Money, Banks and Finance in Economic Thought

What can economics tell us about ethics? Lessons from the Moral Machine Experiment

Salvat Christophe, CNRS, Centre Gilles-Gaston Granger, AMU

The article “The Moral Machine Experiment” published by Nature on 24th October has drawn a lot of attention amongst scientist and philosophers alike. This article is exceptional in more than one way. First, it is published in Nature, whilst dealing on ethics, which is almost unheard of. Second, it gives an account of the biggest experiment undertaken on moral preferences so far and claims to solve one of the most controversial ethical issue raised by Artificial Intelligence: the self-driving car dilemma, or in this particular case, a full range of self-driving car dilemmas. The objective of the survey is to provide policymakers and car manufacturers with socially acceptable moral rules. MME has recorded over 40 million decisions in ten languages, and in 233 countries on a dedicated platform online provided by the MIT. Sessions were each composed of 13 scenarios: one entirely chosen at random and the other 12 sampled from a set of approximatively 26 million possibilities. In each session two dilemmas specifically focused on the importance of gender, age, physical fitness, and social status in the decision-making process. At first sight, these dilemmas are not fundamentally different from those already discussed in the Trolley problem originally presented by Philippa Foot in 1967 and the numerous versions it has produced since. In both cases, we are presented with a (very unlikely) fictional case in which we are to choose amongst two options, each one involving at least the death of one person. In both cases, each questionnaire is accompanied by a diagram, showing the spatial layout of the forthcoming tragedy. But there stops the similarity between these cases. Whilst the Moral Machine Experiment (MME) attempts to build (consequentialist) moral rules out of our intuitions, the Trolley Problem was initially conceived as an illustration of the limits of consequentialism. More importantly, perhaps, MME deeply undermines moral realism, which Philippa Foot, amongst many others, tried to defend. Many points, however, seem to indicate that for their authors, like a great numbers of their readers, these experiments belong to the same category. Most scholars have ceased to see these thought experiments as means to challenge moral theories but prefer seeing them as a way to solve them. This raises, however, a number of theoretical and methodological issues that I propose to discuss. This paper argues, in particular, that moral rules are no more reducible to social preferences than they are to biological laws, and that ethical issues cannot reasonably been treated independently of each other, and maybe more importantly, outside of a sound ethical and meta-ethical framework.


Keywords: experiments, psychology, ethics, consequentialism, self-driving cars