Smartphones Knowing When You’re Drunk Is the Next Frontier of Data Spying

In case your smartphone tracking your travels, web browsing, dating history, job changes, diet changes, shopping habits, and general content consumption wasn’t enough for you, there’s good news. Soon, the little snitch in all of our pockets might be able to tell us when we’re plastered just by measuring a few of our unsteady steps.

A group of researchers from Stanford and the University of Pittsburgh just published a preliminary study in the Journal of Studies on Alcohol and Drugs that used a small group of participants to show alcohol intoxication could likely be detected by the accelerometer that comes standard in most modern smartphones. In previous research, the team found that 50 percent of test subjects don’t realize they’re functions are impaired when they’re drunk, and the hope is that this latest project could lead to an early detection application.

The study was relatively simple: 22 volunteers were given an individually adjusted dose of vodka formulated to achieve a blood-alcohol concentration of 0.2%—a threshold that exceeds the .08% BAC level for the legal limit to operate a motor vehicle. A smartphone equipped with a motion-sensing 3-axis accelerometer was strapped to the volunteers’ lower backs and an app called phyphox was used to monitor the accelerometer data. The participants first walked in a 10-step pattern back and forth while sober for control data. Over the next 7 hours, the volunteers walked the same pattern each hour and had a breathalyzer test administered to monitor their intoxication.

The researchers found that the accelerometer was able to identify intoxication by the changes in the users’ gait with a statistical accuracy of around 90 percent.

While the researchers find the results to be promising, they emphasized in the paper that plenty of more tests would be necessary. In addition to the small sample size and narrow demographics of the participants, the scientists identified the consistent placement of the smartphone on the subject’s body as a weakness and they hope to test more a more natural placement in future studies.

The research also builds on at least three similar studies conducted over the last eight years that used different conditions on a group of volunteers ranging in size from three to 34 participants. Those studies also found strong indications (56-89% accuracy) that an accelerometer can identify impaired walking when a user is drunk.

On the bright side, the scientists state that their work “could offer opportunities for triggering just-in-time interventions aimed at improving prevention and treatment of alcohol use disorders.” But lead researcher Brian Suffoletto acknowledged in an interview with New Scientist that the accelerometer information could be tempting for third-parties in the business of data collection. “If someone wanted to go through the effort to process and analyse it, they could probably make inferences about changes in walking patterns,” Suffoletto said. But he said it would be “a leap” to assume a user is intoxicated just based on the way they walk.

It might be a leap from a legal point of view. No one should expect authorities to start sending bills for drunk driving in the mail after an accelerometer picks up on their calamitous walk through the parking lot—at least not any time soon.

But advertisers and hackers love to make inferences about data. What’s the harm in inferring a user might be drunk when algorithmically deciding whether to serve up some ads for online gambling or porn? Maybe the user is just feeling lucky or horny, anyway. Maybe an advertiser will score really big and get a drunk target to take out a second mortgage on their house. And government-sponsored hackers would surely love to gather data on the alcohol consumption of all kinds of public and private targets.

While these researchers are refining the accuracy of this kind of smartphone monitoring, there are surely bad actors out there who are already taking it for a test drive.