Augmenting Human Creativity in Brainstorming Sessions with Artificial Intelligence

Abstract:

This paper investigates how artificial intelligence, in particular transformer-based language models like GPT-3, provide a highly scalable solution for augmenting human creativity and overcoming the productivity loss in interactive brainstorming groups. We conduct a one-factorial between-subjects lab experiment, in which participants brainstorm ideas for a common question. Additionally, we complement results with qualitative insights. We compare nominal groups (humans brainstorming individually), interactive groups (humans brainstorming together as a team), hybrid groups (a human brainstorming together with an AI), and AI-only control groups. Our results indicate that hybrid groups outperform both interactive and nominal groups in terms of the number of generated ideas and perform on par in terms of brainstorming creativity while requiring only half the human resources. Furthermore, our study suggests that AI could even replace human ideation altogether. We discuss why such an interpretation of our findings might be too hasty and shortsighted, consider important implications, and derive avenues for future research.

The seminar will be delivered in hybrid format.

Please sign up for the format of  your participation here