Can Visual Fixations Explain Context-Dependent Reinforcement Learning?

Authors: Rishi Heggawadi, William Hayes, Melanie Touchard, Nadiah Layne, Andrew Dolinsky

Field of Study: Social Sciences

Program Affiliation: DReaM Lab

Faculty Mentors: William Hayes

Easel: 39

Timeslot: Morning

Abstract: Context-dependent reinforcement learning (RL) suggests decision making is influenced by relative rather than absolute option values. This study aimed to test a model-based account of context-dependent RL based on selective visual attention to choice feedback. Fifty participants completed a learning task, repeatedly choosing between symbol pairs associated with variable rewards, followed by a transfer test involving novel pairs without feedback. Participants' goal was to accumulate as much reward as possible. Using eye-tracking methods, this research recorded which outcomes participants fixated on throughout the learning task. The results reveal that decisions are made based on both relative and absolute values, with visual fixation patterns alone proving insufficient in determining learned values. This study highlights the importance of integrating computational modeling and eye-tracking data to better understand the processes driving reinforcement learning and decision-making.