Detection of Coordination Between State-Linked Actors

Abstract

Powerful actors have engaged in information control for centuries, restricting, promoting, or influencing the information environment as it suits their evolving agendas. In the Digital Age, information control has moved online, and information operations now target the online platforms that play a critical role in news engagement and civic debate. In this paper, we use a discrete-time stochastic model to analyze coordinated activity in an online social network, representing the behaviors of accounts as interacting Markov chains. From a dataset of 31,521 tweets posted by 206 accounts, half of which were identified by Twitter as participating in a state-linked information operation, we evaluate the coordination, measured by the apparent influence, between pairs of state-linked compared to unaffiliated accounts. We find that the state-linked actors exhibit more coordination amongst themselves than with the unaffiliated accounts. The degree of coordination between the state-linked accounts is also much higher than the observed coordination between the unaffiliated accounts. Additionally, we find that the account that represented the most coordinated activity in the network had no followers, demonstrating the power of our modeling approach to unearth hidden connections even in the absence of explicit network structure.