Interpretable Cross-Platform Coordination Detection on Social Networks

Numerous disinformation campaigns are operating on social networks to influence public opinion. Detecting these campaigns primarily involves identifying coordinated communities. As disinformation campaigns can take place on several social networks at the same time, the detection must be cross-platform to get a proper picture of it. To encode the different types of coordination, a multi-layer network is built. We propose a scalable coordination detection algorithm, adapted from the Louvain algorithm and the Iterative Probabilistic Voting Consensus algorithm. This algorithm is applied to the previously built multi-layer network. Users from different social networks are then embedded in a common space to link communities with similar interests. This paper introduces an interpretable and modular framework used on three datasets to prove its effectiveness for coordination detection method and to illustrate it with real examples.