It turns out all those Sunday brunch photos on Instagram aren't just pictures, they're data. Or, at least they were for Georgia Tech researchers looking for a better way to get at the reality of living in a so-called food desert, where access to grocery stores and healthy, affordable food is scant, reports the Atlantic. The most common method of assessing the diets of food-desert residents—occasional surveys of "populations of limited size"—is hamstrung by its inability to really get at what people are eating day in and day out. Enter Munmun De Choudhury and her team, who scouted out 3 million geo-tagged Instagram photos of food to determine what people are eating. What they found: In what the USDA has identified as food deserts, the foods photographed contained 5% to 17% more fat, sugar, and cholesterol than those photos from demographically similar non-food deserts.
The difference was stark enough that researchers could actually predict whether a given Instagram post originated in a food desert 80% of the time, they report at the 19th ACM Conference on Computer-Supported Cooperative Work and Social Computing. "Fruits and vegetables are the biggest difference," De Choudhury says in a release, with 33% of food desert posts mentioning them, compared to 48% of non-food desert posts. (Grist floats one solution here, pointing to the success of farmers markets in food deserts, particularly if vendors accept food stamps.) There was one finding that ran counter to "popular intuition": There was no real difference in terms of the number of calories contained in the photographed foods. The study does have limitations, though: Photos aren't necessarily representative of typical meals, and a picture cannot reveal the quantity a person will then go on to eat. (Of course, "healthy" foods may not be healthy for everyone.)