Community Perspective: Trayt CMO Dr. Carl Feinstein on AI in Diagnosis and Treatment of Autism


For many years, the tech community has made important contributions to the medical sphere. For example, the development of protheses has helped people with disabilities overcome sensory deficits, loss of function and hearing/speaking impairments. And now, technologists are turning their efforts to remediating social communication barriers that hamper people with autism spectrum disorders (ASD) in their adaptation to the community, school, work and cultural/recreational opportunities.

The urgency to invent tech-based interventions for ASD children is heightened by: the frustrating bottlenecks and delays in obtaining quality care, which can be demoralizing to parents seeking help for their children; and the expense of traditional office-based professional care. With early, effective interventions to improve social adaptation, the entire life trajectory of ASD children can be improved. These innovative approaches and initiatives can be found at universities, specialized medical centers, industry and tech startups as well as by people who are themselves “on the spectrum.”

Software and accessible interactive learning devices (including robots and avatars) are attractive options to increase accessibility and decrease costs of treatment, including behavioral and educational interventions. Given the enormous expansion of computing power and sophisticated software engineering, it makes perfect sense that the technology community would seek to tackle such a challenging task. I encourage efforts to enhance the ability of ASD children to pick up social clues better and reduce the barriers to communication, learning and friendship in our “neurotypical” world.

However, it’s worth noting that many of the AI, virtual or augmented reality and avatar-based screen interventions are ambitious and seek to remediate what has proven to be the most intractable aspect of ASD-based adaptive problems: difficulty with social communication.

Areas of caution and advice

Social communication is a pinnacle skill, one of the most impressive and complex attainments of the human brain. It develops from early childhood through interaction with loving and intuitive parenting, which is itself a peak human skill. Most people, including even the brightest and best of our engineers, have only a dim sense about how complex social and communication development in childhood really is. Even the best child development experts would acknowledge that there is much about childhood social development that is still poorly understood. There are bound to be many false starts and blind alleys when skilled engineers, who are novices in child development, tackle the diagnosis and treatment of ASD in children.Over-generalizations about Autism is another conceptual barrier that could interfere with rapid technological breakthroughs. Just as every “typically” developing child is unique, there is no “standard type” of childhood ASD. Every child with autism has his or her unique attributes. A highly tailored treatment that helps one child might be ineffective or even adverse for another. One single-purpose tech tool is unlikely to work for a myriad of problems in a range of children with different profiles of strengths, deficits and atypical behaviors.

Fortunately, there is an approach that will help surmount these obstacles, and that is what Trayt stands for. We start with each family (even better if they are working with helpful clinicians and therapists) carefully identifying a small number of communication, social, medical or behavioral goals that, if achieved, results in meaningful progress for their child. Then, when a treatment is selected, the behavioral outcomes related to those goals will be rated and tracked over the following days, weeks and months. That way, improvement (or lack of improvement) can be seen and the question answered: “Does this treatment work?”

For tech developers, the approach must be the same. Tech treatments must be proven to work, just like any other behavioral or medical treatment. It’s unfair to develop and promote a device or AI-based treatment to families without solid evidence that it helps children in measurable ways. And in the case that the technological approach has been shown to help some children, responsible application to other children must be accompanied by a reliable method of tracking behavior changes over time.

In summary, the world of technology has great potential to help children with ASD, but a poor understanding of childhood social development or treatments based on the fallacy that all ASD children are alike will inevitably have poor results. Furthermore, every child receiving treatment needs to be supported by a method of tracking outcomes to be sure that the treatment is actually working. Clinical outcomes must drive the evolution of technology. Technology only provides a viable solution if it closes the clinical gap that is causing poor outcomes. Technology needs to simplify the treatment, without oversimplifying the condition.