Inspiration booster or creative fixation? The dual mechanisms of LLMs in shaping individual creativity in tasks of different complexity

Abstract The emergence of large language models (LLMs) presents opportunities for stimulating unlimited creative potential. However, how LLMs influence individual creativity remains unclear. Therefore, this paper examines the dual-opposing mechanisms through which LLMs influence individual creativity. In Experiment 1, each participant collaborated with a human partner or a general, unconstrained LLM partner to complete creative tasks. The results showed that compared to collaborating with the human partner, collaborating with the LLM partner significantly improved individual creativity in simple tasks, attributable to inspiration stimulation. However, in complex tasks, collaborating with the LLM partner led to a decrease in creativity, attributable to creative fixation. To mitigate this impact, in Experiment 2, participants were instructed to collaborate with batch-responsive LLM or constrained-responsive LLM to complete creative tasks. We found that constraining the output of LLMs effectively mitigated the creative fixation they induce in complex tasks, thereby enhancing creative performance. However, this constraint may weaken the positive effects of inspiration stimulation in simple tasks. These findings provide insights for the differentiated application of LLMs in creative tasks.