Author: Don Birx
A B S T R A C T
At the Physical Science Laboratory and departments throughout New Mexico State University, we are increasingly involved in developing techniques and tools that work in an agent modeling environment, and mine, integrate, and evaluate the patterns of information in automated ways to deduce the key elements or invariants that enable one to predict future likely outcomes and control for a more desirable result. Other tools, such as formal concept analysis, combinatorics, and dynamics analysis enhance the ability of researchers to clarify and understand relations from incomplete data in a visually compelling format.
Understanding the impact of collective decision processes in a way that leads to an understanding of the impetus for the decisions that are made, requires the ability to model human and computer decision processes as well as symbolic information flow and influences to a level here-to-fore unimagined. Fortunately we live in an age when the science of human influence, interaction, and decision-making, is developing to the point that it may be possible to be understood and even predicted. In the past, modeling was based on the premise of high-level analytical descriptions of behavior and constraints in a brittle and often linearized probabilistic or stochastic format. This type of modeling cannot predict novel events or emergent behaviors, which clearly is needed in the world of the 21st century.
Today it might be possible to construct agent-based models that can realistically simulate a sequence of decision processes and emergent events. This process mines intelligently collected information, which is inputted into an agent modeling environment, and uses a set of tools that includes non-linear dynamics analysis, modal logics, and combinatorics, which structure underlying interactions and discover critical nodes of communication and influence. When fused, this information can yield the critical hidden and underlying patterns. Complex series of events may then be reduced to key generators and chaos and complex systems approaches are utilized to describe dynamic attractors for an understanding of how to influence future events with a minimum resource investment.
In the last few years, we have been building successive generations of models and tools that are targeted at complex systems of individuals with emerging characteristics and behaviors. The latest of these models focuses on the developing terrorist environment and involves the emergence of a network that targets both a nuclear and biological attack. It is currently under construction, and is open to the community for contribution and analysis. It will utilize a fuzzy inference engine with evolutionary programming for an agent’s decision processes and is based on the latest models of psychological behavior (Reflexive control, Reversal Theory etc.) for adaptive realistic modeling of human decision processes, control, and prediction. The presentation this Monday will detail some of the background techniques we have been exploring in the context of this open model environment.