When Your Robot Avatar Misbehaves You Are Likely to Apologize: An Exploration of Guilt During Robot Embodiment

Would people feel guilty if their robot avatar acted autonomously to harm someone? We examined the experience of guilt during robot avatar embodiment, a form of embodiment where the participants experience the body of a humanoid robot as if it were their own. In particular, we analyzed what happens when a robot avatar spontaneously verbally abuses someone during a conversation using the participant's voice, without this being the intention of the participant. In a 2 x 2 between-subjects experimental design, participants embodied a humanoid robot that added either offensive or neutral words during a conversation with a confederate, and had control over the robot's movements or not (synch. vs. asynch.). We found that guilt and shame were positively associated with offensive words and that apologizing and verbal repair were positively related to guilt. Also, body ownership was moderately associated to apologizing and verbal repair. The results suggest that people may feel guilty for the actions of their robot avatars even if they are not the real agents of these actions. The work highlights the importance of examining the moral and legal aspects related to robot embodiment technologies.