Dynamic Scenarios for Organizations in Infrastructures
Paul Whitney, Katherine Wolf, Sandy Thompson, Garill Coles, Jon Young, David Niesen, Cindy Henderson, Alan Brothers, Ken Jarman, and Stephen Walsh
Executive Summary
Central challenges faced in the modern world are a direct result and side-effects of actions of organizations and individuals. Our interest is in computational, predictive modeling for organizational behavior to help anticipate these challenges and the effects of policy. Computational behavior modeling is increasingly addressed as a research challenge. Computational modeling methods used for behavior modeling include agent- based models, systems dynamics models, and cellular dynamics, among others. This project is focused on representing dynamic human and organizational behaviors. We are investigating Bayes networks, and their integration with other modeling approaches, to represent these evolving phenomena.
This project develops models and methodologies in the context of weapons of mass destruction (WMD) and terrorism threats to infrastructures. Structuring and connecting technical process models (for WMD or terrorism activities, forensics, sociology, psychology) with social indicators (economic, political, cultural, and religious drivers) enables assessment, predication and forensics within a repeatable, computationally sound framework.
Approach
Click on image to enlarge.
The team adopted process models to represent the technical steps involved with WMD and terrorism threats to infrastructures. Such models include fault trees, material and informational flows, and graphical models represented as Bayes nets. A graph representation of a Bayes net for the process of deploying an improvised explosive device is shown. For behavior models, the approaches explored include graphical models and expert-driven models.
Projects
The specific technical advances on which this project focuses this year are 1) the integration of the information from two modeling approaches (systems dynamics and Bayes nets), and 2) developing computational methods for validating and calibrating these models with empirical observations.
This project will deliver mathematical modeling methodology to:
- Demonstrate linkages between technical models and social modelsâwhere these linkages take account of the relative certainties associated with these models, and of the overlapping, but not coincident, focus of the models.
- Combine information across expert-driven models obtained via Judgmental Bootstrapping, dynamic stochastic models (in particular, dynamic Bayesian networks), data, and evidence.
- Provide clear and understandable diagnostics regarding the consistency, predictability, and sensitivities across the models, data, and evidence.
Impacts
This project showcases the impact of modeling for technosocial predictive analysis: both empirical data and expert opinions can be used to calibrate and validate models for technosocial settings. It is a step towards designing and executing large-scale experiments to drive and validate behavior models.