Storytelling Robots for Training of Emotion Recognition in Children with Autism; Opinions from Experts

Social robots are being increasingly used in the therapy of children with autism spectrum disorder (ASD). However, robot interaction is often designed by HRI researchers who are not fully familiar with cognitive challenges faced by children with autism. This study aimed to validate a social robot interaction designed for emotion recognition training for children with autism by seeking opinions from ASD educators and experts. A total of 26 participants (13 ASD experts and 13 non-experts) filled out a survey in which they watched videos of six emotional gestures performed by a NAO-robot. The emotional gestures were prepared with and without situational context presented in form of storytelling by the robot. Participants first made a recognition of the robot emotion in each gesture and then evaluated the feasibility of gesture recognition for children with ASD. Results showed that for almost all emotions, addition of context by storytelling significantly increased the feasibility of gesture recognition. Gestures were considered as not feasible for children with ASD when storytelling was missing and that in general, experts gave a significantly lower feasibility score to robot gestures as compared to non-experts. Our findings suggest that creation of context play an important role in the design of robot gestures which can make the training of social skills in children with ASD more effective. Additionally, the observed difference in the evaluation of the two groups suggests that social robot interventions should be validated by professionals who are more knowledgeable about social and cognitive difficulties experienced by these children.