Foundations for Improved Models of Value-Based Decision-Making: Neural Correlates in the Anterior Cingulate Cortex and their Implications

Abstract

How do we choose between different options based on their value? Conceptual frameworks of value-based choice describe the process in several stages, often (though not always) with the assumption that these stages are serial, discrete, and localized to different brain regions. Moreover, while several models of value-based decision-making exist at the cognitive level, models of the neural circuit underlying this process have mostly been limited to modeling value comparison. We sought to gain deeper insight into the neural mechanisms underlying value-based choice by examining the responses of neurons in the dorsal and subgenual anterior cingulate cortices (ACC) of rhesus macaques. Our findings in the dorsal anterior cingulate (dACC) are consistent with its direct involvement in the decision-making process. Moreover, neural responses to value in this region suggest an attentionally-aligned—rather than option-aligned—mechanism of choice; i.e. neurons respond to the value of the stimulus presented to the subject, rather than responding selectively to different options according to their identity. Our results also suggest that options are evaluated and preliminarily accepted or rejected as they are presented, rather than a comparison and choice process that only begins once evaluation of all options is complete. Finally, we find evidence that the attributes constituting the value of an option are not fully integrated into a unified value for that option. Many of these neural correlates were similarly found in the subgenual anterior cingulate cortex (sgACC); an adjacent prefrontal region whose role in decision-making is less understood. These findings recapitulate many of the insights gained from cognitive models of decision-making, suggesting that stages of the choice process proceed in parallel, utilize many of the same neural circuitry, and may be replicated across multiple brain regions. These results also illuminate the contours of a different implementation of value-based decision-making in neural circuits.