CredibilityRank

Abstract

As news media have been expected to provide credible news contents to readers, the credibility of news media is mainly decided by how much more credible contents the news media provide and deliver. In traditional media, staff editors go through articles and arrange news contents to enhance their media credibility. By contrast, in social news services, general users play an important role in selecting news contents through voting behavior as it is practically impossible for staff editors to investigate thousands of articles sent to the services. Therefore, social news services have strived to develop news ranking algorithms that screen valuable news contents. However, the fundamental problem is that users`` participation often stands for popularity rather than credibility. In this paper, we propose a novel way to aggregate users`` votes to effectively find credible news contents. We differentiate each user``s votes based on the users`` voting history by tracing out what credibility values the article which the user voted for has. Then we suggest two aggregation methods, ``one person, one vote`` voting-based method using only elite users`` votes and weighted voting-based method using all voters`` weighted votes. To examine whether the proposed system shows better performance in selecting credible news contents, we adopt the proposed systems in other news pools and have the selected articles assessed by general users and journalism experts. As a result, the credibility scores calculated by the proposed systems show significantly higher correlation with assessed journalistic credibility, while the number of votes by popularity-based ranking system currently used in many social news services does not. Also, top-ranked news contents in the proposed system have significantly higher credibility scores than those in current popularity-based ranking systems.