What if Twitter knew you’d get sick before the first sneeze? New research demonstrates that it’s possible. Researchers at the Department of Energy’s Pacific Northwest National Laboratory has shown that, simply by analyzing the emotions behind tweets, they can predict outbreaks of flu about 14% to 35% of the time. Add in the actual content of those tweets (like “I just can’t get out of bed today!”), and researchers say that figure skyrockets to 95%.
This phenomenon has been described as a “digital heartbeat” by many researchers, including Svitlana Volkova, who led the research. Much like a doctor can hold a stethoscope to your body to hear what’s going on inside, so too can the properly trained piece of software read through the lines of a tweet to identify hidden illness.
Overall, military personnel seemed unhappier, in sickness and in health. “Civilians express more positive and less negative emotions, along with less sadness, fear, and disgust,” says Volkova. But both civilians and the military were affected by flu in largely the same way. “During [flu outbreaks] there’s more sadness and neutral emotion. If there’s no flu, people are angrier and express more positive sentiment.” In other words, when sick, our emotional expressions are muted, and sad. When we’re well, we fire on all extremes of the emotional spectrum.
As wildly absurd as that may sound, a lot of researchers are proving it’s already possible. Volkova herself has been giving talks on the matter at Google, Microsoft, and Facebook. Her research has shown that one can predict social media discourse during a major international crisis, along with the way fake news responds to major events. She’s even working on a study to see if they can predict weather based upon how social media talks about it.
Of course, whether this research is actually predicting the future, or whether it simply predicts the cadence of conversation with which society talks about the future, becomes a bit of a philosophical wormhole. But Volkova is more than aware of the soft ground on which this research is standing, and hopes to solidify it more with further study.
“We can’t claim we can predict a future terrorist attack,” says Volkova. “But we can say, given the discourse of social media in the past, we can see what’s important on social media in the future. And that’s likely to correlate in the real world.”