Facing the artificial: Understanding affinity, trustworthiness, and preference for more realistic digital humans

Abstract

In recent years, companies have been developing more realistic looking human faces for digital, virtual agents controlled by artificial intelligence (AI). But how do users feel about interacting with such virtual agents? We used a controlled lab experiment to examine users' perceived trustworthiness, affinity, and preference towards a real human travel agent appearing via video (i.e., Skype) as well as in the form of a very human-realistic avatar; half of the participants were (deceptively) told the avatar was a virtual agent controlled by AI while the other half were told the avatar was controlled by the same human travel agent. Results show that participants rated the video human agent more trustworthy, had more affinity for him, and preferred him to both avatar versions. Users who believed the avatar was a virtual agent controlled by AI reported the same level of affinity, trustworthiness, and preferences towards the agent as those who believed it was controlled by a human. Thus, use of a realistic digital avatar lowered affinity, trustworthiness, and preferences, but how the avatar was controlled (by human or machine) had no effect. The conclusion is that improved visual fidelity alone makes a significant positive difference and that users are not averse to advanced AI simulating human presence, some may even be anticipating such an advanced technology.