beatrossmy / punishableai

PunishableAI

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.

You can find a project video on vimeo.

The project was published at DIS'20:

Beat Rossmy, Sarah Theres Völkel, Elias Naphausen, Patricia Kimm, Alexander Wiethoff, and Andreas Muxel. 2020. Punishable AI: Examining Users' Attitude Towards Robot Punishment. Proceedings of the 2020 ACM Designing Interactive Systems Conference. Association for Computing Machinery, New York, NY, USA, 179–191. DOI:https://doi.org/10.1145/3357236.3395542