All FSI Projects

Exploration-Exploitation and Age


Samuel McClure
Laura Carstensen
Project Mentor

Decisions are made for two general purposes. A choice is determined either to obtain a desired outcome or to gather new information about other (perhaps more desirable) novel actions. These two motivations are generally referred to as exploitation and exploration, respectively. When confronted with a novel environment, the strategy employed to balance these two demands has tremendous consequence on performance. This study aims to understand how exploration-exploitation strategies change across the life span. While it is recognized that balancing exploration with exploitation of accumulated knowledge is critical for performance, little work has studied how people solve this problem at any age. The investigators will study behavior in bandit tasks in which participants are given a limited number of choices from a variable number of actions with fixed, but initially unknown, payoff characteristics. This is a well-studied environment in statistics. Previous aging research suggests that there may be significant differences in exploration-exploitation strategy across the life span. Socioemotional selectivity theory (SST; Carstensen, 2006) states, in part, that aging is associated with a reduction in the variability of behaviors with greater preference shown for behaviors that produce positive emotions. This suggests that older adults may place relatively greater emphasis on exploitation relative to younger participants. These experiments aim to test this hypothesis under different experimental conditions where greater exploration is alternatively beneficial and detrimental. The investigators aim to demonstrate an interaction such that younger participants perform better in some conditions and older participants perform better in others. The final purpose of this project is to collect pilot neuroimaging data to identify the neural mechanisms that support exploitative and exploratory behavior. These experiments will be performed in parallel with the behavioral tests and will serve as pilot data for subsequent larger-scale research proposals.


Nomita Divi