PROJECT SUMMARY/ABSTRACT Advances in genetics, molecular biology, and cognitive neuroscience offer promise towards personalized treatment and improved outcomes in individuals with Autism Spectrum Disorder (ASD). However, the promise of precision medicine has been hindered by a lack of mechanistic models that explain phenotypic and etiological heterogeneity; instead of using such models to identify subgroups likely to respond to specific treatments, the field relies on availability of service, trial-and-error, and clinical judgment to make treatment decisions. This is a major barrier to effective treatment of ASD. This project addresses this problem by integrating mathematical models of behavior and brain activity across adolescent development, in order to establish a neurocognitive model that can successfully predict individual adolescents' social and nonsocial learning profiles at key developmental time-points. Specifically, this work compares the suitability of various reinforcement learning models to capture selective deficits in social learning of adolescents with ASD, as well as variability in both social and nonsocial learning across typically developing (TD) adolescents and those with ASD. Identifying how these model-based predictions are implemented in brain circuits may allow for characterization of the neural architecture underlying learning in therapeutically relevant contexts. This project focuses on the understudied role of the cerebellar posterior lobe in learning processes of interest, given recent research indicating this region's involvement in updating social information. The proposed work will identify and characterize neurocognitive variability in the substrates of learning within ASD, with the long-term goal of applying these models to inform, refine, and individualize diagnosis, prognosis, education, and treatment of youth with ASD.