Reconstructing Cognitive Sovereignty in the Era of Generative AI: A Mixed-Methods Study on User Correction of AI-Generated Misinformation

Abstract

In the AI-driven infodemic, user feedback is a low-cost, scalable approach to counter misinformation. This study investigates the mechanisms behind user correction of AI-generated misinformation and the effects of explanatory interventions. Study 1 (<i>N</i> = 514), based on Self-Determination Theory, integrates the CASA paradigm and FAccT principles (fairness, accountability, transparency) to build a behavioral model, tested via SEM and ANN. Results show perceived autonomy, parasocial relationships, and AI self-efficacy significantly enhance feedback behavior, shaped by fairness, transparency, and empathy. Study 2 (<i>N</i> = 148) uses a controlled experiment to examine explainability, revealing it induces cognitive closure and weakens the positive effect of autonomy on correction. The findings provide theoretical contributions to understanding user agency in human-AI interaction and offer design insights for promoting user participation in GenAI misinformation governance.

Ähnliche Arbeiten