Using Connected Accounts to Enhance Information Spread in Social Networks

Abstract

In this article, a new operation mode of social bots is presented. It includes a creation of social bots in dense, highly-connected, sub structures in the network, named Spreading Groups. Spreading Groups are groups of bots and human-managed accounts that operate in social networks. They are often used to bias the natural opinion spread and to promote and over represent an agenda. These bots accounts are mixed with regular users, while repeatedly echoing their agenda, disguised as real humans who simply deliver their own personal thoughts. This mixture makes the bots more difficult to detect and more influential. We show that if these connected sub structures repeatedly echo a message within their group, such an operation mode will spread messages more efficiently compared to a random spread of unconnected bots of a similar size. In particular, groups of bots were found to be as influential as groups of similar sizes, which are constructed from the most influential users (e.g., those with the highest eigenvalue centrality) in the social network. They were also found to be twice more influential on average than groups of similar sizes of random bots.