Author Reputation and Text Trust in the Wikipedia

The Wikipedia is a source of information widely used in homes, schools, and offices across California, and across the nation. The Wikipedia is also the most successful, and influential, example of collaborative content creation. Most Wikipedia articles can be edited by anyone. This openness has been a key factor in the success of the Wikipedia, but it also poses challenges. Rogue, misinformed, or misguided contributors can corrupt articles with incorrect information, and it is difficult for readers to judge the reliability of the information they are presented. We propose to develop a reputation system for Wikipedia authors, and to use the resulting author reputation to compute a trust value for the text of Wikipedia articles. The author reputation we compute can be used to limit access to critical or controversial articles to authors of good reputation. The trust value, computed for each word of each article, will enable readers to form an informed opinion on the reliability of the information. We will demonstrate our approach by building a mirror site of the English Wikipedia, where the article text is colored according to its trust value. The Wikimedia Board of Trustees2 has expressed interest in this project as part of the Board’s drive towards reliability and sustainability in the Wikipedia. We expect that the techniques we develop will be useful in a much wider context than the Wikipedia. Our research addresses the core issue of judging the trust value, and tracking the provenance and source reputation, for bodies of evolving information. This can be applied in any setting where information is gathered or created in an open, collaborative, or heterogeneous fashion.