Photo: Eneas @

Researchers the world over are finding social media to be the data-mining gold mine of the century. Twitter in particular has been shown to be more accurate and faster in crisis situations than many official reports of the same situation. Now researchers at the University of Rochester are turning that data fountain to the task of analyzing the relative health of the Twitter citizens that inhabit it.

Like most Twitter data mining apps, this one relies on language analysis to determine if a person is actually saying that they’re sick or just “sick of” something. The results are sketchy at best, but a promising start. For example, training in on our area, one person tracked says “I feel like I’m gonna be sick,” but is that because she is actually sick? Or because she hates her mother’s nail polish? Impossible to say based solely on a single tweet.

Screenshot from GermTracker website, showing red, yellow and green dots where tweets that may be relative to sickness happened.

What is genuinely interesting and new about this particular app is the contextual data that they’re gathering. By analyzing a Twitter user’s proximity to factors such as pollution (using GeoCoded tweets, which reveal GPS data about the user’s location), researchers can infer some things about their lifestyles and correlate that to tweet data. For example, on the issue of whether exercising makes you healthier, there is this tantalizing clue:

For example, even people who regularly go to the gym get sick marginally more often than less active individuals. However, people who merely talk about going to the gym, but actually never go (verified based on their GPS), get sick significantly more often. This shows that there are interesting confounding factors that can now be studied at scale.

To be clear: this is scientifically extracted data, not scientific fact. Whether the above clue means that people who go to the gym really do get more sick than those who do not, or whether people who go to the gym complain about or discuss illness more often than those who do not is an open question, an answer for which would require a lot more research.

Still, researcher Adam Sadilek says that there are many practical uses of the new analytic software, from plumbing the social science facts to rerouting those concerned with influenza outbreaks away from subway tunnels when subway riders are reporting being sicker. Mysophobia? Yes. There’s an app for that.