The Confidence Cage: How Creative Self-Efficacy Hinders Gen AI-Augmented Ideation

Drawing on Associative Theory and Threat-Rigidity Theory, we examine why generative AI's creative benefits are unevenly distributed. We argue that while AI can enhance creative performance by expanding associative ideation, this effect is weaker for individuals with high creative self-efficacy (CSE). AI can threaten the identity of people with high CSE (strong belief in their creative abilities), making them less able to integrate new ideas from AI and more cognitively rigid. In contrast, people with low CSE tend to feel less threatened, maintain their cognitive flexibility, and gain more from AI input. Our moderated mediation model is supported by the results of a pre- and post-test experiment (N=300), which demonstrates that AI support enhances creativity through improved ideation. For those with low CSE, this effect was more pronounced. These results reveal how a psychological strength can become a liability and suggest strategies to foster more effective human-AI collaborations.