Congruency-Dependent Gaze Allocation to Avatar Facial Expressions during Shared Video Viewing in Immersive VR

Abstract

In shared immersive media, expressive avatars can provide social context, emotional feedback, or companionship. However, poorly aligned avatar reactions may also draw attention away from the primary content. We investigate this in a VR eye-tracking pilot in which participants watched 30-second emotional video clips alongside a co-present MetaHuman avatar expressing either matching or mismatching emotions. Across 28 participants, mismatching avatar reactions were associated with higher avatar dwell time than matching reactions, longer individual glances, and stronger early inter-subject synchrony of avatar-directed gaze. These observations suggest that avatar reaction mismatch is a promising factor in VR attention design, with implications for virtual cinemas, remote events, training simulations, and avatar-mediated communication. The study used a fixed presentation order, with matching trials preceding mismatching trials. This limitation is addressed in the discussion and motivates future counterbalanced work.