In this world, no one is obeying the law, or lane markings, and everyone, from distracted drivers to jaywalking pedestrians, is forcing you to make split-second, life-or-death decisions—on purpose.

Subjecting autonomous vehicles to such a world, say engineers, is precisely how manufacturers and regulators will know that they are ready to be handed responsibility for our very lives.

To understand why, it helps to know a little about how the artificial-intelligence algorithms at the heart of self-driving systems are trained to handle tricky situations. Across almost every manufacturer of such systems, much of this training happens in simulation. That is a far safer and cheaper option than gathering data in the real world from actual vehicles, says Henry Liu, a professor of engineering at the University of Michigan and director of Mcity, a facility for testing autonomous vehicles.