Using Computational Measures of Social Communication Dynamics for Children with Autism Spectrum Disorder

Background:  Children with Autism Spectrum Disorder (ASD) exhibit impairments in social interactions and at the core of these impairments are social communication deficits. Recent advances in the quantitative and computational measurement of conversational content has resulted in a novel set of methods that might provide a more objective and reliable way of identifying the conversational biomarkers of ASD, as well as a better understanding of the time-evolving dynamics of social communication in these individuals.

Objectives: The current study has two objectives: (1) to validate the use of newly developed computational measures of conversational interaction for assessing deficits in social communication in adolescents with ASD; and (2) to further identify whether deficits in social communication are interrelated to deficits in the social motor coordination that supports effective social interaction.

Methods: Thirty children previously diagnosed with ASD completed the Autism Diagnostic Observation Schedule, Second Edition (ADOS-2). The conversations that each participant had with the clinician during the ADOS-2 administration were evaluated using Discursis, a computational time-series method that analyzes the conceptual and semantic content of an exchange between two or more individuals. Various measures were extracted to quantify different aspects of the conversation (e.g. self-similarity, other-similarity). Additionally, measures of the child’s and administrator’s social motor coordination were obtained during these conversations (e.g. coherence). Of particular interest is the relationship between ASD symptom severity and the dynamical measures of communication and social motor coordination and the degree to which some of these dynamical measures better predict ASD communication deficits.

Results: Some Discursis measures were correlated with some ADOS-2 sub-category scores (e.g. social affect), as well as the composite score obtained from the test. Stepwise regressions confirmed that Discursis measures can be used to predict composite scores and traditional multiple regression showed that by including a measure of social coordination we are able to account for more of the variability present in ADOS-2 composite scores as well as some sub-category scores.

Conclusions:  This data indicates that Discursis could be a sensible and important addition to our diagnostic procedures that would help us better understand the communicational deficits exhibited by some children with ASD. Furthermore, it seems that the social motor coordination that takes place during conversations is interrelated to the verbal communication and necessary to quantify for further understanding of this deficit.