A trait-based persuasive framework for preference elicitation by social assistive robot

Human preferences are often ill-defined and shaped by contextual and situational factors, posing challenges for effective human–robot interaction. Social Assistive Robots can address this by employing adaptive, context-aware persuasive strategies to better elicit user preferences and support healthier behaviors. In this study, we investigated how a Pepper robot, equipped with a Large Language Model-driven dialog system and a trait-based persuasion model, could estimate user profiles and tailor strategies for preference elicitation and persuasion. In particular, the robot was deployed in a simulated home environment to offer snack options and encourage participants toward healthier choices. System performance was assessed using the Persuasive Robot Acceptance Model, an extension of the Technology Acceptance Model that incorporates social dimensions such as trust, liking, compliance, and psychological reactance. This work is presented as a system-integration contribution, focusing on evaluating the perceptual impact and operational robustness of the integrated framework in a realistic HRI setting, rather than isolating the causal contribution of individual personalization components. Results revealed statistically significant differences (p<0.05, p<0.01) in user impressions across gender, technological experience, and motivation groups. However, no significant differences were observed with respect to trait-based profiles (trust propensity, compliance awareness), suggesting either effective adaptivity of the system or a stronger influence of demographic factors. These findings underscore the importance of adaptive, user-sensitive design in social assistive robots to foster equitable, engaging, and effective persuasion across diverse user groups.