You’ve likely seen it on the highway — a car that’s tailgating, darting between lanes, passing in the shoulder. You, a (presumably) human person, understand what that means: An is behind the wheel. As autonomous vehicle development continues, those cars will eventually need to learn how to respond to that same situation. But, how do you teach road rage to a robot?

Researchers at the University of Warwick over in England are trying. tries to quantify road rage, and shape it into something that autonomous vehicles can understand — pure, empirical data.

Our human understanding of road rage relies on something computers don’t have: Empathy. We can see a car swerve, jerk its way from lane to lane, and intuitively understand this as aggressive behavior — we know what angry humans look like, how they act. But computers lack empathy. They have no means of intuiting a person’s mood. They, instead, need hard facts to work from.

That’s what the Warwick study attempts to build: A factual basis for road rage. By angering drivers, then putting them behind the wheel of a simulator, researchers aimed to identify patterns in reckless and aggressive driving. If those behaviors can be accounted for, then autonomous vehicle companies can develop ways of avoiding them.

Of course, AV development is far from really needing to focus on edge cases like aggressive drivers. Ideally, the rest of us out on the roads would like autonomous vehicles to master things like “left turns across traffic” before paying attention to such niche situations. But hey, maybe you really do need to run before you can walk.