Adaptive decision-making requires the brain to flexibly compute subjective values for choice options across changing environmental contexts and distinct operational domains, such as visual stimuli versus motor actions. This dissertation elucidates the neural architectures supporting adaptive valuation across multiple scales of analysis, leveraging functional magnetic resonance imaging (fMRI) and human single-unit electrophysiology. First, investigating the valuation of multi-attribute stimuli under shifting goals reveals a hierarchical valuation process. While visual cortices represent static, context-independent stimulus attributes, regions within the prefrontal cortex (i.e., ventromedial prefrontal cortex (vmPFC) and orbitofrontal cortex (OFC)) transform these into context-sensitive ``attributes in value space''. These intermediate representations are then integrated into a unified subjective value signal along the dorsomedial prefrontal cortex (dmPFC). Second, single-neuron recordings during structurally parallel action- and stimulus-based tasks uncover a temporally shifting representational architecture. Prior to choice, neurons across the vmPFC, anterior cingulate cortex (ACC), and pre-supplementary motor area (preSMA) broadly track available pre-decision values as to facilitate comparison across both stimuli and actions. However when probing choice-dependent valuation, the network segregates: the ACC specializes in action-based chosen value and the vmPFC tracks stimulus-based chosen value. In contrast to the specialization the preSMA encodes value and chosen identity across both choice domains. Together, these findings demonstrate that human valuation relies on a dynamic sequence of transformations. This prefrontal network balances the abstraction required to compare disparate options with the specificity needed for accurate goal-directed behavior.
Aniek Fransen (Wed,) studied this question.