Classifying interpersonal synchronization states using a data-driven approach: implications for social interaction understanding

This study presents a data-driven approach to identifying interpersonal motor synchrony states by analyzing hand movements captured from a 3D depth camera. Utilizing a single frame from the experiment, an XGBoost machine learning model was employed to differentiate between spontaneous and intentional synchrony modes with nearly 90 % accuracy. Our findings demonstrate a consistent pattern across subjects, revealing that movement velocity tends to be slower in synchrony modes. These insights support the notion that the relationship between velocity and synchrony is influenced by the cognitive load required for the task, with slower movements leading to higher synchrony in tasks demanding higher cognitive load. This work not only contributes to the limited literature on algorithms for identifying interpersonal synchrony but also has potential implications for developing new metrics to assess real-time human social interactions, understanding social interaction, and diagnosing and developing treatment strategies for social deficits associated with conditions such as Autism Spectrum Disorder.