Conference on Cognitive Computational Neuroscience

Phasic changes in dopamine release encode reward prediction errors (RPE), which are crucial for learning stimulus-value associations. Generalizing learned associations to semantically similar stimuli allows the brain to learn efficiently, but we also need to discriminate the stimuli whose values violate the prediction based on semantic similarity. Previous evidence suggests that dopamine D1 and D2 receptors mediate the generalization and discrimination processes, respectively, while D2 receptor is a common target of antipsychotic medicine. It is unclear how the two processes develop throughout life as our knowledge matures and whether their imbalance is potentially related to psychotic symptoms. To answer these questions , we designed a reinforcement learning task in which values of stimuli follow a hierarchical structure based on semantic categories. Two versions of the task are conducted, one on online participants who also responded to questionnaires assessing psychotic symptoms and the other on age-diverse participants at a museum. Our results indicate that discrimination in the early learning phase is positively correlated with psychotic symptoms, while the recruitment of discrimination over generalization also increases over age , suggesting a potential link between abnormality in the developmental process of reinforcement learning and the risk of schizophrenia.