Applying Game Theory to Driverless Cars

There is a new driverless shuttle in Las Vegas.

Mayor Carolyn Goodman was among the first public officials to hop a ride on the vehicle developed by the French company Navya.

The company has shuttles in use in other countries, and one is being tested at a University of Michigan research site.

It uses GPS and electronic sensors that company spokesman Martin Higgins says will stop it if a person or dog runs in front.

That last sentence is what interests us here. These vehicles will stop if someone or something steps out in front of them. Clearly liability laws will require that they stop 100 percent of the time. We can also assume that, once the tech is perfected, with their never-blinking robotic eyes, sonar, radar and robotic reflexes there will be few if any events in which pedestrians get hit.

Down at Iowa City around the University, pedestrians have the right-of-way. When traffic is slow, college students take advantage of this and step right in front of your car. They cross against the light. They hardly look. They know you will stop. That they know for certain you will stop is important. The result is if you are driving the traffic is snarled and you barely move through downtown area.

On nearby streets were traffic is moving more quickly, they are much more cautious. They still have the right of way and they still know that you don’t want to hit them but they can’t be sure that you won’t make a mistake and unintentionally hit them.

When the risk is near zero, the benefit of crossing against the light outweighs the risk and students will walk right out into traffic. When the risk is much above zero, however, they are much more circumspect and usually wait for the WALK light.

So what happens in a world of self-driving robotic supercars reduce that chances of being hit and run over to near zero everywhere?

I submit that pedestrians everywhere will begin acting like students in Iowa City do when the traffic is creeping. Everywhere people will exploit the fact that self-driving cars have reduced risk to near zero step right out into the street. Since there is no risk, there is no benefit to waiting. This behavior would, however, ensure that traffic is always snarled and that cars are barely able to move.

It seems to me this is almost a certainty and represents a risk to the whole project.

Thoughts?