Temporal Webs: Analyzing Dynamics with a Trans-Temporal Network Representation

Aaron L Bramson

University of Michigan

TITLE: Temporal Webs: Analyzing Dynamics with a Trans-Temporal Network Representation

ABSTRACT: Computer simulations of complex adaptive systems hold great potential for bet- ter understanding social phenomena, but progress is hampered by our inadequate ability to capture dynamics and analyze outcomes. In order to help alleviate this limitation I demonstrate a technique to measure properties of a system's dynamics by capturing its processes in a trans-temporal network representation. We can apply static network measures to this trans-temporal network to reveal properties of the processes operating within a system. For example, by applying various well-known network measures such as betweenness, clustering, and diameter we can uncover information about the process' path dependence, efficiency, and percolation thresh- old among other things. By developing novel measures specifically designed for this purpose, we can reveal a great many other properties of system dynamics. Using a simple agent-based SIR model as a running example this paper first presents all the necessary components of the trans-temporal network representation and general instructions for creating one. Then a demonstration of collecting and collating data from simulations for use in the trans-temporal network analysis is performed. After that is made available the network algorithms are applied to the generated structure to uncover several key features of the system's dynamics. Finally these measures are validated against the observed dynamics of the original SIR model. Other measures, applicable models, caveats, and extensions are also outlined for future work.

Back to Swarmfest2010_program