Author ORCID Identifier
https://orcid.org/0009-0003-3534-206X
Date of Award
Summer 2026
Document Type
Thesis (Ph.D.)
Department or Program
Psychological & Brain Sciences
First Advisor
Alireza Soltani
Abstract
Learning in naturalistic settings involves inherent uncertainty about which aspects of a decision predict subsequent outcomes. Choice options contain multiple features, while outcomes are typically obtained through actions directed toward those options. Yet relatively little is known about how the brain determines whether to associate outcomes with stimuli representing choice options or with the actions used to select them, and how these associations are combined to guide decisions. Here, I combine multiple computational methods and analyses to quantify concurrent learning in monkeys performing tasks with different levels of uncertainty about the model of the environment. Using reinforcement learning (RL) framework, I find evidence for dynamic, competitive interactions between stimulus- and action-based learning, as well as single-cell and population-level representations of the arbitration weight, representing the current belief about more reliable model of the environment. This arbitration signal in turn systematically modulates both the strength and geometry of value representations. Moreover, by estimating multiple neural timescales at a single-cell level, I find that the memory timescales of relevant choice attributes are flexibly adjusted, with direct functional implications on the choice behavior. Lastly, by comparing behavior in control monkeys and monkeys with bilateral lesions to the amygdala or ventral striatum, I further validate the computational framework of arbitration and identify dissociable contributions of these subcortical structures. I show that the amygdala adjusts the initial balance between the two learning systems, thereby altering the time course of choice behavior. This previously unrecognized role of the amygdala helps reconcile seemingly contradictory findings and generates testable predictions for future studies. Overall, these findings demonstrate that cortical and subcortical computations operating at multiple spatial and temporal scales support arbitration between stimulus- and action-based learning, enabling adaptive behavior in complex and uncertain environments.
Original Citation
Woo, J.H., Costa, V.D., Taswell, C.A. et al. Contribution of amygdala to dynamic model arbitration under uncertainty. Nat Commun 16, 11704 (2025). https://doi.org/10.1038/s41467-025-66745-1
Recommended Citation
Woo, Jae Hyung, "Neural and Computational Mechanism of Arbitration Between Stimulus and Action Learning" (2026). Dartmouth College Ph.D Dissertations. 532.
https://digitalcommons.dartmouth.edu/dissertations/532
Included in
Cognitive Neuroscience Commons, Computational Neuroscience Commons, Systems Neuroscience Commons
Comments
Part of this work (Aim 3) has been published in the below journal.