As embodied conversational agents (ECAs) reach unprecedented visual photorealism, aligning their communicative behaviors with human social norms becomes a critical challenge for interaction design. This laboratory study N=40 investigates how agent behaviors—specifically gaze patterns, interruption-handling strategies, and expressed personas—interact with user personality traits to shape interaction quality. Utilizing a high-fidelity MetaHuman agent within a Wizard-of-Oz framework, we conducted a multi-modal analysis integrating subjective measures, qualitative interviews, and facial Action Unit (AU) analysis. Our findings reveal that user personality traits, particularly Agreeableness and Conscientiousness, significantly moderate social expectation: while high-agreeableness users exhibit greater "social forgiveness" toward norm violations, high-conscientiousness users prioritize interactional efficiency and are highly sensitive to conversational overlaps. Crucially, we identify a phenomenon of "hidden social friction," where high-fidelity agents compel users to perform "emotional labor" manifesting as masking smiles (AU09 + AU12) during disruptive interruptions. We argue that intelligent agent design should move beyond superficial anthropomorphism toward a "normative alignment" that accounts for individual psychological differences. This work contributes to the CSCW community by providing a nuanced understanding of interactional costs associated with digital human photorealism and offering heuristics for developing more socially resilient and personality-aware autonomous systems.
Ye et al. (Thu,) studied this question.