Characterising gameplay development in autism with swipe pattern networks

Autism spectrum disorder (ASD) is a neurodevelopmental condition affecting at least 700,000 individuals in the UK [1] with an aggregate annual healthcare and support cost of at least £28 billion [2]. Early identification, proceeded by therapeutic intervention, can produce significant, lifelong health and economic benefit. An ASD diagnosis currently requires a trained clinician, but there is a long and growing waiting list for such assessments. To meet demand, and create more accessible means of assessment, bespoke touchscreen games have been developed for early autism detection and trialled for children aged 2.5–6 years. These games focus on recognising ASD through detecting disorder in intentional movements [3]. In this study, we employed a serious iPad game for young children (441 without known neurodevelopmental problems, 373 diagnosed with autism spectrum disorders, and 64 diagnosed with other neurodevelopmental disorders) [4] that allowed for different play patterns, but where the child was encouraged toward a social aspect of gameplay (sharing food) with attractive sensory feedback. Children were encouraged to drag four pieces of food from a serving area (food zone) to deliver them to a set of four children (snap-to-plate zones) to trigger feeding animations and an audible celebration (Figure 1a). By converting gameplay swipes into a graph, we can identify – for the first time – the specific pattern signatures of autistic users. We find that autistic participants employ an indirect, two-step, sharing process in contrast to the direct, singlestep, approach employed by children without neurodevelopmental disorders. These insights into the serial organisation of play actions could form the basis of effective diagnosis and tailored therapeutic interventions.