Erasmus Mundus Masters in Complex Systems ENDOGENIZATION OF NETWORK TOPOLOGY IN METAMIMETIC GAMES

In this work, we aim to investigate the endogenization of topology parameters in metamimetic games and construct a framework that can be easily adapted for the simulation of this kind of games following any kind of topology. Metamimetic games are games where the agents follow dynamics of mimesis or imitation for their behaviour and rules. Moreover, their own rules for imitation can be imitated from nearby agents. We present a new paradigm for metamimetic games by defining neighbourhoods through links instead of spatial proximity, something that has not been done before. We aim to investigate if, when endogenizing some topology parameters (such as the likelihood to change a link), the resulting network would somehow resemble a real-life social network. This hypothesis comes from previous work with this class of games, where it was shown that sustained cooperation with a diversity of agents emerges from the model. As a result, under a certain range of parameters and starting from either a perfectly ordered network or a small-world network we do get a small-world network. However, if we start out from a completely random network, the gains in clustering were marginal, even at very extreme and unrealistic parameter ranges.