Learning while controlling

Playing an unfamiliar instrument means trying to produce the right sound while discovering how the instrument responds. An action can therefore serve two purposes: achieving a goal and learning how to achieve it better. Changing the timing or force of a movement may alter the sound we make and reveal something about the instrument. We ask how people balance immediate success with exploration that improves future performance. Which variations are worth trying, and how does that choice change as an instrument becomes familiar or its response becomes less predictable?

This direction connects dual control—choosing actions partly for what they reveal about how to control a system in the future (Feldbaum, 1960)—with reinforcement learning’s exploration–exploitation problem: when to use a familiar strategy and when to try something new (Sutton & Barto, 2018). These frameworks guide our questions about how uncertainty, feedback, and the opportunity to learn influence action. Our interest is in how people improve their understanding of a changing system while continuing to pursue a goal.