From birth, infants and their caregivers engage in mutually adaptive and mutually reinforcing interactions that serve as the platform and catalyst for development of the social brain and behaviors. These interactions are characterized by species-typical behaviors — that is, repeated patterns of behavior that are highly uniform across infant-caregiver dyads — which create critical opportunities for initiating and maintaining social interactions.
A recently advanced theory specifies that though infants with autism spectrum disorder (ASD) may exhibit predispositions to species-typical social behaviors as neonates, diminished sensitivity to social contingency may lead to increasingly atypical social behaviors and, eventually, clinical features of ASD^1. Though the foundational nature of social skill disruptions in ASD is widely accepted, no studies have investigated infant-caregiver interactions in dyads with infants with ASD before six months of age. As a result, there is a substantial gap in understanding of how early differences in infant sensitivity to social contingency may underlie ASD pathogenesis. Bridging this gap is further complicated by technical limitations of the standard methods used to identify and quantify behavior in early infancy, which prevent a nuanced investigation of the intrinsic dynamics of infant-caregiver interactions and the development of objective, scalable and quantifiable markers of risk for infants with ASD. The majority of the literature relies on manually coding predetermined, recognizable behaviors, a subjective and laborious method that struggles to reliably detect the subtle and/or unanticipated millisecond-scale behaviors common in infancy.
In the current project, Sarah Shultz and Gordan Berman plan to bridge these empirical gaps by harnessing advances in computer vision analysis and deep learning for dynamic behavior prediction to automatically detect the full repertoire of species-typical behaviors produced during social interactions and quantify social contingency, modeled as the extent to which caregiver and infant behaviors predict each other. A well-powered and well-characterized sample of dyads with typically developing infants and infants with ASD will be used to test the recently advanced theory of ASD pathogenesis described above. It is predicted that between-group differences in the type and/or frequency of species-typical behaviors will be minimal at birth but become increasingly divergent over time in parallel with diminished sensitivity to social contingency in infants with ASD. The researchers further hypothesize that atypical interactive dynamics in infants with ASD will ultimately predict increased social disability at two years of age.
By implementing this novel computational approach in a densely sampled, longitudinal dataset of well-characterized infant-caregiver dyads, this project will identify objective markers of interactional dynamics that signify risk for social disability. These findings will offer transformative insight into the behavioral pathogenesis of ASD and have notable clinical value as targets for when and how to optimize early social interventions.
- Shultz S. et al. Trends Cogn. Sci. 22, 452-469 (2018) PubMed
- Electroencephalography and eye-tracking measures as scalable biomarker-based predictors of ASD in high-risk infants
- Home-based system for biobehavioral recording of individuals with autism
- Objective measures of social interactions via wearable cameras
- Electrophysiological, metabolic and behavioral markers of infants at risk