Toward a Better Understanding of Evolving Social Networks

Abstract

In this thesis, we provide methodological contributions toward a better understanding of evolving social networks. In particular, we emphasize the importance and consequences of holistic approaches that exploit all available data and detailed information embedded in structure and time. Multiple case studies demonstrate the added value in doing so and reveal deficiencies in previous interpretations that resulted from more aggregate data views.Initially, we introduce the concept of Simmelian backbones and a corresponding notion of triadic cohesion to reveal primary actor groups in social networks. Next, we discuss a novel visualization technique, gestaltlines, that allows to represent asymmetric longitudinal network data within a single gestaltmatrix. Then, we address the fundamental question to which degree tie-change events in evolving social networks depend on each other and propose a general framework that allows to quantify the implications of assuming conditional independence. Finally, we demonstrate how social vector clocks can be used to exploit indirect information flow based on the ordering and spacing of communication events and accurately predict future interactions.