Moderated Online Communities and Quality of User-Generated Content

Abstract

Online communities provide a social sphere for people to share information and knowledge. While information sharing is becoming a ubiquitous online phenomenon, how to ensure information quality or induce quality content, however, remains a challenge due to the anonymity of commentators. This paper introduces moderation into reputation systems. We show that moderation directly impacts strategic commentators incentive to generate useful information, and moderation is generally desirable to improve information quality. Interestingly, we find that when being moderated with different probabilities based on their reputations, commentators may display a pattern of reputation oscillation, in which they generate useful content to build up high reputation and then exploit their reputation. As a result, the expected performance from high-reputation commentators can be inferior to that from low-reputation ones (reversed reputation). We then investigate the optimal moderation resource allocation, and conclude that the seemingly abnormal reversed reputation could arise as an optimal result. The paper concludes with a discussion of the development of a scientific moderation system with application to academic publishing.