Autonomous systems in ethical dilemmas: Attitudes toward randomization

It is ethically debatable whether autonomous systems should be programmed to actively impose harm on some to avoid greater harm for others. Surveys on ethical dilemmas in self-driving cars' programming have shown that people favor imposing harm on some people to save others from suffering and are consequently willing to sacrifice smaller groups to save larger ones in unavoidable accident situations. This is, if people are forced to directly impose harm. Contrary to humans, autonomous systems feature a salient deontological alternative for immediate decisions: the ability to randomize decisions over dilemmatic outcomes. To be applicable in democracies, randomization must correspond to people's moral intuition. In three studies (N = 935), we present empirical evidence that many people prefer to randomize between dilemmatic outcomes due to moral considerations. We find these preferences in hypothetical and incentivized decision-making situations. We also find that preferences are robust in different contexts and persist across Germany, with its Kantian cultural tradition, and the US, with its utilitarian cultural tradition.