Abstract
Previous work shows that understanding the processes through which groups function requires understanding human behavior at a team level. However, there exist persistent methodological challenges when treating groups, compared to individuals, as the unit of analysis, wherein resources (e.g., sample size, time) needed to draw reliable conclusions grow exponentially with design complexity. In this chapter, we explore simulations using generative artificial intelligence (GAI) as a potential avenue for addressing such challenges. Integrating work on social science research and GAI, we identify criteria that GAI simulations should satisfy. We offer two specific examples of GAI simulated group interactions to highlight their existing limitations.