Temporal Dynamics in instrumental Learning: Integrating Entropy Measures and Computational Models

This research explores the dynamics of decision-making within an instrumental learning framework, combining analyses of response times, entropy measures, and computational modeling. We conducted a study using an RLWM task, involving 27 participants and varying set sizes to examine the relationship between task complexity and cognitive functions. Our methodology incorporates Shannon entropy, Cumulative Residual Entropy (CRE), and Instantaneous CRE to assess uncertainty and mental effort across different task conditions. We evaluate two computational models - RLWM and its enhanced version, $\mathbf{~ p i H}$ - to interpret both choice patterns and reaction times. Our findings indicate that set size significantly impacts response times and entropy measures, with increased complexity leading to higher entropy and slower responses. The temporal analysis of iCRE suggests a dynamic shift in cognitive strategies throughout the task. Model comparisons highlight distinct capabilities of RLWM and piH in capturing various aspects of cognitive processes. This comprehensive approach provides novel viewpoints into how the brain balances complex learning mechanisms with simpler rules for efficient decision-making under limited mental resources, contributing to our understanding of adaptive behavior in complex environments.