A University of Rochester researcher’s work suggests that though an individual may opt out of location tracking through social media use and call records, their movement is still predictable with data from family, friends, and even strangers.
According to Ghoshal’s findings, companies and governments can get a solid picture of a person’s next move by using the data of “non-social co-locators”—people in different social circles who visit similar locations. Individuals in the same social circle carry more information about each other’s mobility than individual non-social co-locators, but multiple co-locating strangers taken together can “provide similar levels of predictability,” the paper states. These co-locators may be people working in the same office, living in the same apartment building, or attending the same school, for example.
“We show that if I have no access to your own information, but I have access to some of your friends’ information, let’s say five or six of your best friends who hang around with you quite often, then I can pretty much predict where you’re going to be next with 95 percent certainty despite me having no information on you at all,” Ghoshal says. “But we also went a step further and said, ‘What about people who hang around you that you don’t know?’
“Surprisingly, it turns out that it doesn’t do as well as your friends, but about 80 to 85 percent of your own behavior can be predicted just by looking at strangers who hang around you a lot. … People who sort of live in the same area tend to have similar patterns.”
The percentages Ghoshal provides are theoretical upper limits for how confidently a perfect algorithm could predict a person’s mobility. No real-world algorithm is perfect, but even existing imperfect algorithms can make high-confidence predictions using behavioral information ingrained in the people around us.
“There are all these articles that have come out in recent times being like, ‘Your location is being tracked, Google is watching you, etc., etc., and here’s how to turn location tracking off,’” Ghoshal says. “The whole idea behind that is that I can keep my information from these companies and then they won’t have any information and I’m all good to go. What we show in this paper is that’s not true at all because, while your behavior is predictable, it’s also embedded in your environment.”
He says this poses big questions about privacy and potential future business models for web, software, and hardware services.
“People, with the internet and web economy, have just gotten used to getting things for free, but nothing in life is free. So, if you want those services, the price you have to pay is giving up your data to these companies,” Ghoshal says. “To be fair to them, harvesting this data is how they improve their products, but they don’t just stop at improving their products. They also sell this information to other companies.”
Indeed, the data trade is booming. Data overtook oil to become the most valuable commodity on the planet five years ago. Estimating exact figures is difficult, but the revenue of the global data economy was ballparked at $3 trillion in 2017 by the World Economic Forum. The United Nations also noted in 2019 that data firms held a disproportionately large share of market capitalization as early as 2016, with the 25 largest tech companies (mostly data firms) representing nearly 20 percent of market capitalization in the United States at $6 trillion.
Still, there are productive uses for the information being collected. The paper notes that mobility prediction using strangers’ data could aid in future pandemic prevention, for example. Ghoshal adds that data collection and prediction has greatly increased the quality of companies’ products.
“The quality of services, without doubt, has improved by an order of magnitude thanks to this. Where would people be without Google Maps? The flipside is that you’re not getting these things for free,” Ghoshal explains. “So, if this is something that bothers you and you want to continue using this service, then you should think about other alternatives like a subscription plan. … This is a function of the economic model of the web.”
But the findings do not just raise questions related to the private sector. It also means there is no clear way to protect oneself from government surveillance, Ghoshal observes.
“There are also authoritarian impulses on the increase globally. This might not be just about how to sell you the best yogurt, but it might be to track you and lock you up,” he says. “You can go off the grid, but if I find your friends, then I can probably pick up on you as well.”
Questions in this realm pop up more frequently when the national security implications are considered. Ghoshal has conducted research funded by the Army Research Office about how the U.S. Armed Forces can apply the information, but these applications can come with heavy trade-offs, if used unethically.
“There is a national security aspect. If you want to look at people of interest, you might not have access to their information, but you can track down those threats ahead of time by just looking at their social networks,” Ghoshal says. “But, again, the flipside of that is, depending on the government, it can be used for other purposes.”
Justin O’Connor is a Rochester Beacon intern. The Beacon welcomes comments from readers who adhere to our comment policy including use of their full, real name.