The Case of the Curious Robot : On the Social Viability of Curious Behavior in Non-Human Agents

Abstract

Curiosity is increasingly being looked at to develop interactive nonhuman agents capable of lifelong learning, particularly in the space of Ai driven agents. However, the way a non-human agent’s mental state is described could prime expectations in human counterparts as well as alter perceptions of an agent’s behavior. We conducted an online experiment examining people’s perceptions of four levels of curious behavior portrayed by a robot agent and whether this behavior matched people’s expectations when primed to anticipate a “curious” vs. “learning” vs. “autonomous” robot agent. As curiosity increased, the ratings of the robot agent’s thinking ability also increased. This was juxtaposed by a decrease in the robot’s ratings as an effective worker and social agent. Further, we captured a unique trajectory of expectation matching showing that agents’ internal mental states should be matched with analogous external behaviors. Our findings have implications for the design and development of non-human agents that engage in learning modeled on forms of human cognition.