COMPUTATIONAL SOCIAL SCIENCE SEMINAR – The Non-randomness of Random Dynamic Networks: A Baseline Modeling Approach – Joseph Shaheen
COMPUTATIONAL SOCIAL SCIENCE FRIDAY SEMINAR
The Non-randomness of Random Dynamic Networks: A Baseline Modeling Approach
Joseph Shaheen, CSS PhD Student
George Mason University
October 21, 3:00-4:30 p.m.
Center for Social Complexity Suite
Research Hall, 3rd Floor
In an effort to model terrorist (ISIS/ISIL/DAESH) social media networks, a foundational model of dynamic social networks is necessary. Over the last decade a concerted effort has been made from the various lines of inquiry concerned with social networks to describe this model, but many would argue that a comprehensive framework to describe changing, fluctuating social networks continues to elude us.
In this session, I propose the beginning and essential elements of a dynamic, stochastic, undirected, mathematically tractable model built through the use of an agent-based model utilizing simple rules. In other words, I propose a generating function for dynamic social networks.
The model proposed preserves the fundamental elements of random network generation and is in line with the early network paradigm described by Erdos, Renyi and Gilbert in the 1950s, but produces skewed and heavy tailed degree distributions in line with more recent network models such as the Barabasi-Albert and Watts-Strogatz Models.
I propose that the emergence of scale-free and small world properties of a network can be produced using simple randomized rules. I also propose that partial conditional dependence, or the relaxation of the dyadic independence property championed by the statistical branch of network research is not entirely necessary for the generation of real world networks.
Finally, I propose a new modeling element in the form of a coupled set of time-based rules, useful in the modeling and simulation of agent-based models that rely on social networks and furthermore argue its importance and showcase its implications.