Exploration and Exploitation Reflect System-Switching in Learning

AbstractMounting evidence suggests that human category learning is achieved by multiple qualitatively distinct biological and psychological systems. In an information-integration (II) categorization task, optimal performance requires switching away from rule and adopting a procedural response strategy. However, many participants perseverate with rules. This article attempts at understanding the difference between optimal and suboptimal participants in II categorization. To this end, we collected data in the Iowa Gambling Task (IGT) and an II categorization task. Performance in the IGT was used to estimate each participant’s sensitivity to reward, punishment, and propensity to explore. The results show that optimal participants in the II task explored more in the IGT than suboptimal participants. However, optimal participants in the II task did not show higher sensitivity to punishment or lower sensitivity to reward. We conclude by discussing the implications of these findings on system-switching and theoretical work on multiple-systems model of perceptual category learning.

Return to previous page