The Effects of Robot-Child Interactions on Patterns of Joint Attention and Verbalization of Typically Developing (TD) Children and Children with ASDs/ADHD Between 4-10 Years of Age

Background: Verbal and nonverbal communication delays are hallmarks of Autism Spectrum Disorder (ASD) and are often addressed by traditional autism interventions. However, motor impairments in coordination, praxis, and imitation are often present yet seldom addressed through intervention. Currently, we are evaluating how an embodied social intervention that involves an imitation game between a robot, an adult, and a child affects the social skills of typically developing (TD) children and children with ASD and Attention Deficit Hyperactivity Disorder (ADHD) between 4-10 years of age. Specifically, we would like to see whether such a motivating context enhances the nonverbal and verbal communication of the child with the adult controlling the robot.

Objectives: To examine the effects of robot-adult-child interaction training on the rates of joint attention bids and duration of verbalization in typically developing children and children with ASDs/ADHD between 4-10 years of age.

Methods: Fifteen typically developing children and four children with ASDs/ADHD were examined. Each child received eight, 30-minute sessions of robot-child interaction training across four weeks. During the training, each child copied a 7-inch tall, humanoid robot that performed four to five actions based on a karate or dance theme. We examined the rates of joint attention bids to the adult as well as the percent duration of verbalization directed to the adult during the first, mid, and last sessions. We also examined the same during the pre and posttests involving imitation of five novel actions. We divided each code into spontaneous as well as responsive forms of communication. 

Results: In terms of verbalization, our preliminary analysis of six children suggests that percent of spontaneous verbalization was greater for the last session (25.0±2.0) as compared to the first session (8.0±2.0). No clear differences were observed for responsive verbalizations. In terms of non-verbal communication, analysis of the entire dataset suggests that total rates of joint attention bids to the tester increased significantly during the baseline and robot-imitation conditions of the posttest as compared to the same in the pretest (p<0.001). These trends were not observed during training due to increased complexity of actions in the latter sessions as compared to the earlier sessions. We are coding the looking durations to the robot, adult, or other to confirm our hypothesis. Moreover, children with ASDs/ADHD showed small improvements to no change in terms of verbal and nonverbal communication.

Conclusions: TD children showed enhancements in social communication as seen by increased spontaneous verbalization durations as well as total rates of joint attention bids after training as compared to their baseline. Relatively, small to no improvements were seen in the children with ASDs/ADHDs. This could be attributed to the short training period of eight sessions. Our future studies will focus on extended periods of robot-child interaction training using a more capable robot.