- Main
Social Learning from Incomplete Information in a Dynamic Decision-Making Task
Abstract
The exploration-exploitation dilemma in dynamic decision-making scenarios is a notoriously hard problem to solve. Having a partner to potentially learn from might make it easier to balance exploration and exploitation. In the current study, we investigate the impact of social information (i.e., about others’ exploration behavior vs. their rewards) and partner performance (optimal vs. random) on participants’ behavior in a dynamic decision-making task that contains a learning trap. We find that observing the exploration behavior of an optimally choosing partner was detrimental to participants’ overall performance and reduced participants’ exploratory tendencies. In contrast, observing a random partner’s exploration behavior stimulated participants’ exploration, though this increase in exploration did not help participants to uncover the reward function. Following previous literature, a reinforcement learning model that contained eligibility traces was able to describe human behavior and helped to uncover potential mechanisms that could explain aspects of the findings.
Main Content
Enter the password to open this PDF file:
-
-
-
-
-
-
-
-
-
-
-
-
-
-