Automated Quantification of Head Motion Synchrony in Peer Interactions of Autistic and Non-Autistic Youth

Peer interactions take on increasing importance for youth in middle childhood and adolescence. Autistic youth face particular challenges interacting with non-autistic peers; thus, it is critical to understand factors that predict interaction enjoyment for autistic and non-autistic youth. One candidate predictor is interpersonal motion synchrony. Experimentally induced motion synchrony leads to interpersonal liking and closeness, but these outcomes from experimental synchrony may not generalize to synchrony that naturally occurs in interpersonal interactions. Computer vision approaches allow for non-invasive quantification of synchrony in developmental populations, and thus provide an opportunity to investigate the role of naturalistic motion synchrony in peer interactions between autistic and non-autistic youth. Autistic and non-autistic youth were paired into three dyad types (autistic with autistic, autistic with non-autistic, and non-autistic with non-autistic) and completed a brief video-recorded "getting-to-know-you" interaction. Head position was extracted frame-by-frame from video recordings using OpenPose. Cross-wavelet coherence was computed to quantify head motion synchrony. We found significant head motion synchrony in interacting (compared to time-scrambled) dyads, but levels of synchrony did not differ by dyad type. Across dyad types, we found evidence that head motion synchrony negatively predicts youth-reported enjoyment and positively predicts youth-reported discomfort in the interaction. Unlike prior findings of relations between synchrony and positive outcomes, here, synchrony predicted negative conversational outcomes. Thus, these findings add nuance to our understanding of the role of synchrony in naturalistic interactions. They also demonstrate the application of non-invasively quantifying synchrony in minimally structured peer interactions including autistic youth using computer vision.