Design of intelligent artificial agents: Its application in joint attention task for children with autism

Recently computer-based applications for skill training is gaining popularity. An important ingredient is a mediator. Thus, the design of realistic artificial agents (agent) enacting the role of the mediator is important. The agent needs to look like a real human being demonstrating naturalistic behavior. The agent can possess complexion, hair color, etc. depicting the user’s nationality. Also, the agents are empowered with intelligence such as in agent-mediated joint attention (JA) task. The JA skill refers to one’s ability to attend to an object in the shared space with a social partner. However, currently existing computer-based agent-mediated JA tasks either use agents that are not human-like or that show only the face, stealing realism. In our present research, we have designed agents with Indian look and capable of behaving naturally. Also, the intelligent engine empowers them to intelligently administer JA tasks. Results of a study with six typically developing (TD) children and six with autism spectrum disorder (ASD) are promising.