Perspective-taking with robots: experiments and models

Abstract

We suggest that to enable effective human-robot interaction, robots should be able to interact in a way that is natural to and preferred by humans. Using human-compatible representations and reasoning mechanisms should help in developing skills which support effective human-robot interaction. In this paper, we present two studies that examine a critical human-robot-interaction component: perspective-taking. We find that when a person asks a robot to perform a task with some ambiguity to the robot, the person prefers the robot to either ask for clarification or take the person's perspective and act appropriately.