After a bit of imbibing, people are more likely to overshare. But if you tweet while drinking, you’re revealing even more than whatever drunken ramblings you’re posting to social media. A team of researchers from the University of Rochester have figured out a way to analyze tweets to track people’s drinking habits, and though this info isn’t being used to personally prove that you’re a lush, it can provide broad insights into when and where people are consuming alcohol.
The paper, published earlier this month, has an intimidating tittle – Inferring Fine-grained Details on User Activities and Home Location from Social Media: Detecting Drinking-While-Tweeting Patterns in Communities – but its goals were far simpler. Researchers wanted to find out if they could determine two things from Twitter: Was the tweeter drinking and, if so, was he drinking at home or somewhere else?
Answering these two straightforward questions was actually quite complex. According to MIT Technology Review, it took about 11,000 geotagged tweets along with analytical help from workers on Amazon’s Mechanical Turk crowdsourcing service to train a machine-learning algorithm to identify which tweets were posted by someone who was actually drinking at the time. Equally difficult was taking users geotagged tweets and determining where they live, honing in on specific words, frequency and times that indicate where the Twitter user probably calls home.
However, the team claims they were able to determine both pieces of information accurately enough that they were able to compare drinking habits in New York City to those in Monroe Country, NY, and create “heat maps” showing where people go out drinking in both cities. For the record, people drink more in NYC – not that you needed to be a scientist to figure that one out.
Though the research can obviously help reveal plenty about alcohol consumption, the authors also speak to the ability to use Twitter as a research tool in general. “While our experiments are focused on alcohol use, our methods for locating homes and distinguishing temporally-specific self-reports are applicable to a broad range of behaviors and latent states,” the paper states.
So next time you’re posting on Twitter while drinking, be careful. Not just because you don’t want to say anything dumb, but because scientists might be watching you too.