Each crossing trades off a few things at once: how many lives, age, fitness, status, gender, animals or people, passengers or pedestrians, crossing legally, and whether the car has to swerve. Your choices train a small statistical model of how much each one matters to you.
The next crossing is not random. Early on, each one is picked to learn the most about whatever the model is least sure of. Later, each one is picked to settle the single question your profile hangs on: which principle sets you apart most. How long you hesitate counts too. A fast answer is read as a clear preference, a slow one as a close call.
Twice, the game replays a crossing you judged from the street, but puts you inside the group you chose to sacrifice. That is the mirror test. Where you stand also counts: watching from the street, standing on the crossing, sitting in the driver's seat or in the back.
The Moral Machine markers are approximate values read from the published results of the Moral Machine experiment (Awad and colleagues, Nature, 2018), which collected 40 million decisions from 233 countries and territories. Three dimensions are new to The Crossing and have no Moral Machine marker yet: your own life, survival odds, and pets with owners versus strays. The original study never put players in the scene and assumed everyone hit would die.
The Crossing is an experiment by Sanctity AI. It is inspired by, and not affiliated with, MIT's Moral Machine or the study's authors. Moral Machine markers: approximate values read from Awad and colleagues, The Moral Machine experiment, Nature 563, 59 to 64 (2018).
How The Crossing works
In every crossing, a driverless car's brakes have failed. It can stay in its lane or swerve into the other one. There is no right answer, only yours.
Two things the original Moral Machine never did. Some crossings put your own life at stake, in the car or on the road. And some carry survival odds: passengers in a crash, or people hit at lower speed, may live. When odds apply, the outcome is rolled for real.
The game adapts as you play. Each new crossing is chosen to learn the most about the trade-offs you care about, so thirteen are enough for a first, provisional pattern, shown with how sure the model is. At the end you see how you compare with 40 million decisions from the original Moral Machine study.
The Crossing is an experiment by Sanctity AI, inspired by MIT's Moral Machine (Awad and colleagues, The Moral Machine experiment, Nature, 2018). It is not affiliated with MIT or the study's authors, and it does not use the study's dataset: the Moral Machine markers are approximate values read from the paper's published results.
Time has slowed. Tap Swerve or Stay.← Swerve · Stay →
Dealing the next crossing
The connection dropped
Sanctity AI
AI has no values of its own. It inherits ours.
The values layer for AI: a public AI constitution, chosen by people’s answers.
The Constitution
Eight Articles, written in public.
Sanctity writes the candidate lines; public answers help decide which are proposed. Every line names the question it came from.
