ESI Lecture by Armin Lak

Dopaminergic and cortical mechanisms of learning and decision-making under uncertainty

Central to the organization of behavior is the ability to predict the values of outcomes to guide choices. The accuracy of such value predictions is honed by reward prediction error signal that indicates how incorrect a prediction was. In this talk, I will describe a theoretical framework for prediction error computation under perceptual uncertainty, and show that midbrain dopamine neuronal responses comply with signatures of our model, indicating that these neurons can signal prediction errors suitable for guiding perceptual choices. I will then present experiments in which we asked how optogenetic manipulation of dopamine neurons influences animals’ perceptual choices. Finally, I will discuss experiments in which we investigated whether frontal cortical neurons can carry predictive value signals during perceptual decisions. Together, these experiments demonstrate that dopamine cells convey teaching signals that can guide perceptual choices, and illustrate that frontal cortical neurons could contribute to such learning by signalling decision value, based on which prediction errors could be computed.