Date: Friday, October 21, 2022
Time: 12:00pm – 1:00pm
Location:SCSB Conference room 46-6011 + Zoom Meeting (https://mit.zoom.us/j/99631032115)

Speaker: Seng Bum Michael Yoo, Ph.D.
Affiliation: Past Simons Postdoctoral Fellow, Mehrdad Jazayeri
Laboratory. Current position: Assistant Professor, Sungkyunkwan University

Host: Dr. Mehrdad Jazayeri

Talk title: Using network models to probe reduced sensitivity to others’ information in observation learning

Abstract: Observational learning is the ability to learn about hidden states of the environment through observing others’ actions and outcomes. Although behavioral and neural signatures of observational learning have been reported, a mechanistic understanding of its underlying neural computations is understudied. Here, we designed a two-player probability reversal-learning task in which both actor and observer need to track the covert change of the reward probability of the options. We find that, even after matching all the information available to each agent, both human and nonhuman primates still learn less from others’ actions and outcomes than their own. We devised a performance-optimized network model to address what types of constraints would bring such suboptimal behavior. Once the model includes independent noise for each agent’s outcome, the model accounted for others’ actions and outcomes to a lesser degree. In sum, the result suggests that asymmetric weights for the action and outcome information would underlie reduced efficacy of observation learning.