Protocol for optimizing robot-assisted autism therapy sessions through gaze analysis: A pilot study investigating optimal trial count for children with comorbid autism spectrum disorder and intellectual disability

Background: Autism spectrum disorder poses challenges in social communication and behavior, while Intellectual disabilities are characterized by deficits in cognitive, social, and adaptive skills, frequently accompanied by stereotypies and challenging behaviors. Despite the progress made in autism spectrum disorder research, there is often a lack of research focusing on individuals with co-occurring autism spectrum disorder and intellectual disability. Robot-assisted autism therapies are effective in addressing these needs. However, there is a lack of consensus on the optimal number of therapy sessions required for effective outcomes, particularly in children with comorbid autism and intellectual disability.

Objectives: This study aims to investigate the optimal number of therapy sessions required for effective outcomes in robot-assisted autism therapy, specifically focusing on children with comorbid autism and intellectual disability. Gaze analysis metrics, including the average number of eye contact and average inter-eye contact delay, are utilized to assess therapy effectiveness.

Methods: This study integrates gaze analysis focusing on the average number of eye contact and average inter-eye contact delay as performance metrics. A pilot investigation was conducted to explore the critical temporal window for optimal trial count in robot-assisted autism therapy for children with autism. A mathematical model was introduced and validated for eight children with autism spectrum disorder, encompassing 56 experiments conducted over 7 weeks (one trial per week). Robots presented 8 cues in 3 categories (visual, speech, motion) with 16 cues per trial.

Results and conclusions: The results, validated using single-sample Wilcoxon signed-rank tests, revealed that the unsaturated state, with a median score of 3.44, exhibited significantly lower average inter-eye contact delay (p-value = 0.014, r = 0.86) and significantly higher average number of eye contact (p-value = 0.014, r = 0.87) compared to the hypothesized medians. These findings suggest that the unsaturated state occurring in the critical temporal window has significantly higher average number of eye contact and lower average inter-eye contact delay compared to the saturated state. This study underscores the importance of tailored and targeted interventions for children with autism spectrum disorder, facilitating the development of cost-effective, time-efficient, evidence-based treatments.