Empathizing with virtual agents: the effect of personification and general empathic tendencies

Abstract

For interactions to be natural, virtual agents should understand humans’ emotions, and humans should have emotional reactions towards them. In human-to-human interaction, this is achieved through empathic processes between individuals. So, improving empathic responses towards virtual agents represents a crucial step in improving human-virtual agent interactions. This study aims to identify whether the presence of a personification story and individual differences in the ability to empathize predict the empathic response towards a virtual agent. Furthermore, it investigates the effect of previous experience with virtual agents and gender on empathy towards the virtual agent. In an experiment, participants witnessed a virtual reality scene in which a virtual agent experienced sadness. Half of the participants were previously presented with a personification story about the virtual agent, and all completed a self-report questionnaire about empathy and a post-experiment survey about their empathic response towards the virtual agent. Results showed that individual differences in empathy significantly predict the ability to empathize with the virtual agent: people who are naturally predisposed to feel more empathy towards others tend to be more empathic towards the virtual agent. The personification story, previous experience and participants’ gender did not affect the empathic response. Implications and future direction for the design of virtual agents are discussed.