Searching the Net for Differences of Opinion

Warren SACK
Film and Digital Media Department, UC Irvine

John KELLY
Center on Organizational Innovation, Columbia University

Michael DALE
Digital Arts/New Media, UC Santa Cruz

     Full text: PDF
     Last modified: May 19, 2005

Abstract
Political theorists, at least since John Stuart Mill in his book On Liberty (1859), have repeatedly asserted that exposure to conflicting viewpoints is beneficial for democracy. Through exposure to political viewpoints contrary to their own, citizens are said to gain political tolerance and an understanding of opposing rationales. Recent empirical work (e.g., Fishkin, 1991; Mutz, 2002) has confirmed these assertions. However, there is no clear means by which a citizen can find opposing opinions. Factors such as the consolidation of media ownership (e.g., Bagdikian, 2004), neighborhood segregation (by, for example, race and class), lack of weak ties in personal, social networks (Granovetter, 1973), proliferation of ideological-exclusive weblogs and radio and television talk shows, and recent technological developments that allow the “filtering” of Internet-distributed news (e.g., Sunstein, 2002) all make it difficult for individual citizens to find significantly different opinions. Contrary to Negroponte (1996), we posit the development of a software technology to facilitate the construction of a “Daily Not Me,” i.e., a semi-automated “clipping service” that, when given a topic (e.g., “abortion”), will return a range of diverse opinions about the topic (e.g., “pro-choice” and “pro-life”) that contradict one’s personal views. In this paper we present some preliminary results towards this long-term goal. Our work bootstraps recent, prior work in which one of the co-authors (Kelly, 2004) used qualitative content analysis to characterize the political leanings of one hundred and twenty, prolific, Usenet newsgroup authors. Software was developed to automatically download, from a Usenet newsgroup archive, tens of thousands of discussion threads containing over one million individual messages. Within these threads of discussion we were able to find several thousand message “tuples” in which known discussants (i.e., two or more discussants identified by Kelly) of differing political opinion exchanged messages. We have performed an empirical analysis of the structural characteristics (e.g., size, branching factor) of the threads surrounding these message “tuples” of conflicting viewpoints. Our goal is to identify a set of computable, search heuristics that might be employed in a “Daily Not Me” technology for finding opposing, political viewpoints as expressed in the archives of online discussion groups.