Dopamine and a Bias for Proximal Action in Cognitive Effort

点击翻译按钮获取中文摘要

Project Summary Tasks with cognitive control demands are treated as subjectively costly. Individuals will avoid higher demands, just like they avoid physical effort. Subjectively exaggerated costs sap motivation for cognitive control, undermining task performance – an effect that has been examined in schizophrenia, depression, Parkinson’s disease, and depression. Yet, despite widespread significance, little is known about mechanisms tracking effort costs or mediating decisions to engage or persist with demanding cognitive tasks. Numerous lines of evidence suggest that dopamine signaling, conveying momentary incentive state, can offset effort costs and thus promote physical and cognitive effort. However, while dopamine has been shown to enhance cognitive control, it also appears to, paradoxically, undermine control by promoting impulsive action. The purpose of this project is to test a hypothesis that can reconcile conflicting effects of dopamine on cognitive control by unifying action selection mechanisms in the cognitive and physical domain. Namely, I will test the hypothesis that dopamine biases benefit over cost information during action selection, but it does so preferentially for “proximal” actions (those that are immediately suggested by the environment). This hypothesis unifies domains in that physical actions are typically suggested by the environment (e.g. levers at hand, or stairs underfoot), while control actions are not. Instead, in the cognitive domain, control actions must compete with what the environment suggests, and will only win out when control mechanisms respond quickly enough. An important corollary of the hypothesis is that very high dopamine levels can amplify even small differences in proximity, thus potentiating proximal “habits” over “controlled” actions and explaining why dopamine can sometimes undermine control rather than promote it. I will test this hypothesis by formalizing the principles in biologically-constrained neural network models, and testing whether they can explain neurophysiological and behavioral dynamics in existing data sets. In a series of experiments, I will measure and manipulate dopamine (with PET imaging and pharmacological interventions) and measure and manipulate proximity (with eye gaze and task design), to determine whether proximity and dopamine determine cognitive action selection as hypothesized. Finally, I test whether the neural network models can predict performance in my experimental data using a common set of parameters. A long-term benefit of this work will be to precisely articulate the mechanisms by which dopamine can affect effortful cognitive action, and generate targets for pharmacological interventions that promote desirable effortful action without also promoting impulsivity.