Social rewards (e.g. smiles) powerfully shape human behavior, starting from early childhood. Yet, the neural architecture that enables differential processing of social and nonsocial rewards remains largely unknown. Few previous studies that directly compared social vs nonsocial stimuli have used stimuli that have low ecological validity or are not matched on low-level stimulus parameters – limiting the scope of inference. To address this gap in knowledge, social and nonsocial reward images taken from the real world were matched on valence, arousal, and key low-level stimulus properties and presented to 37 adults in a functional magnetic resonance imaging (fMRI) study. Individual self-reported preference for social images was associated with the functional connectivity between the left anterior insula (LAI) and medial orbitofrontal cortex (mOFC), as well as that between the left Fusiform Gyrus (LFG) and the Anterior Cingulate Cortex (ACC). Autistic traits negatively modulated LAI – mOFC connectivity and LFG – ACC connectivity. Reduced functional connectivity between these regions may contribute to the lower social reward responsivity in individuals with high autistic traits, as also noted from their lower valence ratings to social rewards. This study provides evidence for a new experimental paradigm to test social reward processing at a behavioral and neural level, which can contribute to potential transdiagnostic biomarkers for social cognitive processes.

Autism Spectrum Disorder (ASD) is defined as a unidimensional condition, and autism traits are measured on a continuum where the high end of the spectrum represents individuals likely to have an ASD diagnosis. However, the large heterogeneity of ASD has thrown this unidimensional conceptualization into question. With the exact underlying cause(s) of autism yet to be identified, there is a pressing need to establish core, underlying dimensions of ASD that can capture heterogeneity within the autism spectrum, thereby better specifying both autistic traits and ASD symptoms. Here we describe one important transdiagnostic dimension, the cognitive rigidity-flexibility dimension, that may impact autistic traits and symptoms across symptom-relevant cognitive domains. We first discuss how diminished cognitive flexibility manifests in core autistic traits and autism symptoms in perception, attention, learning, social cognition, and communication. We then propose to supplement assessments of autistic traits in the general population and autism symptoms in individuals with an ASD diagnosis with a comprehensive batter of cognitive flexibility measures in these symptom-relevant domains. We conjecture that systematic differences in domain-general versus domain-specific cognitive flexibility can distill subgroups within the autism phenotype. While we focus on the cognitive flexibility dimension here, we believe that it is important to extend this framework to other higher order dimensions that can capture core autism symptoms and transdiagnostic symptom severity. This approach can characterize the latent, multi-faceted structure of autism, thereby yielding greater precision in diagnostic classification and the creation of more targeted interventions.

Traditional psychological research has often treated inter-subject variability as statistical noise (even, nuisance variance), focusing instead on averages rather than individual differences. This approach has limited our understanding of the substantial heterogeneity observed in neuropsychiatric disorders, particularly autism spectrum disorder (ASD). In this introduction to a special issue on this theme, we discuss recent advances in cognitive computational neuroscience that can lead to a more systematic notion of core symptom dimensions that differentiate between ASD subtypes. These advances include large participant databases and data-sharing initiatives to increase sample sizes of autistic individuals across a wider range of cultural and socioeconomic backgrounds. Our perspective helps to build bridges between autism symptomatology and individual differences in autistic traits in the non-autistic population and introduces finer-grained dynamic methods to capture behavioral dynamics at the individual level. We specifically focus on how cognitive computational models have emerged as powerful tools to better characterize autistic traits in the general population and autistic population, particularly with respect to social decision-making. We finally outline how we can combine and harness these recent advances, on the one hand, big data initiatives, and on the other hand, cognitive computational models, to achieve a more systematic and nuanced understanding of autism that can lead to improved diagnostic accuracy and personalized interventions.