Amplifying the Good, Muting the Bad: A Social Amplifier Model of Embodied Role Adaptation in Human-Robot Interaction

As robots increasingly collaborate with humans in dynamic environments, the tension between communication clarity and social appropriateness presents an important design challenge. While communicative clarity is traditionally assumed to enhance human-robot interaction, recent findings suggest that clear communication can backfire when robots violate social norms. This study introduces and validates a “social amplifier” model of embodied communication, proposing that the value of role clarity depends on the appropriateness of the robot’s social behavior. We developed a novel confidence-aware role adaptation paradigm where a NAO robot dynamically switches between socially appropriate and inappropriate roles based on users’ confidence levels in an ambiguous video discrimination task. Using a within-subjects design with 64 participants, we manipulated role complementarity, choice alignment, and embodied communication across 512 total interactions. Results revealed that embodied communication significantly enhanced role clarity but had conditional effects on collaborative outcomes. Contrary to the “bad is stronger than good” principle, embodied communication asymmetrically amplified positive effects of appropriate robot behavior while having minimal impact on inappropriate behavior. This pattern emerged consistently across subjective measures (collaboration quality, role dynamics, trust) and objective behavioral outcomes (confidence change, choice change). The findings also demonstrate that robots with socially appropriate roles became over seven times more persuasive when enhanced with rich embodied cues, while inappropriate robots showed no benefit. These results provide evidence-based design principles prioritizing social intelligence over mere expressiveness in collaborative decision-making robots.