Modelling emergent social phenomena
How do people's beliefs and actions change when they are embedded in social structures that can shape the information they receive, and which change when acted upon? In this work, I use a variety of agent-based models, such as multi-agent reinforcement learning (MARL) and generative agent-based models (GABMs), to capture how individual behaviour can produce feedback loops that shape the structure of a social group, and subsequently the beliefs of those within it, producing emergent phenomena such as stereotyping and social convention formation.
Relevant Papers
Gelpí, R. A., Tang, Y., Jackson, E. C., & Cunningham, W. A. (2025).
Stereotypic expectations entrench unequal conventions across generations in deep multi-agent reinforcement
learning. PNAS Nexus, 4(3), pgaf076.
Tang, Y.*, Gelpí, R. A.*, & Cunningham, W. A. (2023). Unequal norms
emerge under coordination uncertainty in multi-agent deep reinforcement learning. In Proceedings of
the 45th Annual Meeting of the Cognitive Science Society (pp. 555–561).