Autonomy Talks - 21/09/20

Speaker: Nicolò Pagan, Automatic Control Lab, ETH Zürich

Title: Game Theoretical Inference of Human Behavior in Social Networks

Abstract: Social networks emerge as the result of actors’ linking decisions. We propose a novel game-theoretical model of socio-strategic network formation on directed weighted graphs, with continuous action spaces, where actors are endowed with individual parametric payoff functions. In our previous work, we provided a theoretical analysis of specific network motifs of homogeneous rational agents. Here, we relax these assumptions and deal with complex, real-world networks of heterogeneous and irrational actors. The goal is to learn the parameters of the payoff functions constructing an inverse optimization problem using the Nash equilibrium condition. In other words, we want to provide the most rational estimate of the heterogeneous individual parameters from an observed state of the network, e.g., a Nash equilibrium. We show that, when agents are rational, the solution set results from the intersection of finitely many half-spaces. Conversely, when agents are irrational, an exact solution dœs not exist and we deliver Least Square estimates of the parameters with confidence intervals analysis. We provide evidence that our results are consistent with empirical, historical, and sociological observations on real-world data-sets. Furthermore, our method offers sociological and strategic interpretations of random network models, e.g., preferential attachment and small-world networks.

Find out the next talks at [ Ссылка ]

свернуть