This project is studying statistical inference from datasets tracking the diffusion of new ideas or behaviors through a population. Game theoretic models for the diffusion are utilized in which members of the population decide to adopt a technology by maximizing a pay-off that depends on an underlying network structure. Example questions include the identification of "first movers" and the most likely series of actions that result in a given observed state of the network. Algorithms are being developed for characterizing the maximum likelihood estimate of first movers for an evolutionary game theoretic framework with smoothed best response dynamics. Additionally algorithms to identify influential nodes and the network graph along with the associate payoff functions are being studied. The associated modeling and analysis build upon foundations in probability and statistics, Markov processes, statistical mechanics, optimization and game theory.
Understanding diffusions in social networks is broadly applicable across society including areas such as marketing, economics and social sciences; efforts are being made to disseminate the results of this work to such fields as well as to incorporate ideas into undergraduate and graduate courses in EECS.