Social Media Could Predict the World's Next Mass Protest

February 18, 2014 // 07:27 PM EST

Image: Ivan Bandura/Flickr

Scientists and data wizards have been trying to predict everything from illness to financial markets to elections and revolutions—and with some success. After social networks’ presumably instrumental role in the Arab Spring came to light, researchers started digging into whether social media platforms could also be used to anticipate major social uprisings. A new study out of MIT and posted on the arXiv preprint server is the latest to give it a whirl. 

Study author Nathan Kallus says he’s found a new approach to predictive analysis that “validates and quantifies the common intuition that data on social media and the web (beyond mainstream news sources) are able to predict major events.”

Image: MIT

Using the 2013 coup d'etat in Egypt as test model, he fed 300,000 web sources in seven languages through a computer, which used natural language processing, keyword search, and sentiment analysis to spot clues that a political upheaval was afoot. He found that by extrapolating the resulting patterns into future, he could forecast upcoming protests with “high accuracy.” 

Image: MIT

Okay. But of course, it’s much easier to spot patterns and clues in retrospect, when you know what you’re looking for. The question of whether the right algorithm can turn Twitter into a cyberspace crystal ball is still unanswered. So is, for that matter, whether we’d even want it to.

If you ask me, that’s the more interesting question. If “massive online-accessible public data,” as Kallus calls it, does eventually enable reliable social forecasts, what would the ramifications be? The MIT study says warning people about major protests could prevent unnecessary violence. Or businesses, governments, and homes could ramp up security if they’re able to anticipate an attack.  

But what if oppressive regimes leveraged the cyberspace crystal ball to evade or thwart public demonstrations? Could too much insight shift the power away from the crowd and back toward the "Man," as we're already seeing online surveillance threatening to do?

It’s not a crazy question. Early last year, before the PRISM scandal hit, news that law enforcement was tracking and analyzing internet users’ info to anticipate crimescaused a major stir in privacy and civil rights circles. A couple years before that, the Pentagon began studying big data and predictive algorithms as part of its intelligence-gathering strategy—specifically focused on political crises, revolutions, and social and economic instability.

But I suppose at this point such theoretical dystopian fears are premature. Future-predicting social media experiments are still riddled with caveats. Despite years of trying, no one's quite figured out how to harness the monster of publicly available online data to predict the future. It’s a formidable task—Twitter alone sees some 58 million tweets every day. 

To take the MIT study for example, there could be a lot of buzz on Twitter leading up to an event that for whatever reason never came to be. Or someone could theoretically decide to game the system with fraudulent tweets to derail a planned protest.

Whether social media will ever be an effective peephole into the future, there’s no way to tell. But the mere idea that it could be possible, that someday it probably will be, is too titillating for data scientists to ignore.