关键词: 冲突性虚假信息, AI幻觉, 验证行为, 生成式人工智能, 启发式—系统式模型

Abstract: [Purpose/Significance] This study aims to explore the dynamic evolutionary mechanisms of users' verification behavior in the context of GAI-generated conflicting misinformation. By revealing the evolving processes across cognitive, emotional, and behavioral dimensions, it seeks to provide a human-centered solution for the governance of AI hallucinations. [Design/Methodology] Qualitative data were collected from 34 GAI users through semi-structured in-depth interviews combined with the Critical Incident Technique. Utilizing the systematic coding procedures of Grounded Theory, the qualitative materials were analyzed and coded, extracting 4 core categories and 15 main categories to construct a user-validated dynamic behavior evolutionary model. [Findings/Conclusion] In the context involving GAI-generated conflicting misinformation, user verification behavior primarily evolves along three distinct pathways: the cognitive path （contextual trigger–cognitive appraisal–verification behavior）, the emotional path（contextual trigger–emotional response–verification behavior）, and the cognitive–emotional path（contextual trigger–cognitive appraisal–emotional response–verification behavior）. Driven jointly by cognition and emotion, users opt for either heuristic or systematic verification, with a dynamic transformation relationship existing between these two verification behavioral modes. Furthermore, the verification outcome serves as feedback that reshapes the initial context, thereby giving rise to a dynamic cycle. [Originality/Value] Moving beyond the static perspective inherent in existing information verification studies, this research constructs a dynamic evolutionary model that integrates cognition, emotion, and behavior. It elucidates the multi-pathway driving mechanisms and the feedback reshaping process underlying user verification behavior in GAI context, and extends the applicability boundaries of HSM theory in human–computer interaction scenarios.

Keywords: Conflicting misinformation, AI hallucinations, Verification behavior, Generated AI（GAI）, Heuristic-Systematic Model