When robots should look away: how gaze direction and anthropomorphism shape psychological responses to human-robot interaction in embarrassing service encounters

Compared with human providers, service robots can alleviate consumer discomfort in embarrassing service encounters, while their social cues shape responses. As robots’ facial design becomes increasingly sophisticated, gaze direction is a social cue in human-robot interaction (HRI), yet its role in embarrassing encounters across levels of anthropomorphism remains unclear. These effects matter for socially adaptive robot design. Grounded in the Computers Are Social Actors (CASA) paradigm, this research examined how robot gaze direction (direct vs. averted) and anthropomorphism (high vs. low) influence consumer continuance intention in embarrassing service encounters. Three between-subjects experiments were conducted ( N = 1,200). Study 1 examined this interaction across embarrassing and non-embarrassing contexts, while Studies 2 and 3 investigated the mediating roles of perceived warmth and competence; Study 3 additionally examined neuroticism as a potential moderator. Data were analyzed using factorial ANOVA and conditional process analysis. Across all three studies, the effect of gaze direction on continuance intention depended on robot anthropomorphism in embarrassing service encounters. Under high anthropomorphism, averted gaze led to higher continuance intention than direct gaze, whereas the opposite pattern emerged under low anthropomorphism. Study 1 further revealed a significant three-way interaction involving service context, with the gaze direction × anthropomorphism interaction reaching significance in embarrassing but not non-embarrassing service encounters. Studies 2 and 3 identified two distinct psychological pathways in embarrassing service encounters: averted gaze enhanced continuance intention by reducing perceived warmth under high anthropomorphism, whereas direct gaze did so by increasing perceived competence under low anthropomorphism. In Study 3, neuroticism did not significantly moderate the interaction between gaze direction and anthropomorphism in embarrassing service encounters. This research demonstrates that the interaction between robot gaze direction and anthropomorphism varies across embarrassing and non-embarrassing contexts, refining understanding of how robot social cues shape consumer responses. It extends the warmth and competence framework by revealing distinct perception pathways across anthropomorphism levels and showing that perceived warmth may inhibit continuance intention in embarrassing consumption contexts. These findings offer practical guidance for configuring robot gaze direction according to anthropomorphism in embarrassing service encounters.