Answers are grounded in this paper and cite their sources.
Generative artificial intelligence enables the rapid production of fluent, plausible, and readily shareable content, but it also increases the likelihood that problematic information will circulate before it can be adequately verified. In such environments, correction is consequential not only because it may challenge questionable content, but also because it may signal how responsibility for problematic information is being governed. Building on this premise, this research examines how visible correction shapes user responses to generative AI hallucinations by comparing platform-issued correction and community-based correction. We argue that these two corrective actors are analytically important not because they are mutually exclusive in practice, but because, within a focal visible correction episode, they are more likely to direct users’ responsibility judgments toward different actors. Across three experiments in an investment-information context, we examine whether, why, and when visible corrective actors matter. Study 1 establishes baseline differences among no correction, platform-issued correction, and community-based correction. Study 2 examines governance responsibility interpretations while accounting for a negative-affect alternative account. Study 3 tests correction strength as a boundary condition. The findings show that visible correction reduces willingness to further disseminate problematic content, whereas platform-issued correction more consistently strengthens trust in platform governance. By reconceptualizing visible correction as a governance-relevant signal, this research extends correction research in generative AI contexts and distinguishes between platform-focused governance evaluation and downstream diffusion response.
No takes yet. Share an insight, caveat, or question.
Huang et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: