Abstract The growing presence and capabilities of robots and artificial intelligence (AI) make it increasingly likely that artificial entities might in the future be involved in situations requiring moral judgements. We conducted a preregistered study with two online samples (N 1 = 162, N 2 = 247) to investigate to what degree people are willing to let artificial agents decide in life-or-death as well as in non-lethal everyday moral dilemmas. All participants responded to eight moral scenarios (four high-stakes and four low-stakes). In life-or-death scenarios we found a clear preference for human decision-makers while in everyday low-stakes dilemmas participants’ responses indicated that either human or artificial entities might decide. Two main predictors emerged for agent preference: acceptance of AI and participants’ own moral choices in the respective scenarios. While greater acceptance of AI was in both high- and low-stakes scenarios associated with a greater willingness to let artificial agents decide, the relationship between agent preferences and participants’ own moral decisions differed between the two dilemma types. In high-stakes scenarios, a less deontological resp. more utilitarian response was associated with greater willingness to let AI or robots decide while in low-stakes dilemmas this association was reversed. Additionally, greater familiarity with AI predicted less willingness to accept artificial decision-makers, but only in high-stakes dilemmas.
Armbruster et al. (Mon,) studied this question.