Influence in a large society: Interplay between information dynamics and network structure

Motivated by the recent emergence of large online social networks, we seek to understand the effects the underlying social network (graph) structure and the information dynamics have on the creation of influence of an individual. We examine a natural model for information dynamics under two important temporal scales: a first impression setting and a long— term or equilibrated setting. We obtain a characterization of relevant network structures under these temporal aspects, thereby allowing us to formalize the existence of influential agents. Specifically, we find that the existence of an influential agent corresponds to: (a) strictly positive information theoretic capacity over an infinite-sized noisy broadcast tree network in the first impression case, and (b) positive recurrent property of an appropriate (countable state space) Markov chain in the long-term case. As an application of our results, we evaluate the parameter space of the popular “small world” network model to identify when the network structure supports the existence of influential agents.