The Moral Machine

What the experiment measured

MIT's Moral Machine presented users with scenarios: an autonomous vehicle with failed brakes must swerve left or right. Each side had a different group of people who would be struck. Groups varied by number of people, age, gender, social role (doctor, criminal, athlete), and whether they were crossing legally. Participants chose which outcome they preferred.

The experiment was not asking what the law requires. It was not asking what a philosopher recommends. It was asking what millions of ordinary people across many countries actually prefer, which is a different question and a politically significant one when car manufacturers must embed some decision procedure into their vehicles.

The cultural variation

Global results showed broad agreement on some preferences: spare more people over fewer, spare children over the elderly, spare humans over animals. But the differences were substantial enough to complicate any single global standard.

Researchers identified three broad cultural clusters. Western nations showed stronger preferences for sparing younger people and for prioritizing pedestrians following traffic rules. Eastern nations showed relatively more weight on sparing passengers. Latin American nations showed stronger preferences for sparing women and for social status.

An autonomous vehicle programmed with values appropriate for Germany may be making choices that would be considered wrong in Japan or Brazil. A globally deployed technology encoding a single moral framework is, in effect, exporting one culture's ethical intuitions as universal defaults.

Should vehicles calculate at all

The harder question the experiment sidesteps is whether autonomous vehicles should be making utilitarian calculations in the first place. The rule-based alternative says vehicles should follow fixed procedures: always brake as hard as possible, never swerve onto a sidewalk, and accept whatever outcomes follow. The rules are simple, predictable, and publicly known.

The utilitarian alternative says vehicles should minimize harm by calculating outcomes. But this requires deciding whose outcomes count, by how much, and by whose values. It turns every vehicle into an ethics engine and every manufacturer into a moral legislator.

The Moral Machine experiment revealed what people prefer when forced to choose. It is less clear that being forced to choose is the right framing.