A02Pathway-specific roles of midbrain dopamine neurons in value-based decision-making
The ability to learn better prediction from the error of prediction (i.e., the discrepancy between actual and expected outcome) is important for selecting appropriate actions in natural environments. It has been thought that midbrain dopamine neurons are relatively homogeneous population and signal reward prediction error which resembles the teaching signal proposed in reinforcement learning theories. In the meantime, growing evidence supports the diversities of midbrain dopamine neurons. This suggests that dopamine neurons are involved in various learning mechanisms implemented in our brain. In this project, we are testing the possibility that dopamine neurons transmit distinct signals to their downstream targets in a projection pathway-specific and environmental context (or model) dependent manner. Combining large-scale multi single-neuron recordings and optogenetics, we will record the activities of individual dopamine neurons with high temporal resolution. We will apply this technique to freely-moving animals performing value-based decision-making tasks and examine how dopamine systems encode information ranging from relatively simple external inputs to more complex internal models of the environments. By understanding the mechanisms of information processing in the dopamine system, we are hoping to develop more brain-like, generalized artificial intelligence systems that solve a wide range of problems flexibly and appropriately.