Debate, Division, and Diversity: Political Discourse Networks in USENET Newsgroups

1 Do online political discussions tend to aggregate diverse voices in cross-cutting debate and deliberation? Or do " audiences " for online discussion tend to fragment into ideological echo chambers? In the wilds of threaded discussion on the internet (as opposed to deliberative polls, moderated discussions, and other designed venues of deliberation), networks of political discourse emerge from billions of individual choices by millions of individual citizens about what to discuss online, where to discuss it, and with whom. Do these choices lead individuals to interact across ideological divides, or to cluster within them? How we understand the value of internet-based forms of political discourse is in part dependent on this question. The danger to a healthy public sphere, well articulated by Sunstein, is that individual-level selectivity biases, enabled by the internet's facilitation of choice in information consumption (" consumer sovereignty " in Sunstein's words), will precipitate a balkanization of political discourse (Sunstein 2001). In this view, individual preferences for reinforcement in political information and dialog are held in check by public institutions—information " commons, " like public parks or the mainstream mass media—in which citizens are exposed to a range of viewpoints they would not otherwise encounter. The internet represents not such much another public commons, but rather a kind of anti-commons which allows citizens to consume information and affiliate with others on the basis of shared values and interests only. The danger to a unified, balanced public sphere is compounded by the tendency of groups even slightly biased to one side of an issue to move further from the center in the direction of the bias. The implication is 2 that internet modes of political discourse are likely to promote the dissolution of the public sphere in to so many self-radicalizing " public sphericules, " to use Todd Gitlin's phrase (Gitlin 1998). Sunstein's dystopian fears are not without empirical merit. Network analyses of Amazon.com " also-bought " lists show an American public neatly bifurcated into conservative and liberal audiences for politically oriented books (Krebs 2004). Link analysis of the " blogosphere " likewise demonstrates the coalescence of blogging networks into distinct liberal and conservative clusters (Adamic 2005). The tendency of individuals to consume (and in the case of blogging, produce) media resonant with their pre-existing political perspectives is mirrored in the dynamics of political discussion within social networks. Face-to-face, individuals tend to discuss politics when they feel they …