PROBLEM: There has been a lot of interest lately in ways we can use Internet behavior to monitor public health. Facebook likes seem to be a good tool for this -- researchers have already figured out how to use interests expressed on the site to make strong inferences about users' race, gender, age, political affiliation, and sexual orientation. They can even indicate mortality rates for hospitals.
METHODOLOGY: Researchers at Harvard Medical School took behavior either positively or negatively linked to obesity -- namely, being active or being sedentary -- and then looked at the proportion of adults who liked related things on Facebook. Relevant pages were broadly categorized underneath "health and fitness" and "outdoor physical activities" for being physically active and "television" for being sedentary. They organized the data by city for all of the U.S. and, in a more focused analysis, by zip code for New York City, and then compared what they found to public data on the prevalence of obesity in those areas.