Robots as Co-Therapists in Behavior Therapy for Individuals with ASD

Recent technological advances have opened the possibility of using robots in therapy for individuals with Autism Spectrum Disorder (ASD). This approach has received a lot of press, but to date most research has focused on developing the robots rather than clinical issues related to the use of robots. These clinical issues include both potential benefits and drawbacks to the approach. For this study, we decided to examine the integration of a robot as a co-therapist in a commonly used behavior therapy approach known as Applied Behavior Analysis (ABA). The goal of this study was to examine whether or not children with ASD responded to the robot co-therapist, in comparison to the same type of treatment with only a human therapist. It is important to note that we are not developing a robot or any type of technology for sale. We purchased an existing robot (NAO, Aldebaran Robotics) to test our clinical questions. It should also be noted that we operated independent of this company and received no benefits from them.

Participants were 19 individuals with ASD between the ages of 6-13 years with varying levels of communication and cognitive abilities. Participants completed 12 therapy sessions, half of which included both a human therapist and a robot co-therapist, and half of which included only a human therapist. We also collected data on how they were doing before the therapy, immediately after the therapy, and three months after the therapy was completed. In the therapy sessions, we taught basic social skills, such as responding to questions, making social comments, and asking appropriate questions. The child would practice the target skills with the human therapist, and then they would practice them with the robot while being coached by the human therapist. The NAO robot was controlled remotely by an unseen person, who observed the session and selected appropriate responses for the robot. This unseen person controlled the robot by initiating pre-programmed responses (“What did you do today?”) in the robot, but also typed in responses for the robot to say in other circumstances. Interestingly, when children were not interacting with the robot, we used the person who had been controlling the robot as a practice partner in the “no-robot” sessions. Essentially, in both of the types of therapy sessions that we were comparing, the child was interacting with the same person, except half of the time that person was communicating through an interactive robot.

Overall, we found that the inclusion of a robot co-therapist lead to increased social skills gains, compared to sessions that did not include the robot (but in many cases included the practice partner). Importantly, we found that children who learned skills while interacting with the robot would then use these skills with their human therapist, and at home as well. It is important to note that there was considerable individual variation in how children responded to the robot; a significant majority of children learned skills more quickly when the robot was present, but a minority of children responded better to the human therapist, and a few children showed no difference between therapists. This is an important point because it should not be assumed that this approach will work for all children with ASD.

There are a number of important implications of this study. First, there seems to be a benefit to practicing simple social skills rules, such as asking questions or responding appropriately to questions, with a robot. It is possible that the robot simplifies the social situation (no facial expressions, limited gestures), and allows the child to focus on the skill that is being taught. It is also possible that the presence of a talking, interactive robot is so novel and socially motivating that it provides an intrinsically interesting situation in which the child can practice skills. We did not specifically test what led to this improvement in our participants, so this is an area for future research.

A second important implication is that this approach will not be appropriate for all children. It will be important to determine what factors lead some children to respond to the approach, and others not to respond. This will allow parents and clinicians to be appropriately informed when deciding whether or not to use this approach.

Should robots be used in therapy right now? We think that there is considerable work that needs to be done before this becomes a widespread approach for teaching skills to individuals with ASD. Minimally, robots (as with any technology) should not replace human interaction. In essence, a robot therapist will not replace a human therapist. We believe that this approach will be beneficial in cases where it is used to target specific skills that can then be transferred to other people in the child’s life. Robots can be quite expensive (the robot in this study cost over $10,000, for example), and it will be important to work on ways to minimize the cost of the approach, without losing its effectiveness, before robots will be recommended for widespread use. Additionally, our robot did not act on its own; it was controlled by someone, which is a setup that requires additional resources. Still, we consider this work a first step toward understanding the potential benefits of this approach.


Joshua John Diehl, PhD, is the William J. Shaw Assistant Professor of Psychology at the University of Notre Dame. Charles R. Crowell, PhD, is Associate Professor of Psychology at the University of Notre Dame. Michael Villano, PhD, is a Research Assistant Professor of Psychology at the University of Notre Dame. Kristin Wier, MA, LLP, BCBA, is Clinical Director of the Autism Learning Center at the Sonya Ansari Center for Autism, South Bend, Indiana. Karen Tang, MA, is a Graduate Student at the University of Notre Dame.

If you have any questions, feel free to contact us at, 574-631-5729, or visit our website at: You can also follow us on Twitter at @ND_FUN_Lab or #funlab.




This work was sponsored by a Rodney F. Ganey Collaborative Community-Based Research Mini-Grant, an NIH Indiana Clinical Translational Sciences Institutes (NIH RR025761) – Notre Dame Project Development Team Grant, the Boler Family Foundation, and the Notre Dame Institute for Scholarship in the Liberal Arts. None of the members of our team have any connections to the Aldebaran Robotics team. We neither offered, nor would we have accepted, any benefits from the company during this project, outside of the standard support provided through the purchase of the robot and paid maintenance plan.

Have a Comment?