Utilitarian or Relational? Exploring Indicators of User Orientation Towards Intelligent Agents

Abstract

When interacting with an agent, some users with utilitarian orientation tend to treat an agent as an instrumental tool, while others with relational orientation find the design of humanlike features more pleasing. Along with technological advances in user modeling and prediction algorithms, intelligent agents nowadays can personalize their interaction by identifying such orientation of users towards them. While prior work has revealed several behavioral signs resulting from such difference in orientation, little attention is directed to more fundamental cues that precede the occurrence of actual interaction. In light of this issue, this study explores intrinsic properties of users related to their utilitarian or relational orientation towards intelligent agents. Qualitative analysis of responses revealed three user propensities contributing to individual differences in orientation: tolerance to unpredictability, sensitivity to privacy, and sensitivity to an agent’s autonomy. We discuss future directions for leveraging our findings to design personalized interaction in intelligent agents.