Using Human-Robot Interaction to assess emotion recognition in dialysis patients: A preliminary study

This cross-sectional observational study examined the relationship between psychological factors and multimodal emotion recognition abilities in patients with chronic kidney disease (CKD), introducing a humanoid robot as an innovative assessment modality. Sixty adults undergoing dialysis and awaiting renal transplantation (63.3 % male; mean age = 53.95 years, range 20-80) completed three emotion recognition tasks based on Ekman facial expressions, dynamic video clips, and emotional displays performed by the NAO humanoid robot, together with validated measures of alexithymia (TAS-20), anxiety (GAD-7), and depression (PHQ-9). Regression analyses showed that performance on the Ekman task was positively predicted by education and anxiety levels (beta =.26, p = .037; beta =.26, p = .043). Recognition of emotions in videos was negatively predicted by age (beta = -.45, p < .001) and alexithymia (beta = -.37, p = .002). For the robot-based task, increasing age (beta = -.30, p = .020) and longer dialysis duration (beta =.31, p = .017) significantly predicted recognition accuracy. Logistic regression analyses indicated that alexithymia negatively predicted recognition of disgust in videos and fear expressed by the robot. Levels of anxiety positively predicted recognition of happiness and fear in the Ekman test. Across the modalities, alexithymia and age emerged as core factors associated with reduced emotion recognition, with video and robot-based tasks capturing these differences more clearly than static facial stimuli. These findings suggest that socially assistive robots may represent a promising tool for detecting subtle socio-emotional processing alterations in medically vulnerable populations, supporting future longitudinal research integrating technological and psychological assessment.