Structure and Strategy: How the Games We Play Influence our Interaction Networks

-----

The co-evolution of structure and strategy in games has been identified as a significant factor in the emergence many social behaviors of interest to philosophers such as altruism, fairness, signaling, and spite. Smead presents a general framework for modeling the co-evolution of social structure (networks) with social behavior (strategies). Smead then explores two applications of this framework. First, Smead applies it to open theoretical questions the evolution of cooperation, spite, and fairness and show that the general framework is able to reproduce serveral existing results. Second, empirical applications are explored. To do this, a graph neural network (GNN) was trained on synthetic data from a minimal version of the model to learn the relationship between the game played and the resulting interaction network. Then, when given a real-world network, the GNN can classify the type of network, project what game is likely being played, and predict unknown edges in these networks.

-----