Item

Abstract

The neuromodulator dopamine is deeply implicated in many aspects of affective decision-making. In particular, take the popular distinction between model-based control, which supports presbyopic planning based on simulations or samples of the potential future, and model-free control, in which previously successful actions are more myopically recapitulated. Dopamine reports a prediction error important for the latter; however, boosting levels of dopamine pharmacologically in humans and rodents has been shown to boost the impact of the former controller, perhaps by enhancing its operation. In general, when controllers disagree, a form of metacontrol is necessary to determine which dominates – for instance, when the decision-maker has little confidence in their decisions, the model-based controller, which is typically statistically preferable, should optimally be favoured. Here, we consider the influence of dopamine in this form of metacontrol. We used the drug L-DOPA to increase dopamine levels in healthy human volunteers as they performed a task that offers an exquisitely fine decomposition of model-based and model-free choice, and also provides rich information about confidence judgements. Unexpectedly, we found that while boosting dopamine neither increased the impact of model-control on choice nor affected the basic structure of our subjects’ confidence reporting, it did lessen the extent to which model-based control was favoured following low confidence decisions. We are currently examining whether this arises from a difference in the identity of the trials on which the subjects have low confidence. By contrast, suppressing norepinephrine had no effect on choice, confidence or metacontrol. We hope this study will open a new window onto the role of dopaminergic and noradrenergic systems in control over control, and point towards directions by which dysfunctions within and between control systems can be ameliorated.