University of Konstanz
Graduiertenkolleg / PhD Program
Computer and Information Science

Graduation Talks

title

Understanding Evolving Social Networks

speaker

Bobo Nick, University Konstanz
Konstanz, Germany

date & place

Wednesday, 23.06.2010, 17:15 h
Room C 252

abstract

Networks are fundamentally dynamic. Mutually dependent, ties are established, gain in strength, may slowly decay or terminate at once. Social networks are especially interesting, since they are typically enriched with a collection of changeable actor attributes, termed \behavior", that co-evolve over time, i.e., the network will influence the dynamics of the behavior (social influence), and the behavior will influence the dynamics of the network (social selection).
Traditonal approaches to modeling and analyzing network formation focus on (re)producing but one static network, though. Namely, random graph models propose likely networks exhibiting empirically observed characteristics, while strategic interaction models produce equilibrium networks that are stable regarding the underlying incentives of involved actors, such as costs and benefits for creating ties.
Attempts to combine random graph models with strategic interaction models and to carry them over onto dynamic networks are in their infancy, with the exception of so called actor oriented models. The latter are based on panel data (i. e., networks observed at two or more discrete points in time), and simulate the unknown evolution between network observations by means of randomly disturbed local best response decisions, involving time-consuming MCMC procedures. Alternatively, information on (the change of) dyadic interaction, may be already given by time-stamped dyadic events, facilitating more intuitive and less expensive ML estimations of underlying dynamics.
My preceding investigations involved network creation games [1], panel vs. event based modeling [3], and gestalt-theory based visual exploration [5]. Moreover, spectral analyses of inherent related networks [2] and foundation and guidelines in use of network centralities [4] were tackled. Future work will concentrate on the combination of the dynamic aspects within these contributions. In the remainder, we provide further background.

[1] Ulrik Brandes, Martin Hoefer, Bobo Nick. Network Creation Games with Disconnected Equilibria. In Proc. 4th Intl. Workshop on Internet and Network Economics (WINE '08), pages 394{401, 2008.

[2] Ulrik Brandes, Jürgen Lerner, Uwe Nagel, Bobo Nick. Structural Trends in Network Ensembles. In Proc. 1th Intl. Workshop on Complex Networks (CompleNet 2009), Studies in Computational Intelligence 207, pages 83{97, 2009.

[3] Ulrik Brandes, Natalie Indlekofer, Jürgen Lerner, Martin Mader, Bobo Nick. Panel or Event Data? Presentation, 6th Conference on Applications of Social Network Analysis (ASNA), 2009.

[4] Ulrik Brandes, Sven Kosub, Bobo Nick. When to use which Centrality Index? Presentation, 6th UK Social Networks Conference, 2010.

[5] Ulrik Brandes, Bobo Nick, Brigitte Rockstroh, Raul Soriano-Hontanilla and Astrid Steffen. Gestaltlines: Extending Sparklines to Multivariate Sequences. Submitted working paper, 2010.