Fil Menczer has quite a title. He is “a Professor of Informatics and Computer Science and the Director of the Center for Complex Networks and Systems Research at the Indiana University School of Informatics and Computing.” For our purposes, though, he’s a leading Twitterologist, as the creator of the Stephen Colbert-inspired Truthy, a site that tracks how memes–often, specious political memes–spread on Twitter. Fast Company caught up with Menczer on the occasion of an update of the Truthy site, which is now more user-friendly.
FAST COMPANY: What was Truthy originally?
FIL MENCZER: Truthy was born mainly to look at “astroturfing,” or the abuse of social media to give the impression that there is a grassroots campaign where in fact there isn’t. After the 2010 elections were over, we used a lot of the data we collected to start asking questions about how information propagates in social media. Can we find regularities in these patterns? Can we understand the mechanisms that underlie these shapes?
We got funded by a National Science Foundation grant to make Truthy into a public tool, something citizens can use. For example, reporters like yourself can access the data we have and really interact with the data to get a better handle on what we’re observing. We designed this in collaboration with the School of Journalism here.
So walk me through what we can learn about a given meme, say the hashtag #mitt2012.
We have access to a sample of Twitter’s data called the garden hose. It’s about a 5-8% sample, we think. When you search #mitt2012 on our site, you get data from our database regarding all the tweets that included that hashtag. For all those tweets, every one has a user that generated it. Then it may have been retweeted by some other users; when that happens, we connect those two nodes. Some nodes are small, and some big, based on how many times a user is retweeted.
You might ask, who is @norsu2? Why are so many people retweeting this person? We load a whole bunch of information about this user that was not available before. You can see this person generates a huge amount of tweets, and also you can see the partisanship of this person–we automatically detect if a person is right-leaning or left-leaning.
How do you do that?
Based on the structure of the network. When it comes to political conversations on Twitter, users are very strongly polarized, clustered into two groups. People on the left only retweet people on the left; people on the right only retweet people on the right. Say I’m a liberal, and I want to say something bad about Santorum. A lot of conservatives might read my message, but they’re not gonna retweet it. They might reply, saying “You asshole,” but retweet networks are highly clustered. We’ve found they let you accurately classify the partisanship of a user who tweets about politics with about 95% accuracy.
So essentially your tool helps us be more scientific about the way we assess influence on Twitter.
It lets any citizen become a researcher and analyze who is promoting an idea. We have to become a little less naïve about social media, just as with any technology. We have to realize there are people behind it, and if there is some political or economic interest, they will try to cheat. We want to drop some of the mystery behind social media and let people get a bit of a zoomed-out view. On Twitter, you can see your friends and who they follow, but it’s harder to get a fisheye view of what the collective users are saying about something. We hope to allow people to get that kind of view.
Diffuse benevolent memes about Fast Company on Twitter.