A Tool but not a Peer: How Framing Affects People’s Perceptions of AI Agents in Teams

In this paper, we set out to explore how people judge the personality of a non-anthropomorphic virtual agent during group interactions. Using a Wizard of Oz (WoZ) based approach, we observed that people judged the acceptability of a virtual agent’s behavior from a tool-based lens, that is, if this robot and its behavior are useful to the team or not. Furthermore, we found that while people were able to acknowledge the virtual agent’s personality and recognize its identity through social cues, the tool-based framing impacts these perceptions into a normative judgment of the robots’ utility. We present two case studies that we think highlight this tool-based interpretation of robotic personality: robots expressing non-factual opinions and robots expressing humor. Finally, we suggest that researchers should consider the impact of this tool-based framing on people’s perceptions of a robot’s identity when designing robots for social interaction.