How do people respond to computer-generated versus human faces? A systematic review and meta-analyses

Computer-generated (CG) beings are rapidly infiltrating the human social world. Yet evidence about how humans respond to CG faces is mixed. The present systematic review and meta-analyses aimed to synthesise empirical evidence from studies comparing people's responses to CG and human faces, across key face processing domains of interest to psychology, neuroscience, and computer science. We tested whether effects were moderated by the perceived realism of CG relative to human faces, and whether CG and human faces showed the same identity or not. We hypothesised that people would be able to tell CG and human faces apart, and that other types of responses would favour human over CG faces. While results supported our hypotheses across several domains (perceptions of human-likeness, face memory, first impressions, emotion labelling), some responses did not differ for CG and human faces (quality of interactions, emotion ratings, facial mimicry, looking behaviour). We also found a reduced inversion effect for CG relative to human faces, though only minimal data were available for hallmark face effects (ORE, N170 and FFA responses). Overall, findings highlight potential strengths and challenges of using CG faces across a range of applications, including e-health, social companionship, video-gaming, and scientific work.