Can a robot bribe a human? The measurement of the negative side of reciprocity in human robot interaction

Abstract

Reciprocity is a cornerstone of human relationships and apparently it also appears in human-robot interaction independently of the context. It is expected that reciprocity will play a principal role in HRI in the future. The negative side of reciprocal phenomena has not been entirely explored in human-robot interaction. For instance, a reciprocal act such as bribery between Humans and robots is a very novel area. In this paper, we try to evaluate the questions: Can a robot bribe a human? To what extent is a robot bribing a human affect his/her reciprocal response? We performed an experiment using the Rock, Paper, Scissors game (RPSG). The robot bribes the participant by losing intentionally in certain rounds to obtain his/her favour later, and through using direct and indirect speech in certain rounds. The participants could obtain between 20%- 25% more money when the robot bribed them than in the control condition. The robot also used either direct or indirect speech requesting a favour in a second task. Our results show that the bribing robot received significantly less reciprocation than in the control condition regardless of whether the request was couched in direct or indirect speech. However there is a significant interaction effect between the bribe and speech conditions. Moreover, just three of sixty participants reported the robot-bribe in an interview as a malfunction, though they did not mention any moral judgement about its behaviour. Further, just 10% of the participants reported the bribe in the online questionnaire. We consider that our experiment makes an early contribution to continue the exploration of morally ambiguous and controversial reciprocal situations in HRI. Robot designers should consider the reciprocal human response towards robots in different contexts including bribery scenarios. Additionally our study could be used in guidelines for robot behavioural design to model future HRI interactions in terms of moral decisions.