Social dynamics in a large-scale online game

Complex systems — when treated as systems accessible to natural sciences — pose tremendous requirements on data. Usually these requirements obstruct a scientific understanding of social phenomena on scientific grounds. Due to developments in IT, new collective human behavior, new dimensions of data sources are beginning to open up. Here we report on a complete data set of an entire society, consisting of over 350,000 human players of a massive multiplayer online game. All actions of all players over three years are recorded, including communication behavior and social ties. In this work we review the first steps undertaken in analyzing this vast data set, focusing on social dynamics on friend-, enemy- and communication networks. This new data-driven approach to social science allows to study socio-economic behavior of humans and human groups in specific environments with unprecedented precision. We propose two new empirical social laws which relate the network properties of link weight, overlap and betweenness centrality in a nonlinear way, and provide strong quantitative evidence for classical social balance assumptions, the weak ties hypothesis and triadic closure. In our analysis of large-scale multirelational networks we discover systematic deviations between positive and negative tie networks. Exploring such virtual "social laboratories" in the light of complexity science has the potential to lead to the discovery of systemic properties of human societies, with unforeseen impact on managing human-induced crises.