Understanding neural coding through the model-based analysis of decision making

Abstract

The study of decision making poses new methodological challenges for systems neuroscience. Whereas our traditional approach linked neural activity to external variables that the experimenter directly observed and manipulated, many of the key elements that contribute to decisions are internal to the decider. Variables such as subjective value or subjective probability may be influenced by experimental conditions and manipulations but can neither be directly measured nor precisely controlled. Pioneering work on the neural basis of decision circumvented this difficulty by studying behavior in static conditions, in which knowledge of the average state of these quantities was sufficient. More recently, a new wave of studies has confronted the conundrum of internal decision variables more directly by leveraging quantitative behavioral models. When these behavioral models are successful in predicting a subject's choice, the model's internal variables may serve as proxies for the unobservable decision variables that actually drive behavior. This new methodology has allowed researchers to localize neural subsystems that encode hidden decision variables related to free choice and to study these variables under dynamic conditions.

Figure 1.

A, Left, The classical paradigm directly correlates neural data with stimuli or behavior. Center, The “model-in-the-middle” approach instead uses a mechanistic model as an intermediary, on the one hand constraining the model to describe behavior and on the other correlating internal variables of the model with neural data. Right, Two examples of triplets of behavior, model, and neural data, one from Sugrue et al. (2004) and another from Daw et al. (2006). B, The basic scheme for using models as intermediaries. Top, Mechanistic models of choice generally take as input the history of past choices and rewards and render a prediction of future choice as an output. Center, Internal variables computed by the model can be extracted and used as proxies for subjective decision variables hidden from the experimenter. Bottom, These proxy variables can be used to identify specific neural representations of the decision variable of interest within the brain areas supporting decision making.