Des Moines Register: UI Sees High-tech Car Devices Improving Road Safety

Monday, November 15, 2010

By Jens Manuel Krogstad
Des Moines Register

Standard safety features on vehicles may one day include a system that detects drunken or drowsy drivers or even early symptoms of Alzheimer's disease, a University of Iowa researcher predicts.

Measurements of driving performance could trigger vehicles to tell a driver to pull over, warn others through car-to-car wireless communication, alert police or even shut off the car's engine, said Timothy Brown, a lead investigator who works at the university's National Advanced Driving Simulator research unit.

The technology raises increasingly common legal issues associated with artificial intelligence guiding and controlling people's lives, a national privacy official said.

University researchers developed formulas that determine alcohol impairment based on driving performance with up to 86 percent accuracy. The best results exceed accuracy rates of standardized field sobriety tests that law officers now use.

The system detected behaviors a police officer would look for: variations in speed, lane departures and large steering corrections, Brown said. Accuracy would increase as the computer learns a person's normal driving behavior based on previous trips, he said.

"There's a lot of potential down the road with what you could do with this sort of technology," he said. "Although the concepts can be somewhat abstract, the ultimate goal is to have a positive effect on real people's lives and keep people alive."

The proportion of fatal accidents that involve drunken drivers has stalled at around 31 percent in recent years, researchers noted in their full report to the National Highway Traffic Safety Administration. Federal grants paid for the research.

In Iowa, the percentage of alcohol-related deaths has plateaued at between 20 and 25 percent of all fatal crashes. Last year, 93 people died in alcohol-related accidents.

Brown said the research could one day push the boundaries at which technology, public policy and public acceptance intersect.

That day may not be too far off, he said. Vehicles sold today warn drowsy drivers to pull over and help keep drivers in their lanes or from following other vehicles too closely.

"We're at the point where we're able to detect things, but we're not at the point of being able to determine what's the best thing to do once you detect," he said.

Transparency of the formulas will be essential as the technology moves forward, to safeguard due process and privacy, said Jay Stanley, who specializes in privacy for the American Civil Liberties Union. This would ensure that a defendant could, in effect, cross-examine the computer that has accused him of a crime, Stanley said.

Keeping the programming public also safeguards against abuses from governments, corporations and other large institutions, he said.

"When you have artificial intelligence increasingly imbedded in our lives and cars, questions arise about exactly how they're programmed, whose interests they serve, what kind of laws they incorporate into the programming," Stanley said. "These are a lot of the 21st century questions that arise when you put computers in the role of regulating human behavior."

Brown, the university researcher, said a colleague in the medical field imagined the technology one day alerting a spouse that the car's driver may be in the early stages of Alzheimer's disease, based on subtle deterioration of driving performance over time. The idea did not sound far-fetched, he said.

"Driving is a complex physical and cognitive task, and a lot of diseases associated with aging have an effect on that," he said.