John Kelly and mapping blogosphere links

John Kelly, a researcher with Berkman and Columbia University, is one of the founders of Morningside Analytics, a company that’s trying to discover and map connections in the blogosphere and media as a whole. He’s presenting his recent research at the Annenberg/Berman Media Re:public conference.

Before showing some of the quantiative research he’s worked on, Kelly offers some observations that guide his research:

– “Bloggers versus journalists” is just a food fight – they’re all part of one system of communications

– Communications is in a long process of deep evolution, moving from the postal service to telephone to internet. (He notes that very little has been written on social impact of telephones.)

– Blogs are a key to understanding networked social spheres.

– Communications research needs an overhaul, because our current methods misunderstand society.

– We’ve always been a networked society, and now we know it.

With these principles in mind, Kelly shows some of his maps. They’re collections of colored dots, each representing media sources (blogs and mainstream media). Some show clusters of blogs that link to one another. Others show “attention maps” – what sources the bloggers are linking to, representing what they pay attention to. At the center of these attention maps, Kelly tells us, are the New York Times and the Washington Post. Attentive clustering allows Kelly to group together bloggers who share common news sources.

He shows us maps f five different language blogospheres – English, Persian, Russian, Arabic, and Scandinavian (Swedish, Norwegian and Danish). Each has a different shape, a different number of clusters. The Iranian blogosphere, which Kelly has studied closely, has some unusual clusters:
– poetry bloggers
– secular, expatriate reformist bloggers (what most people think of as the Iranian blogosphere)
– the conservative, religious blogosphere, which has clusters around shia sects and youth groups

The clusters in the US blogosphere are, unsurprisingly, quite different. Much of his work has focused on looking at what media liberals and conservatives look at and what terms they use. Both left and right-wing bloggers use the term “wingnut”, for instance, while the right uses the term “moonbat” to describe the wacky left, and the left-wingers don’t really know the term. Tools like a “word radius analysis” look at words that appear within five words of a term like “security” – based on cluster analysis, he determines which terms are used by liberals and conservatives. He concludes that conservatives use more abstract terms to talk about security, while liberals tend to talk in extremely concrete terms when using that word.

Some recent work looks at the speed and persistence of linking to certain media objects. Kelly is interested in questions like, “Was the video of Reverend Wright’s viral, or was it merely salient?” If something was simply salient, it would cause ripples in the communication environment, almost like background radiation. His analysis shows that different types of media have different attention patterns: mainstream news stories tend to peak very quickly, while wikipedia articles are linked over very long periods of time. YouTube videos tend to peak as quickly as mainstream media, with a small exception for videos that truly go viral. Kelly believes it may be more common for videos to be put on YouTube by people attempting to set agendas in mainstream media – they seed YouTube, then point to it as a way of arguing that “the bloggers are talking about a story”, even though they’ve planted the story.