Last week, during commencement, Cameron Berg was awarded the first Robert J. Glushko Prize for Distinguished Undergraduate Research in Cognitive Science. Cameron’s research on the role of social cues in human reinforcement learning stood out among the nominated essays.
In his study, participants played a game where they chose to click on one out of two possible buttons. After clicking the selected button, participants were awarded either 1 coin or 0 coins. At the beginning of each participant’s task, one of the buttons was programmed to award a coin 70% of the time, becoming the preferrable option. At some moment during the middle of the task, the award scheme would covertly change so that the preferrable button changed. Crucially, participants were not playing this game in isolation. They were playing with what they thought was a group of conspecifics which varied in size across the task, from 2 other players to 8 other players. Thus, observing other players choices was an intelligent way to figure out which button was the best option.
Cameron’s results show that people use different strategies to integrate the decisions of others into their own decision-making process depending on the number of conspecifics at play. While the models that best described participants’ behavior in the trials with two conspecifics relied on richer representations that involved tracking and maintaining multiple statistics, the models that showed the best fit in trials with four or eight conspecifics relied on substantially simpler measures.
Cameron’s thesis offered an important contribution as to the role of social cues in human learning.