12,445 papers · continuously updated · last export: 10 Aug 2026livingmeta.ai
← Browse all papers
AI evidence extraction

Survey of AI Hallucinations and Mitigation

Ramie Thompson · Journal of the Association for Information Systems · 2026

AI-generated evidence extraction, verified across multiple analytical personas. Not a substitute for the peer-reviewed original.

9/10
Relevance
1/4
Quality (LMQS)
I
Evidence
0
Citations
0.00
FWCI

Methodology & findings

Study design

Structured literature survey combined with bibliometric analysis of ACM publications from 1995 to 2025.

Sample

> 1000

Primary method

Bibliometric analysis of ACM publications (method details not specified in abstract)

Main result

The study reveals that "the phenomenon commonly termed 'hallucination' remains widely discussed yet inconsistently defined across disciplines" and that "a bibliometric study of ACM publications from 1995 to 2025 reveals a sharp increase in mitigation-focused research alongside the rise of large language models." Additionally, the research demonstrates how "hallucinations function as both risks and, in some contexts, sources of creative value" across domain-specific applications.

Reports effect sizes.

Research paradigm

Interpretive/qualitative synthesis with bibliometric analysis

Author conclusions

The authors conclude that "AI hallucinations [should be] positioned as socio-technical phenomena with direct implications for trust, decision-making, and governance, and provide a foundation for their evaluation, mitigation, and responsible deployment."

Risk of bias

Publication bias in bibliometric study (selection of ACM publications only); Disciplinary bias (inconsistent definitions across domains may not be fully captured); Temporal limitation (bibliometric data extends only to 2025); No systematic quality assessment of reviewed studies

Open questions raised

  • The paper identifies that AI hallucinations remain "widely discussed yet inconsistently defined across disciplines" and calls for structured evaluation and governance frameworks across domain-specific contexts.
  • The paper addresses the inconsistent definition of 'hallucination' across disciplines and provides "a foundation for their evaluation, mitigation, and responsible deployment" as an emerging research direction.
  • The paper identifies that AI hallucinations remain "inconsistently defined across disciplines" and emphasizes the need for clearer understanding of their "implications for trust, decision-making, and governance" across different domains including healthcare, law, finance, art, and information systems.
Data: not_statedCode: not_statedExtracted from: pdfAgreement 75%

Explore related topics

Hallucination ControlResearch IntegrityLiterature Review

Related papers

← Browse all papers