Facilitating Collaboration among Children with Autism through Robot-assisted Play

Abstract

This thesis discusses how autonomous robots can be used to foster and support collaborative play among children with autism in a number of different settings. Because autism impairs one’s skills in social communication and social interaction, this makes it particularly difficult for children with this disorder to participate in many different forms of social play, particularly collaborative play due to the interpersonal skills needed to coordinate and synchronize people’s actions through constantly communicating with them. Since these children have trouble playing collaboratively, this further hinders their ability to develop the necessary skills of interacting and communicating with others. I approached this idea from an empirical, behaviourist perspective instead of a theoretical one, in the sense that I conducted three different experiments in which I observed the behaviours of children with autism participating in controlled play sessions both with and without robots. To this end, I designed simple, effective control architectures which allowed LEGO NXT robots and KASPAR the humanoid robot to autonomously interact with people while playing with them. Additionally, I designed many collaborative video games such as arena games, “Tilt & roll”, and “Copycat”, that served as environments in which children with autism could play with the autonomous robots. The experiments in this thesis attempted to show that not only would children with autism improve their social behaviours while playing collaborative video games with autonomous robots, but these improvements would also transfer into similar settings in which the children would only interact with other people. By recording videos of the children’s interactions and performing observational analyses on the children’s behaviours, the data from my first exploratory experiment indi- cated that the amount of enjoyment the children showed in an after-school robotics was more positively correlated with their social behaviour than the number of play sessions in which they interacted. Using similar means, the results from my more streamlined second experiment suggested that children with autism displayed more social behaviours while playing with a typically developed adult after playing with KASPAR than they did beforehand, and the findings from my more rigorous third experiment strongly indicated that different pairs of children with autism showed improved social behaviours in playing with each other after they all played as groups with KASPAR compared to before they did so.