A child with autism spectrum disorder, alongside their caregiver, plays a game with Jibo to practice their social skills.

Brian Scassellati/Yale University

How 30 days with an in-home robot could help children with autism

For many children with autism spectrum disorder (ASD), recognizing and responding to eye contact, body language, and tone of voice is a major challenge. Improving those social skills can take lots of work—putting a strain on caregivers with limited time, resources, and money for therapy. Now, a study shows that just 30 days with an in-home robot that provides social feedback can dramatically improve a child’s interactions with others.

Researchers have long known that robots—and games with automated feedback—can change the behavior of children with autism, at least in the short term. Such interactions have been shown to help children pick up on social cues, such as making sustained eye contact, that they might have missed from their caregivers. But translating these new skills into better person-to-person interactions may require longer and more intensive training, and few studies have been large enough—or long enough—to show significant, long-lasting improvements.

So Brian Scassellati, a robotics expert and cognitive scientist at Yale University, put together an experiment that gave children a long-term relationship with their bots, one they could share with their families. His team provided 12 families with a tablet computer loaded with social games and a modified version of a commercially sold robot called Jibo, which was programmed to follow along with the games and provide feedback. “As a roboticist, that was one of the most frightening things in the world. Leaving the robots there and hoping they would do the things we’d programmed them to do,” Scassellati says.

For 30 minutes each day during the 30-day study, the children sat next to their parents or other family members and played games while Jibo interacted with them. The games were based on clinical therapy techniques for improving different social skills, including social and emotional understanding, taking another’s perspective, and completing tasks in a sequence. For example, in a game called Rocket played on a computer tablet, the child and the caregiver take turns building a rocket ship by dragging parts around the screen. The screen is then reset to hide this design, and the first player must explain to the second player how to recreate the design.

Throughout the games, Jibo would model good social behavior by focusing its eyes on the child or the caretaker and orienting its body toward them. It also told the child to do the same. “The whole point of the robot,” Scassellati says, “is to make the interaction between the child and the parent better. All the skills that they’re learning, they’re learning to direct … to the parent, not to the robot.” At several points, the robot would call the child by name, asking them to engage with their caregiver by looking at them. After each session, the system’s software changed the games’ difficulty based on the child’s performance.

Jibo, a humanlike robot, watched the children play the games, offering directions and modeling good social behavior by rotating its head and body, and making contact with its animated eyes.

Brian Scassellati/Yale University

Caregivers also scored children on how they behaved around other people (including themselves) and how engaged they seemed during the games. Researchers filmed the sessions, looking for changes in body language and social cues among the children. Finally, the researchers tested the children on “joint attention,” which measures how well someone can alert another to an object by means of eye gazing, pointing, or verbal description. That testing happened four times: 30 days before the experiment, on the first day of the games, on the last day of the games, and again 30 days after the end of the intervention.

Caregivers’ reports showed that the social skills of all 12 children improved over the course of the study, the researchers report today in Science Robotics. They were more responsive to communication, initiated more conversations, and made more eye contact with others. The children’s average joint attention scores also improved between the first and last days with the robot, increasing by 33% before dropping slightly 30 days after the study ended.

Scassellati says he isn’t surprised by the drop after playing with the robots stopped. “Even if I put the best therapist that I have in your home, we wouldn’t expect to see a permanent change after just 30 days.” So far, the results of the robot therapy are similar to other therapies for ASD, in terms of changes over time, he says. “We can’t say at this point that the robot actually produces long-term, lasting social change, but what we do see is very promising.”

Even though the study can show changes in individual children’s behavior, what it couldn’t measure is how well the bot therapy works relative to other treatments, says Elizabeth Broadbent, who studies human-robot interactions in health care at The University of Auckland in New Zealand. That’s because each child continued to receive their regular ASD treatment and therapy during the study.

Scassellati says more work is needed to test the effectiveness of the robots with larger groups of children over longer periods of time. He adds that the robots were never intended to replace traditional treatments and therapy, but to magnify the effectiveness of these scarce resources. “Most families can’t afford to have a therapist with them every day,” Scassellati says, “But we can imagine having a robot that could be with the family every day, all the time, on demand, whenever they need it, whenever they want it.”