[論文レビュー] The Verification Crisis: Expert Perceptions of GenAI Disinformation and the Case for Reproducible Provenance

Alexander Loth, Martín Munizaga Kappes+1 more|arXiv (Cornell University)|Feb 2, 2026

Scientific Computing and Data Management被引用数 0

ひとこと要約

A Wave 1 study of 21 experts across AI, policy, and disinformation fields finds multimodal GenAI disinformation is a present, escalating threat, with text posing systemic epistemic risks and deepfake video triggering immediate shock; experts advocate reproducible provenance and infrastructure over purely technical detection.

ABSTRACT

The growth of Generative Artificial Intelligence (GenAI) has shifted disinformation production from manual fabrication to automated, large-scale manipulation. This article presents findings from the first wave of a longitudinal expert perception survey (N=21) involving AI researchers, policymakers, and disinformation specialists. It examines the perceived severity of multimodal threats -- text, image, audio, and video -- and evaluates current mitigation strategies. Results indicate that while deepfake video presents immediate "shock" value, large-scale text generation poses a systemic risk of "epistemic fragmentation" and "synthetic consensus," particularly in the political domain. The survey reveals skepticism about technical detection tools, with experts favoring provenance standards and regulatory frameworks despite implementation barriers. GenAI disinformation research requires reproducible methods. The current challenge is measurement: without standardized benchmarks and reproducibility checklists, tracking or countering synthetic media remains difficult. We propose treating information integrity as an infrastructure with rigor in data provenance and methodological reproducibility.

研究の動機と目的

• Assess perceived severity of GenAI disinformation across text, image, audio, and video modalities.
• Evaluate current mitigation strategies and their effectiveness.
• Argue for reproducible provenance and infrastructural approaches to information integrity.
• Propose a practical framework combining checklists, methods hubs, and knowledge graphs to advance reproducible research.
• Lay out a three-act narrative and actionable path toward an infrastructure for truth verification.

提案手法

• Purposive snowball sampling to recruit 21 experts from AI research, policy, and disinformation fields.
• Two-latent-construct survey measuring Threat Severity and Mitigation Efficacy on Likert scales (1–7).
• Qualitative analysis via reflexive thematic analysis to identify themes like Synthetic Consensus and Black Box Failure.
• Quantitative descriptive statistics to identify central tendencies and consensus; qualitative data coded inductively and triangulated with quantitative results.
• Reproducibility statement aligning with R2CASS: data, survey instrument, and analysis code are to be shared via the GESIS Methods Hub with Momeni et al. checklists.

実験結果

リサーチクエスチョン

• RQ1 What is the perceived threat level of GenAI disinformation across text, image, audio, and video modalities?
• RQ2 Which mitigation strategies do experts view as most and least effective, and why?
• RQ3 Can a reproducible provenance framework address the verification crisis more effectively than detection alone?
• RQ4 How can infrastructure like knowledge graphs and Methods Hub uplift reproducibility in disinformation research?
• RQ5 What structural changes are needed to shift from detection to provenance-centered defenses?

主な発見

• Text-based GenAI content is perceived as a systemic threat with high potential for synthetic consensus and epistemic fragmentation.
• Deepfake video shows high shock value but is considered more debunkable than text as a long-term threat.
• Technical detection tools are viewed as the least effective long-term solution due to black-box limitations and adversarial evolution.
• Experts favor provenance standards and regulatory frameworks, despite implementation barriers.
• A reproducible infrastructure (checklists, methods hub, knowledge graphs) is proposed as essential to countering disinformation at scale.