Ranking Social News Articles Based on Voter Credibility

Given the wide spread of web based tools and social news media services which are facilitating grassroots journalism, there is a growing interest in selecting credible news content among a huge number of articles. Currently, most of social news services rely on reader votes to select articles for their front pages. However, the fundamental problem is that users’ votes often stand for popularity rather than credibility. In this paper, we propose a system to address this problem using a weighted voting system. Specifically, we trace thousands of users and their votes, differentiating them depending on how credible the articles voted for are. We then calculate each user’s voting credibility and use it as the user’s voting weight in our system. The results indicate that our method performs better in selecting credible news articles than other methods replying on a “one person, one vote” system. The results suggest feasible solutions to problems in social news media concerning media credibility. Introduction & Related Work Recently, news media have been going through a huge change by the emergence of the web-based publication tools and social news media which lower the cost of publication and dissemination of news and information. However, it raises difficulties in selecting credible news content. While staff editors can look into well-organized articles of professional journalist, investigating a large number of news articles including ones from amateur writers is practically impossible in the social news services. Instead, general users play an important role in selecting news content usually through their voting behavior. However, the problem is that users’ votes often stand for popularity rather than credibility. Moreover, there has been reported a possible bias caused by a particular group of people, such as so-called digg mafia, or reddit downmod squad. In this paper, we propose a weighted voting based Copyright © 2010, Association for the Advancement of Artificial Intelligence (www.aaai.org). All rights reserved. news ranking system to deal with this problem with collective intelligence. Studies on traditional news credibility focus on finding components for measuring perceived media credibility (Infante 1980; Meyer 1988). As the web changed the ecosystem of journalism, recent studies focus on not only estimating media credibility with the measuring factors specified to online journalism (Kiousis 1999), but also assessing the accuracy of news review instruments (C. Lampe & R. Kelly Garret 2007). However, these studies cannot be computationally applied to news ranking systems. While there have been several computational approaches focusing on the overall media credibility problems (Sohn et al. 2008) and the media bias problems (Park et al. 2009, Munson et al. 2009), most of the studies take advantage of the well-defined news structure or do not cover the entire news credibility, both of which are not guaranteed in non-professional participatory forms of journalism.