Abstract We examine people’s attitudes toward AI-based algorithms as a means to allocate scarce resources. Through a vignette experiment, we confront respondents with five scenarios in which an AI-based algorithm allocates various goods and ask them if they find this morally desirable. We compare people’s moral attitudes toward AI with their attitudes toward a friend, a waiting list, a lottery, and the market. Our results show that people rank allocations through AI as morally clearly less desirable than most alternatives. This is especially true for goods that are nonessential to a person’s survival. One potential explanation for the identified algorithm aversion is that AI is considered more opaque than its alternatives, and that an allocation mechanism’s moral rejection increases if its working is less well understood. Together, our results suggest that using AI to allocate resources, especially nonessential ones, is likely to meet substantial social resistances. To understand the reasons, researchers should systematically study laypeople’s moral intuitions about algorithms, a field we call folk algorithmics.
Michael et al. (Thu,) studied this question.