Eszter Hargittai on blogger insularity

Henry Farrell and Daniel Drezner have invited a handful of academics who focus on blogs to a meeting in Chicago to present papers that are intended to become chapters in a book titled “The Power and Political Science of Blogs”. Today and tomorrow, a number of the authors are presenting drafts of their papers for group critique. Up today were Eszter Hargittai and Cass Sunstein. (I’ll cover Hargittai’s talk in this post, Sunstein’s in the next.)

This was a particularly well planned grouping, as Hargittai’s research focused on testing a theory put forward by Sunstein in his book, Republic.com. Building on an idea from Nicholas Negroponte – “the Daily Me”, a digital newspaper custom-tailored to a reader’s interests and concerns – Sunstein worries that the Internet enables users to cocoon themselves in information, hearing only viewpoints they want to hear and never encountering dissenting voices. (Sunstein’s book is very much worth reading, but you can get a feel for his arguments in a shorter essay available online, “The Daily We”.)

Hargittai and collaborator Jason Gallo set out to test what they call Sunstein’s “fragmentation hypothesis”. If Sunstein is right, and the ability to select the news sources we interact with allows us to hear only our own views, we’d expect to see top political bloggers linking only to bloggers with similar political persuasions.

So Hargittai and Gallo created a list of 40 top bloggers (as selected from three blogranking engines), 20 from the left, 20 from the right, all well-linked and all focused on US politics. Then they retrieved seven weeks worth of post data from the blogs and analyzed interlinking between the blogs. In seven weeks, the blogs generated 13,329 posts and 2,609 links to the 40 blogs being tracked. (I think that’s a remarkable finding – roughly one in five blogposts, on average, linked to another very highly ranked blog – that starts looking like something of an insular conversation.)

Tracking only links in blogrolls, Hargittai and Gallo turn up some evidence that conservatives are more willing to link across ideological lines than liberals… but that neither side is particularly good about it – less than 10% of blogroll links are across party lines. Looking at links on a week by week basis, the news is a bit more encouraging – only 40% of conservative blogs hadn’t linked to a liberal blog in the set at least once, and the pattern with liberal blogs was fairly similar.

Using a nifty algorithm I hadn’t seen before – Krackhardt and Stern’s EI Index – Hargittai and Gallo score liberals and conservatives on their insularity – their tendency to link inside their network, rather than outside their network. Unsurprisingly, both groups are highly insular, linking to likeminded blogs more often than those from opposing viewpoints. In one week’s worth of data, liberals look more insular than conservatives; in another, conservatives look more insular than liberals.

Lots of what Hargittai and Gallo cover has been covered by Natalie Glance and Lada Adamic in a paper called “Divided they Blog” – I blogged Natalie’s presentation of the paper in Chiba a few months back. But Hargittai and Gallo go an interesting step forward and hand-classify the content of some of the posts in their set. (Given the size of the set, and the work involved, it’s not a surprise that they’ve only been able to classify a small subset.) They classify each post as “Agreement”, “Disagreement”, “Redirect” – i.e., a value-neutral reference to another post, “non-Political” – i.e., a reference to another blogger in a non-political context, and “Straw Man”, a dismissive and unfair attack on another position. (This last category proved problematic for some of the audience today – dismissing someone as a jerk is more common in this category than restating a weakened version of their argument, a true “straw man”.)

You’ll be unsurprised – though possibly disappointed – to discover that most cross-ideological blogging falls into the “straw man” or “dismissive” camp – 41% of conservative to liberal links, 62% of liberal to conservative links. Redirection is pretty frequent within ideologies, substantive agreement is pretty infrequent (right on, David!) and substantive disagreement is almost nonexistent within party groupings – there’s one instance of a liberal blog linking to another liberal blog and disagreeing and no instances within the conservative blogs. (Admittedly, it’s a small sample set – it will be interesting to see if the finding holds in a larger set.)

Hargittai argues that links across ideological boundaries are indications that conversations are taking place and that people are encountering different views in the blogopshere, even if they’re responding to these views by writing about them dismissively. I suspect Sunstein might have been more interested in the lack of disagreement finding, a result that paralells some of the findings on group deliberation he spoke about today.

Jay Rosen pointed out that an interesting dimension to the study might be to look at what media sources conservatives and liberals link to. Do both link equally to the New York Times, or do only liberals link there? Or do conservatives link and dismiss or critique coverage? It strikes me that some of the work I’ve been trying to do on my headliner research, looking to see which headlines from the Times, the BBC and other sources bloggers choose to link to, could be pretty easily adapted to try to answer this question. Hmm, more papers to write.

More about the research Eszter presented today on her blog.