AI-driven wearable acts as social guide for children with autism

An AI-powered intervention supported by Google Glass reinforced facial engagement and emotion recognition in a study of children with autism spectrum disorder (ASD), suggesting the digital approach might be a viable way to support those patients without spending five figures on conventional therapy.  

Study lead Catalin Voss, MS, a PhD candidate at Stanford University, and colleagues developed their “Superpower Glass” intervention as an alternative to conventional applied behavioral analysis (ABA), a technique used in kids with ASD to facilitate social skills like maintaining eye contact, paying attention and recognizing facial expressions. Experts recommend 20 hours per week of ABA, but that costs the average family between $40,000 and $60,000 per child per year.

“Furthermore, the increase in prevalence of ASD has outpaced availability of behavioral therapists, creating waitlists of up to 18 months in the U.S.,” Voss and co-authors wrote in JAMA Pediatrics, where their work was published March 25. “Learning aids based on novel ubiquitous technologies using machine learning can begin to address these problems by creating opportunities for therapy that are accessible outside of the clinician’s office.”

Superpower Glass runs on Google Glass—the tech giant’s eponymous brand of smart glasses—and is wirelessly connected to a smartphone app that provides real-time interventions to the child wearing the glasses. Voss et al. confirmed the feasibility of fit and the software’s augmented reality form factor in lab studies and in a field study of children with ASD.

Their field test suggested three 20-minute play sessions per week with the glasses for six weeks could improve social behavior, so the authors tested the efficacy of the tech in children aged 6 to 12 who had a formal ASD diagnosis and were being treated with applied behavioral analysis therapy. A total of 71 children were enrolled in the study, 40 of whom were randomized to standard therapy plus the Superpower Glass intervention and 31 of whom were randomized to standard therapy alone.

Voss and colleagues measured the success of their intervention with four unique socialization measures: the Social Responsiveness Scale (SRS-II), the Vineland Adaptive Behavioral Scale (VABS-II), the Developmental Neuropsychological Assessment (NEPSY-II) and the Emotion Guessing Game (EGG), a game designed by the team at Stanford to evaluate a child’s ability to correctly label emotions.

The authors found the VABS-II socialization subscale score significantly increased between the start and end of the intervention in the study group, with a mean treatment impact of 4.58 points. EGG, NEPSY-II and SRS-II scores also reflected larger positive average changes in treatment participants compared with controls, but those changes weren’t significant.

“Additional gains at posttest 2 were observed for SRS-II and EGG, but given a lack of posttest 2 control data, we cannot rule out a practice effect for these measures,” Voss and co-authors wrote. “Overall, these results support the hypothesis that the Superpower Glass intervention can improve social skills of children with ASD as an augmentation to standard-of-care therapy.”

The authors said these smart tools don’t replace traditional ABA, but they can help generalize kids’ social skills to a broader environment and act as a “care bridge” while they wait to see a therapist.

“The intervention reinforces facial engagement and emotion recognition, suggesting either or both could be a mechanism of action driving the observed improvement,” Voss et al. wrote. “This study underscores the potential of digital home therapy to augment the standard of care.”