Your Photos Say Where You Are

Researchers from the United Kingdom have created a way to predict where a person might travel next based on pictures posted to the photo-sharing website Flickr. The team published their results Tuesday in the journal Royal Society Open Science.

Scientists led by Daniele Barchiesi, a data scientist formerly at the University of College London and now at Capgemini, used computer algorithms to cluster 8 million geotagged photos taken by 16,000 people in the U.K. by location. Using that information, they could map the probability that the photographer would travel from one location to another. For example, a person currently in London is most likely to go next to Edinburgh, the researchers found.

Seem a little creepy? The potential for using this information for cyber-stalking isn’t huge, Barchiesi says, though it is an issue worth consideration.

“If somebody wants to track you, there are better ways of doing that,” agrees Peter Gloor, a MIT professor who was not involved in the research. That may be small comfort to many concerned about the shrinking realm of privacy in the Internet age.

But there’s another point of view to consider: Understanding how groups of people move around a city, a country, or the globe could have more potential for good than evil. Stopping disease outbreaks would be much easier if researchers know at the outset that someone infected at point A is more likely to travel to point B than to point C. And understanding commuters’ favorite routes through a city at rush hour could help the urban planners of the future design less congested traffic patterns.

Being able to make these kinds of global and local predictions from online photos will take time. For one thing, the eye of Flickr is not all-seeing. Not everyone uses Flickr all the time: People post only the photos that they want other people to see online, and most days they’re not going to be snapping photos of hospitals or their commute.

Similar studies have tracked peoples’ movements with geotagged Twitter posts, mobile phones, and money. MIT professor Marta Gonzalez even assessed how appealing cities are to tourists using geotagged photos.

But Gonzalez says she hasn’t seen the kinds of methods Barchiesi used applied to Flickr photos before, and Gloor adds that the real “coolness of this paper” lies in its novelty: “Pictures that you didn’t think could be used for this can be used to predict movements of people,” Gloor says.

And a variety of geotagged photos are readily available. Unlike Facebook, which is a holy grail for data scientists because its relative privacy entices users to share more intimate information about their daily lives, Flickr’s publicly shared photos are easily accessible to researchers. “You can go and actually grab the data and analyze it,” Barchiesi says.

“Sometimes,” he says, “The work in data sciences is about getting the data.”

© National Geographic | Rachel A. Becker