Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster C in Poster Session C: Wednesday, August 5, 9:30 – 11:15 am, Kimmel Center, Shorin & Rosenthal Rooms

Assessing Value Normalization and Habit Explanations in the Reward Pairs Task: a Simulation Study

Viktor Timokhov¹, Hugo Fluhr¹, Philippe N. Tobler¹, Stephan Nebe¹; ¹University of Zurich

Presenter: Viktor Timokhov

Choice frequency effects in the Reward Pairs Task were previously attributed to habit learning, but value normalization provides an alternative explanation. We simulated five reinforcement learning (RL) models, with and without habit learning and value normalization, using a Diffusion Decision Model (DDM) to predict choices and response times (RTs). Models with value normalization failed to capture observed choice behavior or the inverted U-shaped RT-value relationship. Only combined RL and habit learning model, without value normalization, matched choice and RT data in both equal- and unequal-reward trials. Our findings refute the value normalization explanation and support the habit learning explanation for behavior in the Reward Pairs Task.

Topic Area: Decision-Making, Cognitive Control & Event Cognition