12,637 papers · updated 10 Oct 2026livingmeta.ai
← Browse all papers
Research theme

Epistemic Risk

The Epistemic Risk theme comprises 550 papers in this corpus published between 2004 and 2026. Work here is dominated by Conceptual, Empirical Study, Experimental. 5 open research gaps have been surfaced in this area.

Methodology profile

  • Conceptual180 (33%)
  • Empirical Study70 (13%)
  • Experimental58 (11%)
  • Literature Review52 (9%)
  • Position Paper49 (9%)
  • Content Analysis38 (7%)

Research domains

  • Bias & Epistemic Risk288 (52%)
  • Research Integrity202 (37%)
  • AI Governance19 (3%)
  • Human-AI Collaboration17 (3%)
  • Knowledge Synthesis8 (1%)
  • Data Analysis5 (1%)

Frequent sub-topics

LLM personality, language ideology propagation, and societal bias · 1AI hallucination and epistemic authority in information systems · 1judgment-embedded AI outputs and professional decision-making · 1AI-generated image detection and authenticity judgment · 1Indoctrination risk in using large language models for research · 1AI-generated deception detection and cognitive offloading · 1toxic algospeak and content moderation with AI · 1credibility assessment in synthetic media · 1

Open research gaps

  • The authors identify a gap in understanding how to adequately address labeler bias in crowdsourced annotation despite diversity efforts, suggesting future research should focus on practical solutions
  • The authors identify a need to understand why Reverse Turing Adaptation has not materialized despite theoretical conditions existing for cognitive flattening, and suggest ongoing monitoring of linguis
  • Need for direct assessment of reasoning process in LLM-based agents rather than outcome-only evaluation
  • Requirement for changes at the base-model training level to address epistemic reasoning patterns (scaffold engineering insufficient)
  • Necessity for reasoning itself to become an explicit training target in LLM development

Representative papers

← Browse all papersPriority research agenda →