Cognitive Science & Social Psychology Seminar ~ Chujun Lin, Faculty, Columbia University

Cognitive Science & Social Psychology Seminar

Chujun Lin ~ Columbia University, Assistant Professor

Title: From Isolated Psychological Phenomena to Unifying Computational Principles of Person Perception

Abstract: People routinely make a wide range of inferences about others from rich information in daily life. Existing theories have provided important insights into different aspects of social inference, but these aspects may be governed by shared computational principles. My research attempts to identify these principles using computational modeling, naturalistic experiments, and cross-cultural designs. In this talk, I will share three insights.

First, decades of research have asked how people mentally represent the many inferences they form about others, with the low-dimensional account dominating. I argue that mental structure cannot be understood independently of the processes operating over it, and show that social inferences are better captured by a small-world network model in which dimensions emerge as byproducts of dynamic processes rather than serving as underlying mechanisms.

Second, a growing literature documents how people make social inferences from different facial, bodily, and contextual cues, but leaves unexplained why perceivers rely on particular cues in particular contexts. I argue that cue use cannot be understood by studying cues in isolation, and show that perceivers learn, and flexibly adapt to, the broader statistical structure of co-occurring cues in their environment, to guide social inferences.

Third, people learn about both individuals and their groups, yet individual inference and group stereotyping have largely been studied separately. I argue that these two forms of learning are inseparable in everyday social experience, and show that a hierarchical Bayesian model that captures this joint learning explains how maximizing the accuracy of individual inferences can nevertheless give rise to persistent group stereotypes.

Together, these findings illustrate how connecting phenomena traditionally studied in isolation can reveal unifying computational principles underlying how people think about others in the complex social world.