Despite deaths from cell phones and other tech distractions, a huge percentage of the carnage on America’s roads is still due to drinking and driving. Out of the 37,461 people killed on the roads in 2016, 10,497 people—nearly 30 percent—died because they were, or someone else was, drunk. An innovative way is emerging to cut that death toll, and it highlights the burgeoning issues of regulating autonomous vehicle (AV) technology.
Government, academia, and the private sector have long collaborated on a technological breakthrough to reduce drunk-driving deaths. For a decade, the National Highway Traffic Safety Administration and the auto industry’s Automotive Coalition for Traffic Safety (ACTS) have partnered to support the commercial research and development of something called DADSS—Driver Alcohol Detection System for Safety.
“It began with a simple question,” says Robert Strassburger, head of ACTS. “What if we could limit driver blood-alcohol contents to less than the 0.08 legal limit? . . . We have made a lot of progress over the past 30 years”—in the early 1980s, drunk-driving deaths were half the nation’s car-related total, not a third—“but as a percentage, I think that number is pretty significant.” Drunk drivers are a stubborn population: “Probably one-eighth of 1 percent of drivers in this country [produce] 30 percent of the deaths,” Strassburger notes. For these people, it’s hard to reduce that number with traditional traffic-safety programs like education and deterrence.
DADSS would reduce drunk driving through automated, passive data collection and interpretation. The technology works in one of two noninvasive ways. A would-be driver gets into a car and puts his finger on a sensor. Within a fraction of a second, a refracted signal “tells us the alcohol concentration through your finger,” says Bud Zaouk, program manager.
The other sensor is breath-based, based on technology that can differentiate between a driver and a passenger and identify the alcohol concentration in carbon-dioxide exhalation. A computer can use the information provided by the sensors to disable a vehicle’s movement capability (you could still turn the vehicle on for warmth or communication, and the computer wouldn’t record the data for posterity).
The technology evolved in part fortuitously. Inlight Solutions, a company doing blood-glucose work to study how to test diabetics through noninvasive procedures, noticed that one test subject at the University of New Mexico always yielded skewed results, and “it turned out he was showing up to the test drunk,” says Strassburger. The technology has improved markedly over the 15 years. The DADSS venture does much of its research in Massachusetts, drawing on the state’s rich academic community. Says Zaouk, “we have chemists, we have medical doctors, engineers, statisticians.”
Researchers have shrunk the equipment from large boxes to something that can fit unobtrusively into a car. Most important is improved reliability, as a false result risks inconveniencing someone stranded somewhere with a car that won’t start. “We felt . . . that we could not in any way hassle a sober driver. There would be zero tolerance for any inconvenience,” says Strassburger. Something that began as an “audacious concept” could be commercially available in five years.
The impaired-detection technology may be straightforward—but how people can, will, and should use it in the real world is not. One question is how to allow the technology to interact with AV capabilities. For example, should a semiautonomous car (in which a driver would have to take over only in certain circumstances) transport an obliterated “driver” if the car does not expect such circumstances to arise along the journey—for example, along a road that the car has traveled many times before?
Another question addresses who is in charge of driving decisions: the technology, drivers, or regulators? For instance, should alcohol-detection technology allow an inebriated person to override the system? In a society that claims to have zero tolerance for any drunk driving, this seems a bizarre point, but there are circumstances in which a generally responsible driver might prefer an override function. Does the public-health benefit of keeping drunks off the road outweigh an individual’s imperative, in any condition, to flee, say, a would-be rapist?
Answering this question raises others. If a driver knowingly uses an override function, does he or she incur more culpability in a crash, for knowingly—and not just possibly—driving drunk? Does the car or device manufacturer incur any liability if an innocent person in another car or on the street is injured or killed?
A final question arises, and it, too, will apply to more advanced AV technology: Should drunk-driving detection and disabling be optional or mandatory? The technology’s champions are presenting it as optional, noting that three potential markets for AV cars are parents with teen or twentysomething children of driving age, heavy drinkers who are concerned about the possibility of drinking and driving, and people who may eventually be able to save money on their insurance by installing the technology. (Another, not-quite-optional market: drivers with past drunk-driving convictions who might have to install the technology as part of a court order.)
Yet the history of safety-critical features is that they eventually become mandatory. After all, seat belts and air bags were once optional. The same people who, as children, rode lying down in their parents’ station wagons would never dream of leaving the driveway before ensuring that their kids are belted. In two decades, it may seem reckless to leave a party three hours after having your last drink without knowing for a fact that you are sober.
These are cultural, political, and regulatory decisions, not tech ones. How much personal freedom are Americans willing to give up for safety, not just to benefit themselves, but to help other people on the road?