value-add (how far the claim outruns its evidence), then recency and journal tier. Critiquing a paper is not an accusation: strong, influential papers are in scope precisely because they shape decisions.
- 1Tier SInformation systems · 2026-05-08 · 41d old
Unraveling Generative AI from a Human Intelligence Perspective: A Battery of Experiments
Wen Wang, Siqi Pei, Tianshu Sun · Information Systems Research
The headline claim that 'GPT-4 outperforms humans in cognitive, emotional, and creative intelligence' anthropomorphizes benchmark scores into human-like 'intelligence' and then leaps to forecasting job-level labor-market impacts for policymakers, an overreach built on one model and human-derived tests of uncertain construct validity.
Human–AI interactionLabour marketsForesight & AGI transitionpolicy prescription beyond evidenceAGI relevance asserted, not testedbroad generalisation from a narrow settingreproducibility riskScore89Impact ·30%100Value-add ·30%70Recency ·20%88Tier ·20%100 - 2Tier AScience & technology studies · 2026-05-30 · 19d old
From rule of law to rule of algorithm: Generative Artificial Intelligence's threat to democracy
A.T. Kingsmith · Big Data & Society
The sweeping claim that GenAI marks a 'qualitative break' replacing rule of law with rule of algorithm and threatening democracy is a commentary asserted by argument rather than evidence, and judges the EU AI Act 'insufficient' without empirical support.
AI governanceLaw & regulationPolitical economypolicy prescription beyond evidenceAGI relevance asserted, not testedbroad generalisation from a narrow settingScore83Impact ·30%100Value-add ·30%60Recency ·20%100Tier ·20%75 - 3Tier SManagement & organisation · 2026-04-27 · 52d old
More Versus Better: Artificial Intelligence, Incentives, and the Emerging Crisis in Peer Review
Claudine Madras Gartenberg, Sharique Hasan, Alex Murray, Lamar Pierce · Organization Science
The high-profile claim that AI drove a 42% submission surge and declining writing quality rests on single-journal before/after data with AI authorship inferred from detectors, so the causal attribution to AI and the generalisation to all of science (via editor 'conversations') outrun the evidence.
Knowledge productionLabour marketscausal claim on observational designbroad generalisation from a narrow settingreproducibility riskScore82Impact ·30%100Value-add ·30%55Recency ·20%76Tier ·20%100 - 4Tier AScience & technology studies · 2026-05-26 · 23d old
The rise of AI sovereignty: Authoritarian technological imaginaries as a form of reflexive control
Gregory Asmolov · Big Data & Society
The strongest claim that Russian 'reflexive control' shapes global AI-governance debates and can distinguish legitimate Global South concerns from authoritarian influence rests on a single-actor reading of presidential statements, so the normative policy prescription overreaches a thin, non-reproducible evidentiary base.
AI governancePolitical economySurveillance, security & policingpolicy prescription beyond evidencebroad generalisation from a narrow settingreproducibility riskScore80Impact ·30%100Value-add ·30%50Recency ·20%100Tier ·20%75 - 5Tier SManagement & organisation · 2026-06-09 · 9d old
Artificial Collusion: Examining Supracompetitive Pricing by Q-Learning Algorithms
Arnoud den Boer, Janusz M Meylahn, Maarten Pieter Schinkel · Management Science
Concluding that competition agencies need not be suspicious of pricing algorithms rests on debunking one algorithm class (Q-learning) under specific synchronization conditions, a reassuring policy inference that may not generalize to other RL pricing methods.
Law & regulationAI governanceInnovation, productivity & competitionpolicy prescription beyond evidencebroad generalisation from a narrow settingScore80Impact ·30%92Value-add ·30%40Recency ·20%100Tier ·20%100 - 6Tier SManagement & organisation · 2026-06-12 · 6d old
The Cybernetic Teammate: A Field Experiment on Generative AI and Teamwork
Fabrizio Dell’Acqua, Charles Ayoubi, Hila Lifshitz‐Assaf, Raffaella Sadun et al. · Organization Science
The headline claim that 'individuals with AI matched teams without AI' is causally clean given the preregistered RCT, but generalizing a single-firm (P&G) product-innovation task to knowledge work broadly is the likely weak point.
