Coordinated link sharing on Facebook

Malicious actors regularly attempt to manipulate social media using coordinated posting. Many existing methods for detecting this coordination, though, have relied primarily on post-timing, which is trivially easy to change. In this paper, we make a significant methodological advancement in coordination detection, leveraging highly regular statistical patterns in the speed and frequency of sharing. We apply and validate this approach on Facebook, using 11.2 million link posts from a list of 16,169 most popular English-language Facebook pages that referenced at least one of the top eight US politicians in any post, a set of pages that produced more than 91% of all user engagement in this category during our collection period. Our approach can be calibrated and adapted across contexts, platforms, and times, allowing researchers to build valid, testable, but still human-interpretable models of platform manipulations.