Explorations in human vs. generative AI creative performances: A study on human-AI creative potential

Advances in quality and range of generative AI have opened up new possibilities for AI-supported work and human-AI collaboration. Now, researchers are challenged to investigate how, where, and to whom AI can contribute meaningfully. In this paper, we present a study on human versus AI creative performance in the Alternate Uses Test (AUT) and discuss the implications of our results for human-AI collaboration. We analyze how different text-generative AI chatbots compare to human dyads in the AUT regarding creative fluency, originality, flexibility, and elaboration. Our results reveal high ranges in performance within both the human dyad group and the AI chatbot group. Further, humans excel in original and flexible ideation, while AI better elaborates and details responses. Therefore, collaborative creative performance in human-AI teams could benefit from these different but complementary skills. In future work, we will test this assumption and explore the social dynamics of human-AI collaboration to find ways of trustworthy and reliable human-AI collaboration.