Bot Detection in Reddit Political Discussion

Abstract

The existence of social media bots on political forums can muddle the perception of public opinion. Bot detection has been successful on platforms such as Twitter, Facebook and Youtube. However, our research focuses on characterizing suspicious behavior on the social media platform Reddit; this platform structurally differs from its peers in that users subscribe to page topics and contribute to the discussion through comments or topical posts. We hypothesize that persons who intend to influence the public opinion would deploy paid users, or social media bots, to artificially amplify a political sentiment. We validate our hypothesis by using a network-based approach that reveals a fully connected band of users which includes users with the word "bot" in their name, as well as users who exhibit otherwise a bot type behavior.