AI-native critique journal · powered by the AGI Social Scientist
Critical AI
Harsh on claims. Precise on evidence. Cautious about motives.
A critical companion to the scholarly record. Critical AI publishes structured, evidence-linked, author-contestable critiques of social-science research on AI and AGI — the research that shapes policy, regulation, labour-market expectations and public debate. The critique pieces are generated by the AGI Social Scientist, a research engine that enforces provenance in code (every claim cites a verbatim source excerpt), then approved by a named human editor. We critique claims, methods and evidence— never authors’ motives.
Status: founding pilot. The platform, the multi-agent workflow and the public audit trail are live. The standing, multi-disciplinary editorial board named above is being recruited; until it is ratified, critiques are approved under a named founding-editor review, carry the full transparency record, and remain open to author reply and public correction. This status is disclosed on every critique rather than implied.
4
evidence reviews
124
papers coded
87
evidence gaps named
1
paper critiques
1
high+ severity
100%
counts re-derivable
The editorial standard
Publish fast. Critique harshly. Verify before release. Correct visibly. Never speculate about motive or misconduct without evidence.
Evidence reviews
JSON index ↗AGISS scoping reviews / evidence-gap-maps over a body of AI-governance literature: every paper coded into a matrix, every empty cell named, every limitation disclosed, and every count re-derivable offline. Frameworks are proposed far more often than they are tested — these maps show exactly where.
- Evidence reviewEvidence Gap MapCR-REV-001
Evidence gap map: AI governance frameworks in the scholarly literature
Across a 5x6 framework matrix, 21 of 30 cells have no papers in this corpus of 16 included records; the populated cells concentrate in aspect 'conceptual_governance_model' (6 papers) and evidence type 'conceptual_normative' (8 papers).
16 papers coded21 empty cells (70%)13 excludedAI governanceKnowledge production - Evidence reviewEvidence Gap MapCR-REV-002
Evidence gap map: data privacy in the AI-governance literature
Across a 5x6 framework matrix, 24 of 30 cells have no papers in this corpus of 12 included records; the populated cells concentrate in aspect 'health_data_privacy_governance' (5 papers) and evidence type 'conceptual_normative' (3 papers).
12 papers coded24 empty cells (80%)20 excludedLaw & regulationSurveillance, security & policing - Evidence reviewEvidence Gap MapCR-REV-003
Evidence gap map: data protection regulation in the AI literature
Across a 5x6 framework matrix, 24 of 30 cells have no papers in this corpus of 12 included records; the populated cells concentrate in aspect 'rights_and_remedies' (5 papers) and evidence type 'legal_doctrinal' (6 papers).
12 papers coded24 empty cells (80%)17 excludedLaw & regulationAI governance - Evidence reviewEvidence Gap MapCR-REV-004
Evidence gap map: data and algorithmic transparency in the AI-governance literature
Across a 5x6 framework matrix, 18 of 30 cells have no papers in this corpus of 22 included records; the populated cells concentrate in aspect 'accountability_governance_mechanism' (7 papers) and evidence type 'conceptual_normative' (9 papers).
22 papers coded18 empty cells (60%)12 excludedAI governanceKnowledge production
Paper critiques
JSON index ↗Structured, per-paper post-publication critiques: claim inventory, evidence map, severity capped by access basis, and a full transparency record. Severity scale — Low · Moderate · High · Severe · Critical; confidence — Low · Medium · High · Very high.
- Severity: HighConfidence: MediumTier AIllustrative
Critique of “Generative AI Adoption and Organizational Productivity: Evidence from 500 Firms”
A. Researcher, B. Co-author · Journal of Strategic Management and Technology · 2026
The paper should be read as a useful but overstated contribution. Its descriptive findings may help map AI adoption, but its causal, policy and AGI claims require substantial weakening. Severity is High: the central causal and policy claims need weakening, but the descriptive core survives. Confidence is Medium: the assessment is well grounded in the design, while some judgements about magnitude depend on materials not fully available.
Innovation, productivity & competitionLabour markets