Twitter has come a long way since the 2008 election, when the platform was just two years old and had a few million users. Now, 140 million active users around the planet use it to spark revolutions and rant about reality TV shows. Twitter was the secret sauce that made this election's presidential debates palatable, generating 10 million tweets about the first debate alone, 7.2 million for the second, and 6.5 million for the third, not to mention a circus of memes and the hashtags to entertain jaded viewers. But who are these politically minded tweeps? What can we learn from the vast ocean of public status updates out there?
Bluefin Labs, a company that grew out of research on language and machine learning at the MIT Media Lab, has developed tools to filter and analyze this data. Parsing public Twitter and Facebook updates, they can see what different demographics are saying about various brands and put together a profile of what the average, say, Starbucks aficionado looks like. This year they've launched a noncommercial, nonpartisan project called the Crowdwire to apply these techniques to politics. Substitute "Obama" for "Starbucks," and you can take a look at who his fans are and what else they like.
The mobile and social video news network Now This News collaborated with Bluefin Labs to relate their findings in the short, entertaining video below. The research draws from both Twitter and Facebook, but because most Facebook updates are private, the bulk of the data comes from Twitter. The results might surprise you:
You can have a laugh at the expense of the Coors Light-swigging, Hannity-watching Romney bro, or the lobster-eating, skinny jeans-wearing Obama fangirl, but the science behind these portraits is pretty fascinating too. We spoke with William Powers, the author of Hamlet's Blackberry and director of the Crowdwire project, to learn more about how they did it. Powers expressed some skepticism of citizen journalism in a 2007 piece he wrote for The Atlantic, so why would he want to study citizen tweets? On Twitter, he says, "everybody gets a megaphone," but there's more to it than that:
"I think it's revealing in that Twitter comments, tweets, are almost by definition really granular. They're coming from these individuals who are speaking out of their own lives. So you have this conversation about the election, which is this big topic all these individuals now get to have a voice in by responding to debates and so forth. That's exciting, but it becomes even more granular when you can cross-reference it with other things they've revealed about themselves publicly."
First, Bluefin sifted through public tweets and Facebook updates and sorted users into two groups if they showed clear support of either candidate. They looked for certain positive phrases and hashtags to make sure that users weren't just criticizing a candidate. They found 82,000 Romney supporters and 144,000 Obama supporters and they've been tracking them for the past two months. Why were there so many more clear Obama fans? Powers theorizes that as an incumbent, Obama's just been in the spotlight for longer. Based on further analysis, Bluefin determined that the Romney group was 45% female and the Obama group was 53% female. You, in fact, might be in one of these groups, if you show a strong affinity for one of the candidates in your tweets.
Bluefin then sorted through the tweets from these two groups to see which brands they mentioned more frequently. They discarded brands that clearly carried a significance beyond simple preference, like Chick-fil-A, for example. It would be too difficult to sort the real Chick-fil-A fans from its critics, given the storm of attention the fast food chain generated. The results don't necessarily mean that users love Sprite or Walmart, but simply that they discuss Sprite and Walmart often. Powers, who regularly shares new findings on the Crowdwire blog, thinks this kind of analysis will become an important tool as it gets more sophisticated:
"It's a massive conversation. It's sort of the ultimate big-data challenge, or it's becoming that ... it's a frontier of journalism and of democracy at the same time. We're going to need tools like the ones we're using, and they're going to evolve -- they're going to have to evolve to really make sense of it. Basically you get this big, big portrait of democracy in action that is very different from what polls have always shown us because it's in the wild, you know, it's happening out there organically. It's not people answering a given set of questions at dinner time. It's not better than polls; it's just a new way of taking the public's temperature and seeing what they're thinking at any given moment."
This article available online at: