Date: September 19, 2019. 12:00
Location: CCU Seminar Room
Title: Neural basis of decisions guided by sensory confidence and reward value
Affiliation: University College London (UCL)
Deciding between stimuli requires combining one’s sensory confidence with the learned value of their rewards. Armin Lak in our laboratory trained mice in a visual task that probes this combination. He found that mouse decisions conform to a simple model where choices are made based on predicted value, the product of sensory confidence and learned value. The predicted value of the chosen option is encoded so that after outcome it can determine reward prediction error to update learned values. Armin found precise correlates of this predicted value in medial prefrontal cortical neurons. These neurons played a causal role: inactivating them before trial outcome strengthened learning from the outcome. Midbrain dopamine neurons also played a causal role, but only after trial outcome, when they encoded reward prediction errors graded by confidence and influenced subsequent choices. Armin’s results precisely map the hidden variables inferred by a behavioral model onto causal neural signals. They reveal neural signals that combine learned value with sensory confidence and guide learning from the outcome.