This weekend I’ll be at the annual meetings of the International Communication Association meetings in New York. All of the members from my research group will be participating in the conference and we’ll be reporting on several of our projects. Sunday midday we will present a poster summarizing some preliminary findings from our project on cross-ideological conversations among bloggers. I thought I would give a little preview here.
Cass Sunstein in his book Republic.com talks about the potential for IT to fragment citizens’ political discussions into isolated conversations. Borrowing from Negroponte, he discusses the potential for people to construct a “Daily Me” of news readings that excludes opposing perspectives. Sunstein argues that for democracy to flourish, it is important that people continue to have conversations with those in disagreement with their positions. However, he is concerned that with the help of filtering out unwanted content people will fragment into enclaves and won’t be exposed to opinions that challenge their positions. The book is an interesting read, but it does not offer any systematic empirical evidence of the claims.
I have been working on a project this past year with Jason Gallo and Sean Zehnder on empirically testing Sunstein’s thesis. We are doing so by analyzing cross-linkages among liberal and political blogs. You may recall that about two months ago Lada Adamic and Natalie Glance came out with a report on “The Political Blogosphere and the 2004 U.S. Election”. My first reaction was one of panic. Here we had been working on our project for months and someone else came out with the results first. However, a closer read made me realize that our project has some unique elements. And if nothing else, seeing that project has made us more careful and critical in our work showing that more research in an area can be fruitful, because hopefully it inspires the agenda to move forward in a productive manner.
[I updated this image on June 1 when I realized the right graph wasn't displaying exactly what I had described it as.]
Our work has focused on addressing two questions. First, we are interested in seeing the extent to which liberal and conservative bloggers interlink. Second, we want to see what kind of changes we may be able to observe over time. Sunstein’s thesis suggests that we would see very little if any cross-linking among liberal and conservative blogs and the cross-linking would diminish over time. We go about answering these questions using multiple methodologies. We counted links and calculated some measures to see how insular the conversations are within groups of blogs. We also did a content analysis of some of the posts in our sample. We continue to work on this project so these are just preliminary findings.
At this first stage, we report on findings from three week’s worth of posts on 41 blogs (20 conservative, 21 liberal). We classified the blogs ourselves for political affiliation based on their content. We sampled a week of posts from June 2004, October 2004 (the week before the elections) and March 2005. It may not sound like much, but these 41 blogs posted over 5000 entries during these three weeks (5214 to be precise) including over 900 links to each other (excluding links on blogrolls).
We look at blogroll links and links in posts separately. The graph on the left [click for larger version] is a network map of the blogroll links across these blogs. It is certainly the case that liberal blogs (blue) are more likely to link to other liberal blogs and the same is true for conservative blogs (red) as they are more likely to link to other conservatives. However, there is also interlinking on the blogrolls.
The graph on the right [click for larger version] is a representation of links in posts to other posts on the blogs in our sample (during a week in March). Again, there is more linking within the groups, but there is also linking across liberal and conservative blogs.
We analyzed the linking structure for each of the three weeks to see how the insularity of the conversations changed over time. This image shows a segment of our conference poster with the relevant figures. Certainly both conservative and liberal bloggers are more likely to link to those who share their political orientation. A figure closer to -1 suggests more insularity. There is no clear trend toward becoming more isolated in conversations over time, however. (We are in the midst of coding data for additional weeks to allow for more detailed analyses of the trends.)
Of course, it is possible that all interlinking across liberal and conservative blogs happens in a manner that is void of substance. To address this point, we undertook a content analysis of a subsample of the posts in our study (140 for now to be exact). We found that about half of the links represent what we classify as strawman arguments. The liberal bloggers in our sample are more likely to engage in such cross-linking than the conservative bloggers. However, we also found some evidence of substantive cross-linking. In these cases bloggers may either agree or disagree with the other person, but they do address the content of the other blogger’s post. Also, we did not find that bloggers address the substance of those who resemble their point-of-view very often either. (We present specific figures on the conference poster.)
Overall, it would be incorrect to conclude that liberal bloggers are ignoring conservative bloggers or vice versa. Certainly, liberal bloggers are more likely to address liberal bloggers and conservative bloggers are more likely to link to conservative bloggers. But people from both groups are certainly reading across the ideological divide to some extent.
Much remains to be done and we continue to work on the project. We are expanding our sample to a much larger set (120) that will also include blogs that fall into an “other” category on the political scale. We have been working on a spider to automate some of the process. If you can come by the conference poster session, Sean will have his laptop on hand to showcase the program. We are also coding more weekly segments, as mentioned above. Longer term, there are all sorts of additional ways in which the project can be expanded.
There are numerous methodological issues to consider with this type of a project. I don’t want this post to get unwieldy so I have not gone into details. Please do not assume that just because I did not mention something here we did not consider it. Chances are we did consider it and addressed it in some manner. Of course, we may have missed some issues and certainly welcome feedback.
I would like to thank my great undergraduate research assistant Matt Kane for his assistance with this project. Also, the great visualizations are thanks to Valdis Krebs‘ InFlow software. Thanks to NICO for its support as well.