Towards designing a social interaction model based eXplainable Autism Spectrum Disorder (ASD) care agent

Autism Spectrum Disorder (ASD) is a neurological disorder that impacts a subject’s ability to be involved in a social interaction. A large body of work exists on the detection of ASD using machine learning (ML) and deep learning (DL) algorithms. Further the use of eXplainable artificial intelligence (XAI) algorithms is being advocated in the healthcare domain (on account of the ‘black-box’ nature of DL algorithms), to the best of our knowledge, no social interaction model based eXplainable Autism Spectrum Disorder (ASD) care agent exists for social interaction training of an ASD subject. Therefore, we make a beginning in this direction through this work. We present the interaction protocol and the social interaction training model for the design of an ASD care agent and elucidate its working through six representative social interaction scenarios. Further, we also put forward interesting discussions that emanate from the behavior of the ASD care agent in these scenarios and some practical considerations that need to be accounted for when the agent is put in practice with the ASD subject. We hope that the interested researchers can develop further on the work in future.