Labour marketsHuman–AI interactionInnovation, productivity & competitionbroad generalisation from a narrow settingScore74Impact ·30%100Value-add ·30%15Recency ·20%100Tier ·20%100 - 7Tier SInformation systems · 2026-04-30 · 49d old
How Costs Influence Preferences for Control in Generative Artificial Intelligence (GenAI): Human-Guided vs. GenAI-Based Delegated Search
Lei Wang, Ho Cheung Brian Lee · Information Systems Research
The claim that salient usage costs make users more deliberate cocreators and raise service value is drawn from observational analysis of 1.8 million prompts where cost salience is likely confounded with user type and task, so the causal interpretation of charging-improves-outcomes is fragile.
Human–AI interactionPolitical economycausal claim on observational designbroad generalisation from a narrow settingScore74Impact ·30%84Value-add ·30%45Recency ·20%79Tier ·20%100 - 8Tier AScience & technology studies · 2026-06-01 · 17d old
Generative AI, propaganda, and digital authoritarianism: Comparative insights from six democratically weakened countries
Gabrielle D. Beacken, Inga K Trauthig, Samuel Woolley · Big Data & Society
The argument that GenAI propaganda spreads 'digital authoritarian tactics' into semi-democracies is built from 93 interviews across six purposively chosen countries, so the comparative generalization and qualitative-coding reproducibility are the soft spots.
Surveillance, security & policingPolitical economyAI governancebroad generalisation from a narrow settingreproducibility riskScore73Impact ·30%100Value-add ·30%25Recency ·20%100Tier ·20%75 - 9Tier SInformation systems · 2026-05-21 · 28d old
Can ChatGPT Kill User-Generated Q&A Platforms?
Junzhi Xue, Lizheng Wang, Jinyang Zheng, Yongjun Li et al. · Information Systems Research
The headline causal claim that ChatGPT cut Stack Overflow question volume by ~14% (up to 27.9%) is identification-dependent on isolating the LLM shock from confounding platform trends, and the 'niche partitioning not displacement' conclusion generalizes from one platform to knowledge ecosystems broadly.
Labour marketsKnowledge productionInnovation, productivity & competitionbroad generalisation from a narrow settingScore72Impact ·30%92Value-add ·30%15Recency ·20%100Tier ·20%100 - 10Tier SMarketing · 2026-05-05 · 44d old
Made With AI: Consumer Engagement with Social Media Containing AI Disclosures
Stephan Carney, Ignacio Riveros, Stephanie Tully · Journal of Consumer Research
The central finding that AI-disclosure labels reduce engagement via lowered parasocial connection is well-identified via experiments, but the implied policy guidance that effort-signaling disclosures should be used to 'mitigate' engagement loss edges toward prescribing how to blunt the very transparency the disclosures exist to provide.
Human–AI interactionLaw & regulationpolicy prescription beyond evidenceScore71Impact ·30%88Value-add ·30%25Recency ·20%84Tier ·20%100 - 11Tier SManagement & organisation · 2026-05-04 · 45d old
Backfiring AI? AI Deployment in Workplace
Di Yuan, Manmohan Aseri, Narayan Ramasubbu · Management Science
The claim that AI-facilitated knowledge transfer can backfire and lower firm productivity is derived from a stylized game-theoretic model, yet the abstract issues concrete managerial policy recommendations as if empirically validated.
Labour marketsPolitical economypolicy prescription beyond evidencebroad generalisation from a narrow settingScore70Impact ·30%72Value-add ·30%40Recency ·20%83Tier ·20%100 - 12Tier SInformation systems · 2026-05-08 · 41d old
When Influencers Delegate Replies: How Social AI Agents Shape User Engagement
Maggie Mengqing Zhang, Yang Gao, Jingjing Li, Steven L. Johnson · Information Systems Research
The strongest claim that delegating replies to an LLM agent causally increases user engagement rests on a single-platform staggered DiD whose 'AI reply received' assignment is likely endogenous to influencer behavior, and the proprietary, undisclosed-platform data makes the effect hard to reproduce or generalize.
