Robots provide high-level autism therapy
By Achilleas Martinis
In Menomonee Falls, Wisconsin, Logan – a school child with autism — regularly attends therapy sessions. But now, his therapist might not be who you’d expect.
He receives therapy from Milo, an AI-powered robot.
Logan used to hit and bite his teachers. The staff in his school even wore long sleeve shirts to protect themselves. After his school introduced therapy with Milo to Logan, he started engaging with other students and initiating conversations. His mother was impressed by the impact of the robot on him. In an interview she said, “[Logan]’s coming out of his shell that it seems autism had him in.”
There are many more success stories like Logan’s and until now the outcomes of robot-assisted therapies seem more promising than those of human-led therapies.
To understand why the robot-assisted therapy is better than the human-led therapy, we have to pinpoint the edge the robots have over humans in different areas of teaching.
Robots get children with autism to make eye-contact
Children with autism do not engage in conversations and do not make eye-contact with others. But robots excite them and keep them engaged.
Brian Scasselatti, professor and leader of the Social Robotics Lab at Yale had an autistic child have a conversation with a therapist, then exposed him to a robot dinosaur, and then had him have another conversation with the same robot. In Dr. Scasselatti’s demonstration, the robot dinosaur got the autistic boy from making no eye-contact to making the level of eye-contact you would expect from a typically developed child.
Robots get children to make mistakes they can learn from
People ask and answer fewer questions when they are alone with a human therapist than when they are with a robot therapist. This is because they are afraid of making mistakes in front of the human therapist, while they are more comfortable with making mistakes in front of a robot therapist. As a result of making more mistakes with a robot therapist, people learn more with them.
Robots make simpler facial expressions that are easier to read by autistic children
Children with autism find it hard to recognize the wide-ranging emotions expressed by a therapist’s voice and face.
According to a study by doctorates from the University of Singapore, interactions of autistic children with a human therapist “can trigger excessive sensory stimulation, causing severe distress to the child with autism.”
This makes it almost impossible for an autistic child to learn to recognize emotions from a human therapist. But this doesn’t happen with robots like Milo because they are carefully programmed to express and teach only basic emotions.
Although robot therapists are better than human therapists in these three areas, they are not yet perfect.
The robots for the therapies come at a high cost. Robokind, the company that makes Milo, reports that its robot has an initial cost of $5,000 plus an additional cost of $4,500 every year after.
This is lower than the $29,000 cost of human therapy, but Robokind’s cost estimates are still too optimistic. A 2015 study found that parents have a preference that there is a human complementing the robot in autism therapy. If humans have to complement robots in robot-assisted therapy, the cost might even be higher than the human-led therapy cost of $29,000 per year.
In addition, we don’t have large studies on robot-assisted therapies to know what specific cases of autism they would and wouldn’t work on. According to a report by Dr. Scasellati’s lab, the studies we have on robot-assisted therapies right now have small sample sizes and are done over a short time period.
So before we fully convert from human-led therapies to robot-assisted therapies, we have to get a more complete understanding of autism therapies.
Although these shortcomings are obstacles to treating the symptoms of autistic children, researchers are working to overcome them with high hopes about the future of autism therapies.