Hidden Social and Emotional Competences in Autism Spectrum Disorders Captured Through the Digital Lens

Elizabeth B Torres ^1* Joseph Vero ² Neel Drain ² Richa Rai ² Theodoros Bermperidi ²

• ¹ Psychology, Computer Science, Cognitive Science, Rutgers, The State University of New Jersey, New Brunswick, New Jersey, United States
• ² Rutgers, The State University of New Jersey, New Brunswick, New Jersey, United States

Background-Objectives: The current deficit model of autism leaves us ill equipped to connect with the person on the spectrum, thus creating disparities and inequalities in all aspects of social exchange where autistics try to participate. Traditional research models also tend to follow the clinical definition of impairments in social communication and emotions without offering personalized therapeutic help to the autistic individual. There is a critical need to redefine autism with the aim of co-adapting and connecting with this exponentially growing sector of society. We here hypothesize that there are social and emotional competences hidden in the movements' nuances that escape the naked eye. Further we posit that we can extract such information using highly scalable means such as videos from smartphones.Methods: Using a phone/tablet app, we recorded brief face videos from 126 individuals (56 on the spectrum of autism) to assess their facial micro-motions during several emotional probes in relation to resting state. We extracted the micro-movement spikes (MMS) from the motion speed along 68 points of the Open Face grid and empirically determined the continuous family of probability distribution functions best characterizing the MMS in a maximum likelihood sense. Further we analyzed the action units across the face to determine their presence and intensity across the cohort.Results: We find that the continuous Gamma family of probability distribution functions describes best the empirical face speed variability and offers several parameter spaces to automatically classify participants.Unambiguous separation at rest denotes marked differences in stochastic patterns between neurotypicals and autistics amenable to further separate autistics according to required level of support. Both groups have comparable actions units present during emotional probes. They however operate within parameter ranges that fall outside our perceptual umwelt and as such, do not meet our expectations from prior experiences. We cannot detect them.Conclusions: This work offers new methods to detect hidden facial features and begin the path of augmenting our perception to include those signatures of the autism spectrum that can enhance our capacity for social interactions, communication and emotional support to meet theirs.