Human–AI interactionLabour marketsbroad generalisation from a narrow settingreproducibility riskScore70Impact ·30%84Value-add ·30%25Recency ·20%88Tier ·20%100 - 13Tier SManagement & organisation · 2026-05-06 · 43d old
Artificial intelligence adoption and the demand for managerial expertise
Liudmila Alekseeva, José Azar, Mireia Giné, Sampsa Samila · Strategic Management Journal
The headline claim that AI adoption reconfigures managerial roles toward interpersonal/growth skills rests on cross-sectional associations between job-posting-derived adoption measures and vacancy composition, yet the framing implies AI is causing the skill shift rather than co-varying with unobserved firm strategy.
Labour marketsInnovation, productivity & competitioncausal claim on observational designScore70Impact ·30%80Value-add ·30%30Recency ·20%86Tier ·20%100 - 14Tier SMarketing · 2026-05-05 · 44d old
Leveraging Generative Artificial Intelligence to Create Visual Content in Digital Advertising
Remi Daviet, Yohei Nishimura · Marketing Science
The claim that a Bayesian active-learning GenAI framework efficiently discovers effective brand-aligned ad designs is asserted in a one-sentence abstract with no reported sample, baseline, or effect size, making both the generalisation and reproducibility unverifiable.
Innovation, productivity & competitionHuman–AI interactionbroad generalisation from a narrow settingreproducibility riskScore67Impact ·30%76Value-add ·30%25Recency ·20%84Tier ·20%100 - 15Tier ACommunication & media · 2026-06-01 · 17d old
The politics of artificial intelligence alignment: Public reactions to AI moderation in the case of Google’s Gemini
Adrian Rauchfleisch, Andreas Jungherr · New Media & Society
The claim that a visible AI product failure (Gemini images) shifts public support for AI moderation is from a preregistered experiment, but one of two stimulus sets failed to reach significance, so generalizing the 'focusing event' effect beyond this single controversy is fragile.
AI governanceHuman–AI interactionLaw & regulationbroad generalisation from a narrow settingScore67Impact ·30%92Value-add ·30%15Recency ·20%100Tier ·20%75 - 16Tier ACommunication & media · 2026-05-04 · 45d old
Refusal as silence: Gendered disparities in Vision-Language Model responses
Sha Luo, S Kim, Zening Duan, Kaiping Chen · New Media & Society
The finding that GPT-4V refuses transgender/non-binary personas at higher rates is a striking fairness claim, but it rests on a single proprietary model and a constrained counterfactual-persona task whose generalisability across models, versions, and prompt designs is unestablished.
Inequality, bias & fairnessHuman–AI interactionbroad generalisation from a narrow settingreproducibility riskScore67Impact ·30%92Value-add ·30%25Recency ·20%83Tier ·20%75 - 17Tier AScience & technology studies · 2026-05-20 · 29d old
Making GenAI valuable: Benchmarks, singularities, and the enrichment economy
Claudia Aradau, Tobias Blanke · Big Data & Society
The strongest claim that benchmarks 'singularise' LLMs by situating them near a 'future perfect of AGI' treats AGI as an assumed horizon mobilized for valuation rather than a tested construct, so the AGI framing is asserted and theory-driven rather than evidenced.
Political economyKnowledge productionForesight & AGI transitionAGI relevance asserted, not testedScore66Impact ·30%84Value-add ·30%20Recency ·20%100Tier ·20%75 - 18Tier ACommunication & media · 2026-06-04 · 14d old
AI meets politics: Examining the effects of different targeting strategies across 15 countries
Sanne Kruikemeier, Svenja Schäfer, Alice Hamilton, Puck Guldemond et al. · New Media & Society
The claim that AI-generated targeting only persuades along pre-existing political orientation (not age/personality) is experimentally grounded, but the null results on personality targeting risk being read as 'AI microtargeting doesn't work,' an overgeneralization from one EU-election survey-experiment.
