Improving Similarity-Based Methods for Information Propagation on Social Networks

Social networks are a means to deeply and quickly propagate information. However, in order not to vanish the advantage of information propagation, the amount of information received by users should not be excessive. A typical approach operates by selecting the set of contacts to which propagate the information on the basis of some suitable similarity notion, trying thus to reach only nodes potentially interested. The main limit of this approach is that similarity is not completely suitable to propagation since it is typically non-transitive, whereas propagation is a mechanism inherently transitive. In this paper we show how to improve similarity-based methods by recovering some form of transitive behaviour, through a suitable notion called expectation. The non-trivial combination of similarity and expectation in a nice mathematical framework, provides the user with a flexible tool able to maximize the effectiveness of information propagation on social networks.