How does this black magic work, exactly? According to Patrick Meier, the Director of Social Innovation at the Qatar Computing Research Institute, it’s all powered by big data. Researchers at QCRI and the Indraprastha Institute of Information Technology in New Delhi had previously logged and analyzed millions of tweets sent after international crises, including the disastrous Chilean earthquakes in 2010 and the London riots in 2011.
Using that information, they were able to identify certain factors that predict credibility — things, such as the use of swear words and emoticons, the number of pronouns and unique characters, and the follower-base of the tweeter. This new tool is basically just a non-academic application of all that very academic research.
The big question, naturally, is whether it really works. Meier cautioned in a blog post that the tool is experimental, and will improve over time as it receives feedback from users. There’s an interpretation problem here, too, however: TweetCred ranks tweets on a scale of 1 to 7, and it’s not entirely clear what the practical distinction is between, say, a 3-star tweet and a 5-star one.
You’d also think TweetCred would assign higher ratings to major news outlets or other credible sources, but a scroll through @washingtonpost’s recent output uncovered nary a 7-star tweet.
So maybe TweetCred isn’t all there yet — but it’s a valuable first step. As Meier explains in his introduction of the technology, studies show people are far less likely to spread Twitter rumors if other users have flagged them as fakes. So while this won’t necessarily stop false information from spreading, it could, initially, slow it down. Frankly, in the Sisyphean battle against online shenanigans, we’ll take what we can get.