Avatars with faces of real people: A construction method for scientific experiments in virtual reality

Abstract

Experimental psychology research typically employs methods that greatly simplify the real-world conditions within which cognition occurs. This approach has been successful for isolating cognitive processes, but cannot adequately capture how perception operates in complex environments. In turn, real-world environments rarely afford the access and control required for rigorous scientific experimentation. In recent years, technology has advanced to provide a solution to these problems, through the development of affordable high-capability virtual reality (VR) equipment. The application of VR is now increasing rapidly in psychology, but the realism of its avatars, and the extent to which they visually represent real people, is captured poorly in current VR experiments. Here, we demonstrate a user-friendly method for creating photo-realistic avatars of real people and provide a series of studies to demonstrate their psychological characteristics. We show that avatar faces of familiar people are recognised with high accuracy (Study 1), replicate the familiarity advantage typically observed in real-world face matching (Study 2), and show that these avatars produce a similarity-space that corresponds closely with real photographs of the same faces (Study 3). These studies open the way to conducting psychological experiments on visual perception and social cognition with increased realism in VR.

Face identification in the laboratory and in virtual worlds.

“…Forty avatars paired with high-quality digital photographs served as stimuli for this experiment (20 matches and 20 mismatches). Each avatar was constructed using a 3D scan of a real person’s head that was acquired using a state-of-the-art 3D scanner, which was subsequently rigged onto a pre-made body and animated for movement in VR (see Fysh et al, 2021). For each avatar, a high-resolution screenshot was acquired, which was subsequently paired alongside a high-quality digital photograph of the scanned subject’s real-life counterpart (i.e., a match trial) or a different person who was matched for gender and approximate age and broadly similar in appearance (i.e., a mismatch trial).…”

“…And while some professionals excel at face matching (Towler et al, 2019;White et al, 2015), screening for such professionals using conventional laboratory methods is difficult (Bate et al, 2018;Fysh et al, 2020). This difficulty is compounded further by studies showing that realworld factors, such as passenger volume and time pressure, influence face-matching accuracy in the laboratory (Bindemann et al, 2016;Fysh & Bindemann, 2017;Wirth & Carbon, 2017; for a review of factors, see Fysh and Bindemann, 2021). However, while researchers can use laboratory data to estimate how these factors influence performance in practice, unfamiliar face matching has not been studied in the critical real-world context of passport control at airports, due to the security-sensitive nature of these environments.…”

“…We have recently developed a VR airport-based method to study face identification at passport control (Tummon et al, 2019), and demonstrated the potential of this approach to investigate contextual factors in ways that extend laboratory research in this domain (Tummon et al, 2020). Alongside this innovation, we have developed a methodology to construct avatars with photo-realistic faces for psychological experimentation in VR (Fysh et al, 2021). Here, we combine these approaches to study the identification of realistic person avatars at the passport control checkpoint of a VR airport.…”

Feeling close to a Crab-Thing in virtual reality: Does avatar appearance always matter in forming meaningful connections? A case study

“…This suggests that the use of more cartoonish avatars may influence interpersonal assessments in a more positive direction generally and may have had a role in producing the effects reported here (but see also Volante et al, 2016, which concluded that interactions with more cartoonish humans in VR may suppress the formation of emotional bonds). New methods to improve realism of avatar faces (including the integration of actual facial photos of users; Fysh et al, 2021) are under development and will necessitate confirming the current findings using effectively different human appearing avatars than those available to VR users today. Hyper-realistic human appearing avatars may well generate greater levels of affiliative outcomes than observed here.…”

Digital body, identity and privacy in social virtual reality: A systematic review

“…This study sheds light on whether people can recognize identity misappropriation when encountering imposter avatars. Fysh et al (2021) provided a user-friendly photorealistic avatar generation method for the psychological research community, and demonstrated a series of studies exploring the identification of the avatar faces with respect to the correspondence of the avatars with their real-life counterparts. Their research provides off-the-shelf workflows and research methodologies for future psychological research on digital bodies and identity.…”