Political polarization is tricky to measure, but those looking for tangible evidence of it in American politics can find it on Twitter, the proving grounds for every aspiring pundit. A group of computer scientists recently published a study (PDF via The Monkey Cage) analyzing how politically active Twitter users interacted with one another in the weeks preceding the 2010 midterm elections. They developed an algorithm to identify 250,000 tweets with hashtags related to the midterm elections--anything from #whyimvotingdemocrat to #palin12--from about 45,000 Twitter users.

The first network, on the left, is made up of retweets--a line is draw between two Twitter users if one retweeted something from the other. This network segregated itself into two homogenous groups--left-leaning individuals, in blue, and right-leaning, in red. What it suggests is that people tend to rebroadcast information on Twitter that they agree with ideologically. The second network shows who mentions whom. This network, in contrast, is one big heterogenous group of accounts. These same Twitter users don't hesitate to @-mention someone whose politics differ from theirs.

The results are fascinating, but their implications are not entirely obvious. Our hypothetical Twitter user won't retweet messages he disagrees with since retweets, after all, are sometimes thought of carrying an endorsement (which is why so many journalists insert a "Retweet does not imply endorsement" disclaimer in their Twitter profiles). But simply mentioning opponents is fairly common. So, it shows that while cross-ideological interactions clearly exist on Twitter (perhaps even more so than on partisan blogs), they mostly take the form of shots at political opponents.