Exploring AI-Based Reciprocal Eye Engagement Training to Enhance Dyadic Behavior in Individuals With Alexithymia

The significance of socio-emotional difficulties in various neurodivergent and mental health conditions has led to an investigation into the co-occurrence of alexithymia in autism spectrum disorder (ASD), depression, anxiety, and schizophrenia. The study aimed to improve reciprocal eye engagement in individuals with Alexithymia, and hence enhance dyadic behavior and mutual understanding. This study introduced an innovative individual-paced progressive training involved 4 levels from cartoon faces, still facial images, self-videos, to trainer-led reciprocal eye engagement video sessions. The computer vision based real-time feedback on gaze dwell time, eye movement, blink rates, and reciprocal eye engagement achieved higher spatial and temporal resolution. 10 subjects, ages ranging from 7 to 10, were recruited who have clinical assessment of ASD with level 1 support and were also assessed with alexithymia using Children's alexithymia Measure (CAM). 6 subjects made statistically significant improvement over the self-paced training period between 4-8 weeks with p-value<5% for both gaze dwell time and number of reciprocal eye engagement, 3 subjects showed slight improvements, and 1 subject showed no improvement. Notably, this real-time AI-based annotation on screen reinforced the association between eye engagement and positive social interactions which is missing in alexithymia individuals. Furthermore, the trainer-led synchronous training, level 4 specifically, enhanced dyadic behavior by emulating real-life interactions, led to increased engagement, confidence, and skill retention. The life video cameras transcended the physical boundaries. This research also underscored its potential applications in human-computer interactions and future direction on emotion recognition to promote neurodivergent relationships and inclusive society.