Punishable AI: Examining Users' Attitude Towards Robot Punishment

To give robots, which are black box systems for most users, feedback we have to implement interaction paradigms that users understand and accept, for example reward and punishment. In this paper we present the first HRI experience prototype which implements gradual destructive interaction, namely breaking a robot's leg as a punishment technique. We conducted an exploratory experiment (N=20) to investigate participants' behavior during the execution of three punishment techniques. Using a structured analysis of videos and interviews, we provide in-depth insights into participants' attitude towards these techniques. Participants preferred more abstract techniques and felt uncomfortable during human-like punishment interaction. Based on our findings, we raise questions how human-like technologies should be designed. A video documentation of the project can be found here: https://vimeo.com/348646727