Imagine you're driving down a busy street when the car in front of you slams on its brakes out of nowhere. Your brain has to process what's happening, decide what to do, and move your foot to the brake — all in the span of a heartbeat. Scientists have long tried to build computer models that could predict how real humans would react in these split-second moments. Now, a team from Delft University of Technology in the Netherlands and Waymo, the self-driving car company, says they've done exactly that.
On June 10, 2026, they published their new artificial intelligence model in the journal Nature Communications. What makes their approach different is that it combines three things — seeing the danger, deciding what to do, and actually doing it — into one unified system. Older models could only handle one piece at a time, like measuring reaction speed or steering angle, but not both together.
To test whether their model thought like an actual person, the researchers compared it against human behavior in three tricky situations: a car ahead braking suddenly, an oncoming vehicle swerving into your lane, and a car failing to give you the right of way. The model was given exactly the same information a human driver would have seen. The results matched up — the fake driver reacted in about the same time as a real one and made similar choices between hitting the brakes and steering away.
"Our new model brings all these components together," said Arkady Zgonnikov, an assistant professor at Delft University of Technology who worked on the project.
The team says this isn't just an interesting science experiment — it could actually change how self-driving cars are tested and approved. Regulators around the world have long struggled to answer one simple but important question: are autonomous vehicles actually safer than human drivers? This model could give them a scientific yardstick to measure against.
"It can help address whether autonomous vehicles are safer than human drivers, a key question in regulation," Zgonnikov said. "At the same time, it becomes possible to formulate clear and measurable requirements for manufacturers."
Mauricio Peña, Waymo's chief safety officer, said the model could help the entire industry "move toward a shared, scientifically grounded approach to assessing collision avoidance." And Waymo is already putting it to use, comparing how its own autonomous vehicles perform against the predicted behavior of human drivers.
The hope is that better models of human driving will lead to safer roads for everyone — whether they're behind the wheel or in the back seat.