Christopher L. Barrett Webcast : Read the Summary

Complexity Science, Simulation and Public Health

We will describe a very large, detailed, agent-based simulation and analysis of epidemics such as pandemic influenza. The interplay between individual behaviors, societal and built structure, and contagious disease dynamics creates epidemics. Public health ande individual responses to epidemics are timely and classic examples of decision-making in complex socio-technical systems and serve to illustrate general problems addressed by complexity science in both theoretical and practical settings.

In socio-technical systems, physical-, social-, psychological-, and biological processes and built infrastructure are coupled and create co-evolving systems that are not understandable in any of the "languages" dedicated to these individual aspects. Complexity science seeks to provide a transdisciplinary setting, lawfulness, and language for these kinds of circumstances. There are numerous foundations of this nascent integrative science; a reasonably encompassing minimal list includes chaotic systems with their sensitivity to initial conditions, topological mixing and dense phase spaces, interaction-based concepts involving composition of local dynamical systems and networks, computational methods, especially simulation, and the related problems of irreducible representation and issues of algorithmic complexity and algorithmic information theory.

When confronted with a complex system-of-systems, it is now common to employ so-called agent-based methods and create a computer simulation. The simulacra interact and generate the composite dynamical properties of a system; its composed dynamical properties are caused by interactions among the agents. The idea of explanation here is constructive and related to the ability of a composition of certain elements to computationally generate a property. Moreover, since individual agents are constructed of specified relations among constituent symbolic components and instantiated computationally, the position taken is apparently that the causal potency of a thing is not inherently related to its embodiment, but to the relations expressed by an embodiment. Although this is hardly a new view, it is rather extreme and involves cherished issues in physicalism and the representational theory of mind, among other things. Nevertheless, interaction-oriented methods such as agent based simulation frames systems problems in an integrative complexity science. The local dynamical systems interact and define a network which can be characterized, the network properties influence the dynamical behavior of the local systems defining constraint and co-evolution, these dynamics can be understood in terms of the properties of their phase space, the algorithmic properties of the system can be assessed, etc.

We will take as an example a high performance micro-simulation for epidemics that scales to hundreds of millions of individuals, its rationale, use and some analysis.