Statistical physics approaches to large-scale socio-economic networks

Michael Szell
2011 unpublished
In the past decade a variety of fields has been explored by statistical physicists, leading to an increase of our quantitative understanding of various systems composed of many interacting elements, such as social systems. However, an empirical quantification of human behavior on a societal level has so far proved to be tremendously difficult due to problems in data availability, quality and ways of acquisition. In this doctoral thesis we compile for the first time a large-scale data set
more » ... ing of practically all actions and properties of 350,000 odd participants of an entire human society interacting in a self-developed Massive Multiplayer Online Game, over a period of five years. We describe this social system composed of strongly interacting players in the game in three consecutive levels. In a first step, we examine the individuals and their behavioral properties over time. A scaling and fluctuation analysis of action-reaction time-series reveals persistence of the possible actions and qualitative differences between "good" and "bad" players. We then study and model the diffusion process of human mobility occurring within the "game universe". We find subdiffusion and a power-law distributed preference to return to more recently visited locations. Second, on a higher level, we use network theory to quantify the topological structure of interactions between the individuals. We focus on six network types defined by direct interactions, three of them with a positive connotation (trade, friendship, communication), three with a negative one (enmity, attack, punishment). These networks exhibit non-trivial statistical properties, e.g. scale-free topology, and evolve over time, allowing to test a series of long-standing social-dynamics hypotheses. We find qualitative differences in evolution and topological structure between positive and negative tie networks. Finally, on a yet higher level, we consider the multiplex network of the player society, constituted by the coupling of the single network layers. We quanti [...]
doi:10.25365/thesis.19044 fatcat:m4tdj7ebzvbzbi57jhcgsjn74q