Autism spectrum disorder (ASD) is a developmental disorder that affects communication and behavior. The diagnostic tasks are often complex and cumbersome due to the mental health conditions which children with ASD suffer from, some of which are higher levels of anxiety, depression, attention deficit hyperactivity disorder (ADHD), and disruptive behavior disorders. In this work, we investigate how emotion recognition can be leveraged to achieve sustained attention during the diagnostic tasks in children with autism. We focus on capturing the emotions of the child during diagnostic tasks and the corresponding facial signals that the agent has to display during diagnostic task administration eventually leading to the successful completion of these diagnostic tasks and better diagnosis.