Unveiling the complexity of learning and decision-making

Reinforcement learning (RL) is crucial for learning to adapt to new environments. In RL, the prediction error is an important component that compares the expected and actual rewards. Dopamine plays a critical role in encoding these prediction errors. In my thesis, I investigate how human behavior in RL can be explained by RL models that incorporate factors such as novelty and abnormalities in the dopamine system. To broaden our understanding of RL and decision-making, I extend the study to include neural correlates of social dominance, examining their impact on decision-making processes. First, our model suggests that humans tend to adopt novelty-seeking strategies in RL tasks. Indeed, behaviorally, participants are often distracted by the emergence of novel states, especially among those who are more optimistic about potential rewards. Given the critical role of dopamine in RL, we examined its impact on two populations typically associated with dopamine dysregulation: patients with schizophrenia and the elderly. Our model suggests that restricted dopamine levels lead to deficits in reward-based learning due to poor encoding of the reward prediction error. On the other hand, the model predicts that punishment-based learning abilities should be preserved or even enhanced compared to controls. We could verify this prediction. Older adults demonstrate robust RL capabilities, with no significant differences compared to young adults, challenging the common belief about cognitive decline in healthy ageing. Finally, we investigate the neural correlates of social dominance in female participants. Our findings reveal an EEG component in dominant females similar to the previously reported one in dominant males, suggesting a potential universal neuromarker for social dominance during decision-making scenarios.