Wearable sensors can track social interaction in groups of autistic and non-autistic children

People often spontaneously synchronise with one other during social interactions, but this synchrony may be weaker in autism. There are few methods available for measuring social synchrony outside the lab, which makes it hard to know what natural patterns of synchrony should look like. Here we present SocSensors, a system that uses wearable sensors to evaluate interpersonal synchrony, and use it to uncover social behaviours in young autistic people with learning difficulties. We used wrist-worn accelerometers to collect data from the interactions of 3 groups of children (and adults) during school activities: autistic children aged 5-6, autistic children aged 12-17, and neurotypical children aged 4-5. We evaluate a wavelet-based method to calculate interpersonal synchrony between all possible pairings. The output of this analysis and the proposed visualisations provide a convenient way to estimate social engagement. We compare our measures to blind independent video ratings of social engagement and find a clear positive correlation, which validates the use of sensors for in-the-wild studies of social behaviour in autism. We also demonstrate a range of analyses for which SocSensors can be used, including quantifying individual differences in social behaviour, uncovering social relationships within a group, and uncovering group differences of interpersonal coordination between autistic and neurotypical children. We also provide an analysis toolbox for others to build on our approach. The results show how wearable sensors enable a new type of research on real-world social interactions and advance our understanding of social synchrony.