Spotting trendsetters: Inference for network games

Network games provide a basic framework for studying the diffusion of new ideas or behaviors through a population. In these models, agents decide to adopt a new idea based on optimizing pay-off that depends on the adoption decisions of their neighbors in an underlying network. Assuming such a model, we consider the problem of inferring early adopters or first movers given a snap shot of the adoption state at a given time. We present some results on solving this problem in the low temperature regime. We conclude with a discussion on reducing the complexity of such inference problems for large networks.