Different level of virtualization of sight and touch produces the uncanny valley of avatar’s hand embodiment

Abstract

Humans increasingly often act through virtual and robotic avatars, which can feed back to their user only virtual sensory information. Since avatar is user’s embodiment and body image is mostly based on senses, how virtualization of sensory inputs affects avatar self-attribution is a key question for understanding nowadays human behavior. By manipulating visual and tactile inputs in a series of experiments fashioned after the rubber hand illusion, we assessed the relative weight of the virtualization of sight (Real, Robotic, Virtual) and of touch (Real, Virtual) on artificial hand embodiment. Virtualization decreased embodiment, but unexpectedly lowest embodiment was found when only one sense was virtual. Discordant levels of virtualization of sight and touch elicited revulsion, extending the concept of the uncanny valley to avatar embodiment. Besides timing, spatial constraints and realism of feedback, a matched degree of virtualization of seen and felt stimuli is a further constraint in building the representation of the body.