Do We Trust Embodied Agents who Look Like us?

With the rise of automation and the metaverse, there's a notable increase in the deployment of embodied agents (EAs). These embodied agents uniquely interact with human counterparts as team members rather than just acting as tools. A limited number of research studies have attempted to examine how such agents in a computer-mediated environment within virtual reality (VR) environment. Previous research has determined attributing anthropomorphism is one of the most common designs associated with human-agent collaboration that significantly impacts the level of trust and acceptance of embodied agents. While acceptance of agent deployment is strongly determined by the level of trust humans place in the agents in a computer-mediated environment, it is critical to also understand how these human users remit trust toward embodied agents. Humans tend to trust others that portray similar physical elements to themselves. Our study therefore examined whether a human would exhibit higher trusting behavior in an embodied agent with a virtual depiction designed to have similar features. To validate this hypothesis, participants engaged in a coin entrustment game with agents varying in resemblance to human facial features. Our results demonstrate that there is a strong relationship between similarity and trust in the VR context, which suggests that users are more likely to trust EAs that have similar facial features as themselves visually. In addition, these findings offer important implications for future embodied agent design. Lastly, we identify some research gaps and propose future research directions beyond the current state of the field.