Texting while driving is about as self-evidently stupid as watching TV while running a chainsaw. Everyone knows that, but millions of motorists type all the time while behind the wheel, regardless. In a survey released in December by the National Highway Traffic Safety Administration, 18 percent of all drivers — and nearly half of those ages 18 to 24 — admitted to sending texts or emails while behind the wheel. The agency estimates that texting while driving has increased by 50 percent in the past year. That’s despite the fact that 35 states have outlawed the practice and that texting has been implicated as a factor in countless fatal accidents.
Mike Watkins wants to do something about it. In October 2009, Watkins was in the kitchen of his Richland, Washington, home, idly answering emails. He was half-listening to a TV news segment about the dangers of distracted driving, but his ears perked up when he heard Transportation Secretary Ray LaHood call texting while driving a “deadly epidemic.”
“My two daughters were teenagers then, and I was very aware of their texting habits, because I pay the bills,” Watkins says. And as a bicycle commuter who rides eight miles each way to his job at the Energy Department’s Pacific Northwest National Laboratory, he is nerve-rackingly aware of how many drivers pay more attention to the gadgets in their hands than to what’s in front of them on the road.
So Watkins, who heads the lab’s applied physics research program, talked his employers into giving him some funding to find a way to help. Two years later, he and his colleagues have just published their first peer-reviewed paper, in IEEE Spectrum, on a solution: a computer algorithm that enables a cell phone to recognize when a driver is sending a text message.
The March-April 2012
This article appears in our March-April 2012 issue under the title "OMG UR Phone Knows UR Txting + Driving!" To see a schedule of when more articles from this issue will appear on Miller-McCune.com, please visit the
March-April magazine page.
Watkins started by having six test subjects send text messages on a specifically outfitted phone while standing still, and then while using driving simulators. A program on the phone captured the speed and rhythm of their keystrokes. The data showed that the subjects entered letters more slowly while driving than while still. This in itself was not surprising, but the key finding was that the difference was fairly consistent: texting drivers typed about 7 percent more slowly than standers, with the speed measured in milliseconds. Moreover, the ways those keystrokes bunched and clumped also changed according to discernible patterns.
Watkins and his team then devised an algorithm that could be programmed into phones to detect those changes. “You gather a baseline from each person of what their texting is like normally and then look for variations,” he explains. It proved more than 90 percent accurate with their experimental group, he says. Paired with a GPS to spot when a phone is moving faster than a certain speed, indicating it’s in a car, Watkins believes the algorithm would accurately identify texting drivers 99 percent of the time or more. “What you do with that information is another question,” Watkins says. The phone could send the user a warning, shut itself off, or send a tattling message to the driver’s employer or parent.
A fistful of applications and gadgets that perform a similar trick are already on the market. Watkins’s approach, though, offers a more finely tuned way of sorting out who is a driver and who is a passenger. That’s an issue not only for those getting a lift in a car, but for people riding buses and trains. Most existing applications use GPS alone to determine when the phone is in a moving vehicle, then lock it up and offer clunky systems for users to reactivate their phones — by solving timed puzzles, for instance, or getting permission via text from an employer or parent. And of course, none of these apps work in the millions of phones that don’t have a built-in GPS. Watkins’ algorithm can detect a texting driver all by itself.
At least, it can in the lab. But if it proves out in the real world, Watkins’s algorithm could have all kinds of other uses. “The idea of assessing someone’s cognitive state while they’re interacting with a machine or other device has many interesting applications,” he says. “You could potentially use it to detect fatigue in heavy-equipment operators or to detect an anomaly in grandma’s cognitive functions due to a stroke.” More ominously, it could also be used by corporations to track when a computer-using employee’s attention starts to wander.
Watkins is now hoping to attract a partner interested in commercializing the technology. “This solution might be ahead of where the market is right now,” he acknowledges, “but people are literally dying every day. Someone needs to address the problem.”