Article Thumbnail

How Social Media Can Predict the Next Big STD Outbreak

If you’ve ever tweeted about your itchy crabs, you’re now part of the solution

Just when you thought oversharing on social media would be the downfall of modern society, two studies out of UCLA find a silver lining. It turns out, people’s social media and internet tendencies might help pinpoint and prevent the spread of sexually transmitted diseases.

The studies, led by Sean Young (not that Sean Young), founder and director of the UCLA Center for Digital Behavior and the UC Institute for Prediction Technology, “found an association between certain risk-related terms that Google and Twitter users researched or tweeted about and subsequent syphilis trends that were reported to the CDC.” According to Young, the additional data on when and where outbreaks are happening, provided via social media, can be a significant factor in prevention.

As someone who constantly complains on Twitter (and, let’s face it, is one breakup away from haphazardly Googling where to find free and immediate sex), I asked Young what exactly people were searching, and where this research is heading.

So… why did you start researching people who tweet about having syphilis?
Our research began when we built these closed online communities for people with HIV or other diseases, and quickly learned that people would share all kinds of personal information in there. When we started looking through the data of what they were talking about, we thought, Let’s look to public sources of data like Twitter and see if people are sharing there, publicly.

Eventually the CDC provided their weekly syphilis data, and when we matched that to our data, we came up with a list of types of keywords that people might search for, or tweet about, that would suggest they could be at risk for syphilis in the future with pretty good accuracy.

You say you identified 25 keywords, including “find sex” and “STD.” What were the rest?
In accordance with the Internal Review Board, we can’t release the full list — as you can imagine, with things like Facebook, you could actually reverse-engineer and identify people, or people at risk. But in the name of good, transparent science we did release some types keywords that indicated people who might be engaging in unprotected sex.

For example, people will search for, “How to find free sex,” and “How to hook up immediately.” They’ll tweet and say, “I need sex now” — that was literally one of the tweets on there. They will share anything and everything publicly on Twitter, and even more so in our semi-private online communities.

How did you parse that data between a real event and someone maybe just tweeting nonsense, like “I hope my ex gets syphilis”?
We manually screened them to see if the keywords were actually indicative of people’s intentions. We built the model to make sure they wanted to have sex, not just saying, like, “I love the song ‘Sex.’”

Do people who tweet about “needing sex immediately” actually find that sex immediately? (Asking for a friend.)
We do have some evidence from the HIV online communities that people who talked about getting an HIV test had 11 times the odds of actually requesting an HIV test. So we do find when people say they’re going to do something, they’re much more likely to do it.

In the future, then, if someone is tweeting about having STDs, would you say, “Let’s watch the area they’re from”?
If someone just says they have STDs, it doesn’t really matter. In fact if 1,000 people say it, and 1,000 people say it every month, it doesn’t matter. What matters is if 100 people say it, and then the next week, 600 people say it. That indicates something is happening, and we need to pick up on it and anticipate some issues that may result.

Can you use this data to find the first tweet, tracing it back to patient-zero?
We don’t have the ability to do that right now. But by incorporating other types of data — genetic data especially — eventually researchers will have the ability to find that information.

Would there be consequences for that person?
Absolutely. I mean, we’re seeing that now. Companies are building tools to monitor medication adherents — pill sensors that let you know someone is taking their medication. What are the ethical, policy and criminal implications of knowing someone didn’t take their medication that could’ve [led them to] hurt someone else or society? Same type of issue here.