[dlan1000]

Topological properties of an individual (such as network degree or betweeness centrality) are not necessarily predictive of influence. The author acknowledge this and suggests observation of peer response to messages sent by FB on behalf of the user. What is really necessary here to determine influence is a randomized trial. See for example the methods described here: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1564856 (full disclosure - I am a co-author of this work and of some upcoming work that describes techniques of identifying influential and susceptible users in social networks).