Factors influencing disinformation avoidance behavior among generative artificial intelligence users: A cognition-affect-conation framework

This study aimed to uncover the mechanisms driving disinformation avoidance behavior among generative artificial intelligence users to reduce negative impacts and support sustainable use. Using the cognition-affect-conation framework and heuristic-systematic model, this study analyzed 769 valid responses through covariance-based structural equation modeling and fuzzy-set qualitative comparative analysis. The results showed that perceived risk and information hallucination heightened artificial intelligence disinformation anxiety, while perceived mind and performance efficacy enhanced affective commitment. Artificial intelligence disinformation anxiety facilitated disinformation avoidance behavior, whereas affective commitment reduced it. Systematic cues, such as information hallucination and performance efficacy, proved more influential than heuristic cues, such as perceived risk and perceived mind. Psychological resilience directly affected disinformation avoidance behavior and also moderated the relationships between artificial intelligence disinformation anxiety, affective commitment, and disinformation avoidance behavior. Four pathways leading to disinformation avoidance behavior were identified, highlighting substitutability effects among perceived risk, information hallucination, and perceived mind. These substitutability effects were observed under specific configurational conditions identified through fuzzy-set qualitative comparative analysis. They should be interpreted within the scope of this study rather than as universally generalizable. This study advances the literature on generative artificial intelligence user behavior by shifting the focus from proactive to defensive perspectives. It embeds the heuristic-systematic model into the cognition-affect-conation framework to capture the interplay between cognitive and affective processes. Practical recommendations are provided for generative artificial intelligence developers to reduce avoidance behaviors, rebuild user trust, and promote the sustainable development of generative artificial intelligence.