What does Twitter have to say about politics?

Here’s another Twitter-analysis post for all you network science junkies out there.

And although I’m a bit late to the table (New Scientist reported on this a week ago) I couldn’t resist.

Bruno Gonçalves, a postdoc in Alessandro Vespignani‘s research group here at Northeastern, and three colleagues at Indiana University expanded their research on partisan Twitter networks with a paper to be published in the open-access journal EPJ Data Science (where Vespignani is a co-editor in chief).

The team looked at Tweets from over 18,000 Twitter users during the 18 week period leading up to the 2010 midterm election. They separated the users into “left-leaning” or “right-leaning” groups based on their use of political hashtags like #tcot (top conservatives on twitter) and #p2 (Progressives 2.0). The “substrate” network is defined by follow relationships and shows how the two partisan groups keep to themselves. This isn’t new, Gonçalves’ team and others have shown similar results previously. I blogged about it not too long ago, actually.

What is surprising is that the activities within the segregated groups are rather unexpected based on the social-media activities of right- and left- leaning groups leading up to the 2008 presidential election. Back then, according to the article “the Obama campaign enjoyed twice as much web traffic, had four times as many YouTube viewers and five times more Facebook friends compared to the McCain campaign.”

In 2010, the right-leaning crowd took the social-media lead, at least on Twitter. The team found that, compared to left-wingers, right-wingers are more likely to self-identify as such in their bios, spend more time talking about politics on Twitter and do so more often, as well as share more politically-related information in the form of hyperlinks.

The right wing group is more tightly interconnected, has a larger network of politically-minded followers and has “a higher degree of engagement with the Twitter platform itself.”

Finally, they used self-reported location data (as opposed to geolocation data) to determine where the groups are Tweeting from, geographically speaking. While there is some overlap between the Twitter map and political party maps, the team accounts for the discrepancy by saying that perhaps outliers in a politically homogenous community turn to Twitter for their discourse, which seems like a reasonable enough explanation to me.

What I’m still confused about though, is how much this means about the political conversation as a whole. Couldn’t it be that left-leaners are more active on Facebook, for instance, instead of Twitter? What does it mean to have an active Twitter community when there are several other types of communities out there…like for instance the so-called “grass roots” community which may have once been called people-talking-to-each-other-in-real-life.

I’m new to the Twittersphere and quickly realizing how powerful a tool it is for information dissemination. So powerful that it scared me away for years before my boss (Hi Lucy!) said I had to get on it. But while Twitter could be a good predictor of total political communication behaviors, I’m not sure whether we’ve proved that yet. Maybe you know.

Anyway, all of this is based on the 2010 election and I’m getting really excited to see what comes out of the network science labs regarding the upcoming presidential election.