Matthew Goodwin uses CS to craft innovative support for autistic people

It might start with something as simple as an itchy sweater.

A neurotypical person could manage this annoyance, but to an autistic person with sensory processing issues, the sweater can feel like steel wool scouring their arms. Individuals with profound autism might also lack the ability to pinpoint what is causing their discomfort or the verbal skills to ask for help taking the sweater off. Discomfort can quickly become pain, then panic, and the person begins to “melt down” — screaming, crying, hitting themselves or the people around them, anything to alleviate the distress. It’s often not until the meltdown has begun that the autistic person’s family, friends, or caregivers realize there is a problem.

How much suffering could be spared if that autistic person could communicate their discomfort to the people around them, even just minutes before it becomes unbearable?

Enter Matthew Goodwin — an interdisciplinary associate professor jointly appointed in the Khoury College of Computer Sciences and the Bouvé College of Health Sciences, a founding member of the Personal Health Informatics doctoral program, and director of the Computational Behavioral Science Laboratory. His research uses wearable sensing and machine learning to develop tools and techniques that make life easier for autistic people and their caregivers. Goodwin recently received major grants from the National Science Foundation (NSF), Department of Defense (DOD), and National Institutes of Health (NIH) to work on predicting aggressive behavior minutes before it occurs in minimally verbal autistic youth, as well as helping more cognitively able autistic people regulate their stress through paced breathing, progressive relaxation, and other strategies.

“People with profound autism are a segment of the population I think we need to support the most, and yet we have the least amount of information to guide us in their needs and what the best interventions are,” Goodwin said. “It creates a focus and urgency to use computer intelligence to advance our understanding and care for human health.”

Smart and Connected Health

Most severe behavior management techniques focus solely on stopping behaviors already in progress by any means necessary. “That tends to be major sedative medications,” Goodwin explained, “which can cause weight gain, impact working memory, and reduce learning potential. Or autistic people get mechanically restrained; they wear braces so they can’t bend their arms to hurt themselves or others, or they wear helmets so that when they hit themselves, the force is reduced.”

Mechanically restraining a child in distress is traumatic for everyone involved, and the constant possibility of having to do so prevents families from engaging in public activities together. The alternatives, though, are worse.

“If the meds or mechanical restraints don’t work, or if the stress on a family becomes too great as that child becomes stronger than them, they’ll get pulled out of their homes and educational settings. They’ll find themselves in group homes, hospitals, the ER, or psychiatric inpatient units,” Goodwin continued, “These are not long-term solutions — these are acute crisis management facilities.”

To potentially avoid these devastating outcomes, Goodwin’s long-term vision is to alert autistic people and their caregivers of rising stress using data from commercially available, wrist-worn, wireless devices that measure real-time changes in the body’s cardiovascular, electrodermal, and physical activity, to avoid or preemptively manage overwhelming stimuli and distress in their environments. By developing machine learning and predictive analytics techniques and applying them to this data, Goodwin and his team can forecast aggression toward others, self-injury, and meltdowns three minutes in advance with 80 to 90 percent accuracy.

The NSF Smart and Connected Health study, the first of Goodwin’s recent grants, focuses on enriching his behavior prediction models with advanced machine learning methods,intended to reduce reliance on humans to label behavioral data. Working with the Marcus Autism Center at Emory University, Goodwin and his team will spend the next four years testing their new models on biosensor data gathered in an outpatient clinic.

The fact that computer science and machine learning techniques have scaled his research as much as they have encourages Goodwin. From the beginning, he aimed to create tools and aids for disenfranchised people whose unique experiences of the world prompted him to reconsider everything from small talk to knowledge acquisition. Now he finds himself engaging in work with enormous potential considering that one in 36 children is diagnosed with autism by the age of eight.

“I didn’t set out to be a quantitative person or a machine learning person. I went to the places that seemed like they had tools and methods that can answer questions about a very complex clinical condition,” he said. “Most parents [of autistic children] will say to me, ‘If I could have this capability, it would be priceless. It would be a game changer for our ability to lead a more normal life and have some hope that our child will be as independent and self-determined as possible.”

Regulating Together

Autistic people’s experiences can be roughly categorized into three groups, Goodwin explained. The first group,those targeted by the Smart and Connected Health study, have autism that severely impacts their daily lives and independence. This group often has significant social, cognitive, and verbal challenges; physical and psychological comorbidities; and difficulty self-regulating.

Members of the second group have more mental and verbal abilities, but still deal with depression, anxiety, and stress management issues; it is for this group that Goodwin’s other research is designed. (The third group has even more social, cognitive, and verbal abilities, fewer comorbidities, and greater independence.)

Regulating Together is a program developed at Cincinnati Children’s Hospital that teaches cognitive behavioral techniques, progressive muscle relaxation, and stress-reducing mindfulness activities to parents of autistic children. The parents are trained to teach their child these techniques, perform them together, and hopefully manage stressful situations that might trigger a meltdown. The program performed well in a smaller, randomized clinical trial, and Goodwin and his colleagues at Cincinnati Children’s Hospital are now conducting a more extensive, four-year follow-up study with support from the DOD. Goodwin’s role is to collect and process ambulatory psychophysiological signals such as heart rate, heart rate variability, and electrodermal activity to determine Regulating Together’s impact on participants’ physical stress reactions before, during, and after the intervention.

The third and final study, supported by the NIH, asks a simple question: Would adding a dog to the Regulating Together program help? It’s been established that emotional support animals can benefit people with mental health conditions. Goodwin and his collaborators seek to determine whether Regulating Together methods have an even greater impact on participants’ stress reactions when a friendly puppy is in the room.

Because Regulating Together can be conducted in the home by parents, Goodwin points out, the barrier to entry is lower, allowing more autistic people — including those who are more significantly impacted — to access this service.

“It’s not like medicine, where you take a pill that works for some time, contradicts with other drugs, has side effects, and you get dependent on it,” Goodwin said. “These approaches are highly scalable. You can train parents to train their kids. It’s not expensive; it doesn’t require dependency. If we determine it’s effective at a large scale, it will be easier for many people who don’t have the same financial resources to get the same benefit.”

His more verbally able autistic research participants have been similarly encouraged by this approach.

“Most will say, ‘I’ve had such a hard time demonstrating to people that I’m not bad. I experience the world differently, and this is how I cope with it. You’re providing evidence that I’m not a bad or broken person; I just have a harder time managing the world,’” Goodwin relayed.

Goodwin’s focus has always been on that outcome: giving autistic people better tools and support to lead richer, happier, and more independent lives.

“When we talk about equity, diversity, and inclusion, we’re often thinking about race or sex, but diversity is also variation in the human experience,” Goodwin said. “I’d like to see these solutions get into the hands of the people who need them most — those who really need this tech and can’t afford it, those who are struggling in life.”

Subscribe to Khoury News