Negotiating Creativity with AI: Effects on Self-Efficacy and Task-Performance
Generative AI is increasingly integrated into creative work, yet its impact depends not only on its capabilities but on how it participates in the creative process. This study examines how collaborating with an AI agent framed as a conversational teammate shapes creative self-efficacy (CSE) and task-performance in an ideation and ranking task (N=28). Results show that AI collaboration increased CSE and led to greater revision of initial ideas ranking, indicating deeper engagement in evaluation without promoting blind reliance on AI suggestions. Qualitative findings reveal that these effects were driven by negotiation. When the AI challenged and defended its evaluations, participants engaged in reflection, justification, and reconsideration. This interaction introduced productive friction, supporting deeper thinking while preserving human creative control. We argue that creativity in human–AI collaboration is shaped by social and interactive dynamics, highlighting the importance of designing AI systems that foster engagement through dialogue rather than passive assistance.