By Ben Schiller

To you and me, Twitter is just another means of communication. To Adam Sadilek, it is an enormous sensing network to predict when people might get the flu, or experience food poisoning.

A researcher at the University of Rochester, Sadilek recently analyzed the content and location of a month's tweets in New York City. He was looking at factors such as how often people go to the gym or take the subway, and how that might affect their health. Not surprisingly, people who visited venues with high concentrations of people were more likely to get the flu. And people living in polluted areas were more at risk than those who living elsewhere.

Sadilek says the mapping is potentially useful for policy-makers who want an instant, real-time snap-shot from the ground. It is a way of surveying people passively, and without spending lots of money to do so.

"The key is that now we can measure these factors in real-time and at scale. For instance, it is intuitively understood that pollution is bad. But how bad exactly, for who, when, and under which conditions is poorly understood. Methods like ours enables us to study the important details."

Working with Professor Henry Kautz, Sadilek has developed an app called Germtracker that people can use to avoid locations where they are more likely to get sick. He says the predictive-accuracy of the app will increase as people use social media more. Kautz and Sadilek plan to track food-related illnesses bad products, unsanitary restaurants, and mental health (like depression and suicide) soon.

The collaborators claim their algorithm is sensitive enough to distinguish between a range of moods, and the different ways we communicate. Sadilek says it understands "that someone who tweets 'I'm sick and have been in bed all day' should be characterized as sick, but 'I'm sick of driving around in this traffic'" shouldn't be.

"The news about many important health hazards often reach the general population when it's too late for people to act," Sadilek says. "Our app is an additional channel they can use to get real-time insights."

Comments