Social phenomena are complex and therefore difficult to analyze quantitatively. In fact, our present understanding of processes such as spreading of a popular idea or diffusion of a new radical opinion is still only at a descriptive level. Various statistical and computational models have been proposed in the recent decades, attempting to capture these processes through the framework of dynamical systems and complex networks. Despite being intuitively clear, these models are usually able to predict only a limited range of behaviors of a given social system. In this talk, we propose a different approach in examining these models. In particular, we suggest to search for specific networks topologies, on which the behavior of a given model of social dynamics resembles more closely realistic scenarios.