By now most folks have heard about using Tweets to try and predict things like fashion trends, disease, political change (ex. Arab Spring) and customer buying behavior. There’s been some pretty neat research on these topics. But what about trying to predict Tweets themselves?
A colleague named Adina recently sent me this article from the Sydney Morning Herald about two professors who've built a model to predict how often Tweets will get re-Tweeted. The algorithm doesn't look at the content of the Tweet itself, instead it focuses on how Tweeters (i.e. people who Tweet) behave.
According to a related article, the MIT professor who co-authored the Tweet algorithm said “what happens in the first couple of minutes [following a Tweet] potentially informs you about what’s going to happen the rest of the day.” You can see a bunch of these researchers’ predictions and results on their site Twouija.
Sometimes the results were really good, other times they were off mark. Nevertheless it’s a provocative idea.
What does this mean for Tweeters? Apparently, it opens the possibility for not-so-famous people to start writing more popular Tweets. Ashton Kutcher beware!
Fascinating research to be sure, yet I keep thinking about it this way… in a sense this algorithm is predicting the popularity of Tweets… meanwhile Tweets themselves are being used by some to predict other things (ex. fashion, politics, consumer behavior). String it all together and you can easily end up with algorithms predicting algorithms.
Is this interesting? Definitely.
Is it fun? Possibly.
Is it Tweet-worthy? I don’t know… you tell me.