The Effects of Human-AI interaction on creativity

ABSTRACT Creativity, the ability to generate ideas that are both novel and useful, has long been considered a hallmark of human cognition. With the rapid emergence of large language models (LLMs), artificial intelligence has become a prominent collaborator in many tasks, as well as creative endeavors. Yet, this raises pressing questions about how such interaction affects human originality and diversity of thought. Previous work points to possible homogenization of Human-AI collaborative outputs. In the present research, we build on such homogenization findings, to investigate whether human-AI interaction affects human creativity. Study 1 introduces HAICCO, a novel experimental platform that embeds turn-taking interaction with an LLM into established creativity tasks, including Forward Associations and Divergent Associations Tasks. We hypothesize that repetitive human-AI collaboration will reduce semantic variability and constrain originality in post-collaboration creativity performance compared to baseline. We find that originality decreased post HAICCO, and collaborative rounds showed a more creative pattern of semantic movement, especially in the Divergent Association task. An exploratory analysis revealed that a less homogenous pattern of semantic change late in human-AI collaboration predicts higher post-collaboration originality. Study 2 introduces CHAP, a platform designed to examine the effect of AI on the creative process, by contrasting human and AI involvement during idea generation versus idea evaluation. Preliminary results show no difference in the pre-and post- collaboration measures. However, as expected, AI collaboration was related only to idea evaluation and not idea generation. Together, these studies provide unique empirical evidence for the impact of repetitive human-AI interaction on the creative product and the creative process.