Competition in Creative Tasks, Gender Gaps and the Mitigating Role of AI: Experimental Evidence

We examine gender differences in competitiveness when performance depends on creativity, which is central to many high-skilled jobs, and investigate how generative AI may affect them. Our study is based on a laboratory experiment in which participants complete a creative task and then choose whether to enter a tournament compensation scheme, with access to ChatGPT varying across treatments. While men and women achieve similar levels of creative performance, they differ in competition entry, with women displaying a lower willingness to compete in environments without AI assistance. This gender gap in competition becomes smaller when AI is available. To investigate the mechanisms behind this pattern, we construct a measure of underconfidence based on the difference between participants’ beliefs about their relative performance and their actual rank. We show that underconfidence predicts competition entry and that, besides improving creative performance, AI is associated with lower underconfidence, particularly among women.