Peer recognition increases user content generation but reduces content novelty

Platforms that depend on user-generated content spend a great deal of effort crafting policies and mechanisms that can yield a steady stream of useful content. In this work, we consider the effects of awards offered by peers, a feature that platforms often provide to enable users to recognize the quality of peers' contributions. We set out to conduct a large-scale field experiment on Reddit, one of the biggest social news aggregation and discussion platforms in the world, evaluating the effect of peer recognition on content generation as well as content novelty. Leveraging the plentiful textual content and its innovative peer recognition mechanism, Gold Award, we purchased and then randomly assigned Gold Award to approximately 900 posts on Reddit, anonymously, over the course of two months. Collecting and analyzing users' behavioral trace data and content postings in the post-treatment period via Reddit's API, our results suggest that the probability of submission increased by 6.6% among treated users, relative to control users, and posts were 95.3% longer on average. Interestingly, however, the content that users post becomes more similar to their awarded content, and thus declines in its novelty. Based on this result, we conclude that peer recognition is a double-edged sword. On the one hand, peer recognition does indeed foster increased engagement and content production among users, but the additional content that results is less novel than content that arrives in the absence of recognition.