Robot Interaction May Help Youngsters

USC studies document that children with autism disorders actively interact with robots. Creation of therapy tools is the next step.
By Eric Mankin
Recent experiments examined the interplay of a bubble-blowing robot and autistic children.

Photo/Eric Mankin
Papers delivered at three conferences in the United States and Europe report on new research at the USC Viterbi School of Engineering featuring interactions of children with Autism Spectrum Disorders (ASD) and bubble-blowing robots.

The preliminary studies, by professor Maja Matarić and Ph.D. student David Feil-Seifer of the USC Interaction Laboratory, confirm what has been widely reported: In many cases, ASD children interact more easily with mechanical devices than humans.

For video of the robot, visit and type Bubblebot.

Matarić and Feil-Seifer, both specialists in Socially Assisted Robotics, are engaged in further research to confirm their findings, and to develop a robot “control architecture” which will tailor robot interactions to the specific needs of ASD children. Their goal: to help therapists treating the condition.

The initial study, reported at the Conference on Interaction Design for Children with Special Needs held in Chicago, tested whether interaction as opposed to simple passive observation was going on between ASD children and a bubble-blowing robot on wheels.

The robot had two settings. In one, it carried on its rolling and bubble blowing on its own internal schedule, regardless of the behavior of the child. In the other, “when the child pushes a button, then the bubbles blow,” in the words of the Chicago presentation.

The researchers watched the children and observed differences.

“We found that the behavior of the robot affects the social behavior of a child (both human-human interaction and human-robot interaction). Social behavior with a contingent robot was greater than with a random robot,” the researchers wrote in their study.

“Total speech went from 39.4 to 48.4 utterances, robot speech from 6.2 to 6.6 utterances and parent speech from 17.8 to 33 utterances. Total robot interactions went from 43.42 to 55.31, with button pushes increasing from 14.69 to 21.87 and other robot interactions going from 24.11 to 28. Total directed interactions (interactions that were clearly directed at either the robot or the parent) went up from 62.75 to 89.47. Generally, when the robot was acting contingently, the child was more sociable.”

While only four children were part of the initial study, Feil-Seifer and Matarić believe the work clearly demonstrates the ability of robots to actively engage with ASD children or to “offer a doorway into their attention,” Matarić said.

A much more extensive follow-up with more subjects is in progress in collaboration with Childrens Hospital Los Angeles and the Autism Genetic Resource Exchange.

Two other presentations by Feil-Seifer and Matarić at the 11th International Symposium on Experimental Robotics 2008 in Athens, Greece and at the IEEE Proceedings of the International Workshop on Robot and Human Interactive Communication announced these results in more detail, particularly in regard to the “Behavior-Based Behavior Intervention Architecture” they have developed to make the robots flexible.

This architecture is based on an ASD therapy format called DIR/Floortime in which a therapist shares the floor with various toys used to try to engage the child.

For years, Matarić has been working in the field of socially assisted robots to help a variety of other user populations, including patients with Alzheimer’s disease and stroke victims receiving help in rehabilitation. She noted that ASD is now at “epidemic” proportions in the United States.

“I am gratified by these preliminary results,” she said. “I believe that Socially Assistive Robotics has a part to play in helping families, both the affected children and their parents and siblings.”

Clara Lajonchere of Childrens Hospital of Los Angeles and Michele Kipke of the Autism Genetic Resource Exchange played key roles in the work and will continue to collaborate with the USC researchers.

The research was funded by the USC Provost’s Center for Interdisciplinary Research, the Okawa Foundation and an National Science Foundation Computing Research Infrastructure Grant.