Joseph M. Whitmeyer

Advancing Social Science through Agent-Based Modeling

We discuss three issues in this chapter. First, we argue that agent-based models are necessary in social science. This also entails the contention that the more common ways of doing social science—the established methods of statistical analysis and the newer technique of data mining—are insufficient. Second, we advocate building into ABMs in social science the capability of choosing and combining theories or models. We argue that this will improve and accelerate theory advancement. Last, we discuss how to model the individual in ABMs in order to facilitate incorporation of theories choice and their combination.

The ACSES Model of Afghanistan: The Model Operation, Synthetic Population, Calibration, and Surprises

In this chapter, we continue the description of the ACSES model of citizen allegiance in the presence of an insurgency in Afghanistan. Here, we show how the social theories and actor models presented in the previous chapter were implemented for the specific case of the Taliban insurgency in Afghanistan in an agent-based model. We address four main topics. First, we present and narrate a flowchart of the model. Second, we describe creation of the synthetic population, so that the model would have some fidelity and relevance to the target scenario. Third, we discuss how we calibrated the model using real-world data, concentrating on the genetic algorithm technique that was most fruitful. Finally, we discuss some instances of surprise that we found using the ACSES model, as an example of some of the benefits of this approach.

The ACSES Model of Afghanistan: Introduction and Social Theories

In this chapter we present an example of the principles and approaches we have described to this point. The example in question is a model of population allegiance in Afghanistan in the face of the Taliban insurgency, and effort by the Afghan government and coalition forces to combat the insurgency. DARPA supported our development of this model as a pilot project, which we refer to it as the ACSES (“Actionable Capability for Social and Economic Systems”) model.

Towards a Characterization and Systematic Evaluation Framework for Theories and Models of Human, Social, Behavioral, and Cultural Processes within Agent-Based Models

An article in Science magazine (Bhattacharjee, 2007) discussed how the U.S. military is interested in enlisting the help of multidisciplinary scientific experts to better understand “how local populations behave in a war zone.” The article mentioned the “Human Social Culture Behavior Modeling” (HSBC) program at U.S. Department of Defense and indicated, through a few anecdotal examples, the types of prior research emanating from multidisciplinary fields that may be considered the state of the art.

Running the System: Prediction and Intervention

Among the most valuable uses of simulation models is making predictions of how a real-world situation will develop either on its own or if interventions are carried out. Decision makers are often confronted with developing situations for which being able to predict the outcome with a sufficient degree of likelihood would be extremely useful, such as forecasting the development and course of a disease or predicting the course of an economy.

Lessons Learned from the ACSES Model

A number of potentially useful lessons emerged from our construction of the ACSES simulation. We list these, with some elaboration, below. Lessons 1-6 concern how to embed social theories. This was the most novel aspect of our simulation exercise and this is where we have the most to say about the lessons learned. Lessons 7 and 8 are about implications of the ACSES model for the technical construction of such models. Lessons 9-11 concern surprise and emergence, that is, some of the more interesting outcomes that agent-based models can produce.