The human eye plays a crucial role in shaping social judgments; for instance, maintaining eye contact enhances social interactions and influences how others perceive us. At a more subtle level, the pupils also play an important role; recent research shows that pupil mimicry fosters trust. In our digitizing world, people frequently interact with virtual agents. Therefore, it is of crucial importance to investigate whether the positive effects of eye gaze and pupil mimicry generalize to virtual interactions. In the current study, we investigated whether eye contact and pupil dilation mimicry influence the evaluation of virtual humans and whether these effects vary with the participants’ levels of social anxiety. Six virtual humans told a story which participants watched on a monitor screen. Three of them were programmed to mimic the participants’ pupil size at moments when participants displayed natural pupil dilation, while the other three did not mimic these instances and instead, their pupil sizes dilated at random moments. After each story, we assessed how the virtual human was perceived by taking multiple social evaluation measures. The results show that the more the participants made eye contact, the more positively they evaluated the virtual human, especially during trials where the virtual humans mimicked the spontaneous instances of pupil dilation in participants’ eyes. Social anxiety level did not influence the results. This study demonstrates that the cumulative positive effects of eye contact and pupil dilation mimicry generalize to interactions with virtual agents, which may inspire potential applications.

During social interactions, people often unintentionally mimic each other's behaviors, which can occur at the motor level (e.g. smiles or head movements), and the autonomic level (e.g. pupil size). Mimicry has often been linked with increases in liking and trust and tends to smoothen the bond between interaction partners. However, people with social anxiety are generally less expressive and they may also have altered mimicry behavior. The current study aims to explore the relationship between mimicry, person-perception, and social anxiety levels by having participants interact with virtual humans. Compared to real interaction partners or photographed or filmed actors, virtual humans can be controlled and behaviors very precisely manipulated. Participants observed ten virtual humans telling a neutral story on a computer monitor. At predetermined time points, the virtual humans showed four types of expressions: smiles, head movements, pupil dilation, and blushing. After each story, we measured to what extent participants liked and trusted the virtual humans via the avatar evaluation questionnaire, the desire for future interaction scale, and an investment game. In addition, we assessed participants’ social anxiety levels. The results show that participants, regardless of anxiety levels, mimicked the virtual humans’ smiles and that smile mimicry is associated with an increased liking and trust toward the virtual humans. These results build on previous research showing that the mimicry of smiles acts as a social glue and also benefits virtual social interactions.