Epistemic drive and memory manipulations in explore-exploit problems

AbstractPeople often navigate new environments, and must learn about how actions map to outcomes to achieve their goals. In this paper, we are concerned with how people direct their search, and trade off between selecting informative actions and actions that will be most immediately rewarding, when they are faced with new tasks. We find that some people were able to learn new reward structures efficiently, selected globally informative actions, and could transfer knowledge across similar tasks. However, a consistent proportion of participants behaved sub- optimally, caring more about collecting new information instead of maximizing reward. We examine how memory constraints and prior knowledge affect this drive to explore, and study the exploratory strategies of people across four experiments.

Return to previous page