Political economyHuman–AI interactionSurveillance, security & policingbroad generalisation from a narrow settingScore66Impact ·30%88Value-add ·30%15Recency ·20%100Tier ·20%75 - 19Tier ACommunication & media · 2026-05-26 · 23d old
Beyond disruption and invisibility: Interactional continuity in everyday AI use in India
Emilia Edwards, Dhiraj Murthy · New Media & Society
The 'interactional continuity' framing of everyday AI use is conceptually central but rests on 28 interviews in a single Indian workplace with AI-assisted coding, so the Global South generalization and reproducibility of the qualitative inference are the weak points.
Human–AI interactionLabour marketsKnowledge productionbroad generalisation from a narrow settingreproducibility riskScore65Impact ·30%76Value-add ·30%25Recency ·20%100Tier ·20%75 - 20Tier AScience & technology studies · 2026-05-22 · 27d old
Crafting computer vision through human eyes: An AI laboratory ethnography
Luqing Zhou · Big Data & Society
The strongest claim that CV knowledge production is fundamentally a sensory, processual system generalizes from one 9-month single-lab ethnography, so the leap to AI epistemics writ large and the non-replicable design are its weak points.
Knowledge productionHuman–AI interactionbroad generalisation from a narrow settingreproducibility riskScore65Impact ·30%76Value-add ·30%25Recency ·20%100Tier ·20%75 - 21Tier AScience & technology studies · 2026-05-19 · 30d old
From prompt engineering to prompt design: Research strategies for visual generative AI
Gabriele Colombo, Sabine Niederer, Carlo De Gaetano · Big Data & Society
The strongest claim that prompt-design strategies reveal stable model 'house styles' and biases rests on a single biodiversity case demo across a couple of years, so the generalization to a method for auditing generative AI broadly and the non-systematic, hard-to-reproduce setup are the weak points.
Knowledge productionHuman–AI interactionInequality, bias & fairnessbroad generalisation from a narrow settingreproducibility riskScore65Impact ·30%76Value-add ·30%25Recency ·20%100Tier ·20%75 - 22Tier AScience & technology studies · 2026-05-10 · 39d old
Into the black box: Laypeople's folk theories about generative artificial intelligence chatbots
Li Z, Nuri Kim, L Chen · Big Data & Society
The strongest claim that laypeople's folk theories of GenAI chatbots fall into three constructed areas that shape interaction strategies generalizes from a handful of focus groups, so the leap to 'laypeople' broadly and the non-reproducible qualitative base are the weak points.
Human–AI interactionKnowledge productionbroad generalisation from a narrow settingreproducibility riskScore63Impact ·30%76Value-add ·30%25Recency ·20%90Tier ·20%75 - 23Tier AScience & technology studies · 2026-05-15 · 34d old
Working the algorithm: Contextual skills of on-demand gig workers
Xinyi Hong, Xinyi Cheng, Dong Liu · Big Data & Society
The strongest claim that gig workers cultivate a nine-indicator framework of 'algorithmic skills' affording meaningful agency is built from just 20 interviews, so the generalization of a tidy taxonomy and the contention that it outperforms the control-resistance framework outrun the small, non-reproducible sample.
Labour marketsHuman–AI interactionbroad generalisation from a narrow settingreproducibility riskScore63Impact ·30%72Value-add ·30%25Recency ·20%96Tier ·20%75 - 24Tier ACommunication & media · 2026-05-03 · 46d old
Being literate, behaving literate? A mixed-methods approach to adolescents’ algorithm literacy and behavioral strategies on social media
Larissa Leonhard, Ruth Wendt, Claudia Riesmeyer · New Media & Society
The claim that higher algorithm knowledge correlates with reduced interactive behavior is presented with a causal mechanism ('reluctance to engage with profiling') from a cross-sectional survey of German 14-17 year-olds, conflating correlation with explanation and limiting generalisation beyond that cohort.
