The dissemination of immersive virtual reality (IVR) with body tracking offers new opportunities for communication across distant places by embodying virtual avatars or actual robots. Thus, when people meet via IVR, different forms of embodiment for humans and artificial intelligence become possible, which may influence the perception of the self, the encoding of information and the evaluation of communication partners. In this research-in-progress paper, we draw upon research on self-presence and social identity theory to investigate how robotic embodiment influences intergroup bias in the context of a user who sees a group discussion of individuals embodied in avatars with different degrees of visual and mental humanness. We hypothesise that less humanness leads to higher misattribution of debaters‘ contributions and more negative evaluations of these debaters. Additionally, we assume that this effect is diminished when the users who watch the group discussion are embodied in an avatar with low humanness.