Computational Mechanisms of Social Reward and Effort Learning Predict Anhedonia and Translate into Real-World Difficulties in forming new Friendships in Young People

<p>Background: Depressive and anhedonia symptoms are thought to arise from altered computational mechanisms of reward learning. However, most studies use monetary rewards; therefore, it is unclear how learning about social rewards specifically, and the effort to obtain them, underpin symptoms or real-life social experiences.Methods: We fitted Q-learning models that separately account for the learning to maximise rewards and to minimise the effort needed to obtain them to data from a probabilistic reward and effort learning task. The task was completed by 122 young people (aged 18–25) with a range of depressive and anhedonia symptoms. Parameter estimates were used to predict depressive and anhedonia symptoms, as well as real-life social experiences.Results: For both reward and effort learning, the besting-fitting model contained learning rate and temperature as free parameters. We found that increased temperature (more random, exploratory choices) during reward and effort learning was related to social anhedonia and consummatory anhedonia, respectively. Further, greater difficulties in forming new friendships in real-life was associated with decreased learning rate (more gradual updating of expectations after unexpected outcomes) during the learning to maximise social rewards.Conclusion: These findings provide novel evidence that computational mechanisms of reward learning can be translated into the social experiences of young people with depression and anhedonia in their everyday lives.</p>