Robots help improve social skills of children with Autism Spectrum Disorder, Yale study shows

Advertisement

The social skills of children with Autism Spectrum Disorder (ASD) greatly improved after a month of working with robots, according to a study led by Yale researchers.

The robots, modeling eye contact and other social behaviors, guided the children through storytelling and interactive games designed to promote social skills such as emotional understanding, taking turns, and seeing things from others’ perspectives.

Children from 12 participating families participated in the study, which was led by Brian Scassellati, professor of computer science.

The study began with a baseline assessment of the children’s skills and emotional development. A month later, a robot was placed in each of the children's homes, and for 30 minutes every day for 30 days, the children worked with the robots. For the last part of the study, the children were assessed again by researchers one month later to measure their progress.

Clinical data showed the improvements in social behaviors of the children, including the scores of the six interactive games, researchers say. By the end of the study, the children were also noticeably better at making eye contact and initiating communication, the children’s caregivers reported.

“The children showed improved performance across the board,” Scassellati says. “This was more than we had hoped; not only did the children and parents still enjoy working with the robot after a month, but the children were showing improvements that persisted even when the robots were not around.”

The robots used for the study are known as social robots, as they are designed specifically to interact with humans. The robots are controlled by a “sophisticated program” that allows the robots to adapt to an individual child’s ways of learning. This means that the robots’ games could undergo a great change depending on how well the children did.

​An important part of the study was to make sure that the sessions were both fun and challenging for the child, according to scientists involved in the study. Yale says that previous research has shown that short sessions with robots can have “significant short-term learning results” for children with autism, but children can often lose focus by the third or fourth day. 

The robots in this study did not run into that issue, though, as they did a much better job at maintaining the children’s interest, with their caregivers reporting that the children were happy to work with the robot throughout the 30 days.

Researchers say that the study also showed that the robots could work well outside of the laboratory setting, and adapt to the unique environments of each home.

With years of studying human-robot interactions under his belt, Scassellati says he has noticed that children with ASD respond particularly well to robots, as many of them are intimidated by common social situations. The children tend to feel safer interacting with robots.

“These are kids who have years of experience with the idea that social interaction is challenging and something they don’t understand,” Scassellati says. “When they interact with the robot, though, it triggers social responses but it doesn’t trigger a lot of the other baggage they’ve come to associate with social interaction.”

Scassellati adds that the study represents a significant advance in the lab’s work with social robots, and says that future studies will look to give the robots an even more prominent role.

“We have been working toward this result for almost 15 years, but always in the past with robots that operated for only short periods and only under very controlled laboratory conditions,” Scassellati says.

“This is the first study that put fully autonomous robots directly into homes, but we will need these robots to operate for more than a month in order to support social skill learning more broadly.”

Five universities participated in the study, including 15 faculty members that come from a wide range of disciplines.

The U.S. National Science Foundation under the Expeditions in Computing program supported this research.