Abstract
Simulating information flow through populations of Bayesian agents is a powerful technique for understanding some of the environmental and cognitive mechanisms that underlie aspects of information flow and cultural change. In this talk I show that the formation of “echo chambers” (subpopulations of agents who do not trust and do not talk to each other) is very likely given sufficient population heterogeneity, even when the agents are fully Bayesian and constrained not to lie. Even small amounts of an agreed-upon “ground truth” is sufficient to disrupt polarisation, but less so if not all agents have equal access to the “ground truth.” These findings are robust to changes in assumptions about representation, population size, and the goals and cognitive abilities of the agents. This work provides a framework for better understanding how to prevent polarisation, and what aspects of today’s information environment drive it.
Bio
Andrew Perfors (formerly Amy) is an Associate Professor in the School of Psychological Sciences and Deputy Director of the Complex Human Data Hub at the University of Melbourne. Andy’s research involves combining computational models and controlled experiments to better understand higher-order cognition. His research topics include how people learn and form concepts and categories; how these concepts (and the language used to express them) changes in response to changing linguistic and social environments; how people make decisions within these environments; and how these decisions shape those environments themselves. His research focus is on the what and the why within these topics. What goals are human learners and reasoners trying to achieve in particular situations? What constraints (cognitive, informational, environmental) do they operate under? How do these factors shape their behaviour? What are the system-level emergent effects of multiple learners interacting in a complex and dynamic environment? You can check out more at his website or by following him on twitter.