Identifying Super-Mediators of Information Diffusion in Social Networks

We propose a method to discover a different kind of influential nodes in a social network, which we call “super-mediators”, i.e., those nodes which play an important role in receiving the information and passing it to other nodes. We mathematically formulate this as a difference maximization problem in the average influence degree with respect to a node removal, i.e., a node that contributes to making the difference large is influential. We further characterize the property of these super-mediators as having both large influence degree, i.e., capable of widely spreading information to other recipient nodes, and large reverseinfluence degree, i.e., capable of widely receiving information from other information source nodes. We conducted extensive experiments using three real world social networks and confirmed that this property holds. We further investigated how well the conventional centrality measures capture super-mediators. In short the in-degree centrality is a good measure when the diffusion probability is small and the betweenness centrality is a good measure when the diffusion probability is large, but the super-mediators do depend on the value of the diffusion probability and no single centrality measure works equally well for a wide range of the diffusion probability.