Full Program »
50. Psychotic-like experiences and model-free versus model-based reinforcement learning
Recent work separates reinforcement-learning into two systems: model-based and model-free. The model-free system learns about the value of actions by reflecting on previous reward outcomes, while the model-based system learns by considering both previous outcomes and the likelihood of future outcomes. In individuals with schizophrenia (SZ), studies show that there is a reduced reliance on model-based learning relative to controls. Moreover, recent research indicates that this decreased reliance on model-based decision-making may be modulated by cognitive impairments in SZ such as working memory deficits. However, it is unknown whether this same pattern is shown in individuals who experience psychotic-like experiences (PLEs), but who do not have a clinical diagnosis. We examined reinforcement-learning, working memory, and PLEs in a large university and community sample (n= 183). Results showed that greater working memory capacity was related to greater use of model-based learning (p<0.01). Further, we found a negative relationship between model-based learning and PLEs, with higher PLEs associated with reduced reliance on model-based learning (p=.05). These findings suggest that reduced reliance on model-based learning may exist across the psychosis continuum, which could inform our understanding of the role reinforcement-learning plays in the positive symptoms of psychosis.