Helping robots imitate

Abstract

Towards informing the design of more ‘natural’ (non-expert) human–robot interaction (HRI) systems, this chapter presents two studies, focusing on a scenario where a human has to teach a robot how to perform an everyday household task.The first study considered the humans’ acknowledgment of feedback from a robot, to which they had to demonstrate how to do a task. The research questions were if (and how) the participants would modify their teaching instructions when the robot stated a misunderstanding, and whether they would remain consistent for the remainder of the task.In the second study, the robot feedback increased in complexity, in order to elaborate on what the human had demonstrated. The research question was if the participants, based on the robot feedback, would be able to resolve any misunderstandings by modifying their teaching instructions.Besides illustrating that people express a wide variety of teaching behaviors when faced with the task of demonstrating to a robot, the results from the second study support our design suggestion that robots should actively and accurately advertise their cognitive and manipulative capabilities via form (appearance) and function (behavior) to improve the dialogic interaction.