EducationHuman–AI interactioncausal claim on observational designbroad generalisation from a narrow settingScore63Impact ·30%60Value-add ·30%45Recency ·20%82Tier ·20%75 - 25Tier ACommunication & media · 2026-04-29 · 50d old
Charismatic machines: On the epistemic power of generative AI within platform convergence
Mauro Barisione · New Media & Society
The central claim that generative AI wields 'epistemic power' as 'charismatic machines' perceived as both human-like and superhuman is a Bourdieu/Weber-grounded theoretical construct asserted rather than empirically tested, attributing authority-conferring effects to AI by analogy.
Knowledge productionHuman–AI interactionAGI relevance asserted, not testedScore62Impact ·30%84Value-add ·30%20Recency ·20%78Tier ·20%75 - 26Tier AScience & technology studies · 2026-05-20 · 29d old
Algorithmic responsibility in PPC practice: Interpreting black boxes in digital advertising work
Natalia Chrobak · Big Data & Society
The strongest claim that PPC work is shifting from a technical to an interpretive profession under algorithmic opacity generalizes a broad theory of 'algorithmic responsibility' from 27 interviews and forum posts, making the leap to the wider data economy and the non-reproducible qualitative base its weak points.
Labour marketsHuman–AI interactionbroad generalisation from a narrow settingreproducibility riskScore62Impact ·30%64Value-add ·30%25Recency ·20%100Tier ·20%75 - 27Tier ACommunication & media · 2026-05-19 · 30d old
Resilience and disempowerment in algorithmic systems
Samantha M. Jones, Erin A. Heerey · New Media & Society
The strongest claim that adaptive recommendation algorithms cause more homogeneous user selections comes from a small (N=263) controlled feed experiment, so generalizing from a stylized lab feed to real social-media algorithmic influence on agency is the weak point.
Human–AI interactionKnowledge productionbroad generalisation from a narrow settingScore61Impact ·30%72Value-add ·30%15Recency ·20%100Tier ·20%75 - 28Tier SManagement & organisation · 2026-06-05 · 13d old
Markovian Search with Ex Ante Constraints: Theory and Applications to Socially Aware Algorithmic Hiring
Mohammad Reza Aminian, Vahideh Manshadi, Rad Niazadeh · Management Science
The claim that ex ante fairness constraints yield 'socially desirable outcomes' in algorithmic hiring is an optimization-theory result whose normative weak point is equating demographic-parity-style constraints with genuine equity, demonstrated only via a numerical study.
Inequality, bias & fairnessLaw & regulationScore58Impact ·30%60Value-add ·30%0Recency ·20%100Tier ·20%100 - 29Tier AScience & technology studies · 2026-05-04 · 45d old
The pragmatic frames of spurious correlations in machine learning: Interpreting how and why they matter
Samuel J. Bell, Skyler Wang · Big Data & Society
The claim that ML researchers judge spuriousness through four 'pragmatic frames' is an interpretive taxonomy drawn from an unspecified 'broad survey' of literature, so the framework's completeness and coding reliability are not demonstrable from the abstract.
Knowledge productionInequality, bias & fairnessreproducibility riskScore54Impact ·30%64Value-add ·30%10Recency ·20%83Tier ·20%75
How the score works
Each 0–100 score is 30% impact (AI/AGI centrality + societal salience), 30% value-add (the claim–evidence fragility flags below), 20% recency (full for the first month, decaying to zero by four), and 20% journal tier. Value-add flags and their weights: causal claim on observational design (30%); policy prescription beyond evidence (25%); AGI relevance asserted, not tested (20%); broad generalisation from a narrow setting (15%); reproducibility risk (10%).
Candidates are drawn only from 137 monitored top-tier venues across 12 social-science fields — see the source-journal scope. A place in the queue is not a verdict; it is a transparent ranking of where a critique would add the most.
Most of these venues are paywalled, so a critique defaults to the abstract (severity capped at moderate). For a queued paper, lawful full text can be manually extracted and provided by someone with access — lifting it to a full-text–calibrated critique. See full-text provision.