What the research reveals: measuring distraction

Our central finding can be stated plainly: Based on the combined evidence from our studies and the wider scientific literature, the AZT estimates that distraction may contribute to one in three road crashes. It is not always the sole cause, and it rarely leaves a trace in the official record — but it is frequently the factor that turns an ordinary moment behind the wheel into a critical one, and a critical one into a crash. The three findings below set out the evidence.

The starting point was the German In-Depth Accident Study database (GIDAS). Working with Traffic Accident Research at TU Dresden (VUFO), the AZT examined more than 34,000 injury crashes recorded since 1999, comparing cases with an indication of distraction against those without. The gap between what is recorded and what is visible in the detail is stark.

If distraction were only about the driver, the same task would carry the same risk in every car. It does not. In a real-vehicle study, the AZT had 20 drivers per car run identical everyday tasks — entering a destination, choosing a song — in two current electric compacts, on a country road, while an eye-tracker measured how long their eyes left the road.

Overall, total glance time was up to 25 percent lower with the more efficient interface — a difference of roughly three seconds. That difference did not come from a better or faster driver; it came from the interface. Consistent menus across functions, controls placed where the hand expects them, a system that responds without lag, familiar navigation and sensibly sized touch areas all shortened the time drivers spent looking away. In an emergency, three seconds is not a detail. It can be the distance between staying in control and a collision.

How a driver interacts matters as much as where the controls sit. In a driving-simulator study with the Technical University of Munich, the AZT compared basic touch interactions and found they are not equally demanding. Typing pulls the eyes away the most; drag-and-drop takes the longest to complete; a simple swipe or a pinch-to-zoom is far less costly. The lesson is not that touch is unsafe, but that different touch interactions impose very different levels of visual demand. Complex interactions should therefore be avoided or restricted while driving. Every strand of the work points the same way: the problem lies in poor interaction design, not in touchscreens themselves.

To turn glance behavior into risk, the AZT worked for the first time with the modeling firm Cognibit, using a cognitive driver model calibrated with real gaze data from the vehicle study. The model simulates how drivers perceive, decide and react, and it ran four critical situations for attentive, distracted and intoxicated drivers (the last at 0.8 per mille blood alcohol). The scenarios were steady driving on a rural road with oncoming traffic, a sudden emergency braking ahead on the motorway, a forced lane change past a stationary obstacle and a child running onto the road in town.

When the car ahead braked hard on the motorway, the share of runs ending in a collision rose from 18.2 percent for attentive drivers to 54.3 percent for distracted ones. Safety margins collapsed with it: the minimum time-to-collision fell from 1.61 to 0.74 seconds, the minimum gap from 6.9 to 3.1 meters. When a child ran into the road, the collision share climbed from 3.9 to 11.2 percent — and in 39 percent of those distracted collisions the driver did not react to the child at all.

The rural road produced the starkest comparison. Distracted drivers drifted across the center line and into oncoming traffic in 5.4 percent of runs. Attentive drivers never did — and neither did the drivers modeled at over the drink-drive limit. In this situation, distraction proved more dangerous than alcohol. Across the four modeled situations, the distracted driver came off worse than the one over the drink-drive limit in all four of them. The forced lane change told the same story in a quieter register: distracted drivers failed the maneuver more often or cut their margins finer, with more of them dropping below a 1.5-second time-to-collision.