Modeling User Dynamics in Collaboration Websites

Numerous collaboration websites struggle to achieve self-sustainability—a level of user activity preventing a transition to a non-active state. We know only a little about the factors which separate sustainable and successful collaboration websites from those that are inactive or have a declining activity. We argue that modeling and understanding various aspects of the evolution of user activity in such systems is of crucial importance for our ability to predict and support success of collaboration websites. Modeling user activity is not a trivial task to accomplish due to the inherent complexity of user dynamics in such systems. In this chapter, we present several approaches that we applied to deepen our understanding of user dynamics in collaborative websites. Inevitably, our approaches are quite heterogeneous and range from simple time-series analysis, towards the application of dynamical systems, and generative probabilistic methods. Following some of our initial results, we argue that the selection of methods to study user dynamics strongly depends on the type of collaboration systems under investigation as well as on the research questions that we ask about those systems. More specifically, in this chapter we show our results of (1) the analysis of nonlinearity of user activity time-series, (2) the application of classical dynamical systems to model user motivation and peer influence, (3) a range of scenarios modeling unwanted user behavior and how that behavior influences the evolution of the dynamical systems, (4) a model of growing activity networks with explicit models of activity potential and peer influence. Summarizing, our results indicate that intrinsic user motivation to participate in a collaborative system and peer influence are of primary importance and should be included in the models of the user activity dynamics.