Here are a few models I have developed. I used the first two to study the dynamics of interest articulation in large groups; the remainder are "toy" models that illustrate some basic social scientific principles. You can view and use the following models through your web browser. Please contact me for the source code.
Preference Cycling in International Organizations: An Agent-Based Model of Two-Level Games
International Studies Quarterly, vol. 52, no. 2. June 2008. pp. 363-382
How do states solve n-choice cooperation problems? Although classical game theory offers useful insights into how states achieve cooperation, its focus on equilibria means that it is largely silent on the processes of negotiation, particularly when multiple solutions to a cooperation problem exist. To explore the path histories of international cooperation, this article uses an agent-based model to explore how states solve a coordination game with three equilibria. To further enrich the analysis, the model explores how domestic constraints may affect coordination games. Such state-level constraints may present negotiators with considerable informational costs and high barriers to understanding each other's preferences. The model finds that the process of communicating preferences can greatly affect the prospects for cooperation, independent of distributional consequences and concerns about cheating. It also finds that, contrary to collective goods theory, a large number of states coordinate a solution to the problem easier than small groups. This is because larger groups create denser networks that communicate preferences more efficiently. The agent-based model thus illustrates how simulation as a method can challenge our extant thinking about cooperation among nation-states. These findings have implications for research on interstate coordination, since it suggests transnational linkages may help states overcome information barriers that frustrate agreements. Click here to try the model.
Voting Complexity and Electoral Outcomes: An Agent-Based Model of Condorcet Social Choice Problems
Prepared for the Workshop on Complexity and Policy Analysis.
Co-sponsored by the Department of Government, The University College Cork; The Institute for the Study of Coherence and Emergence; and Penn State University-Harrisonburg.
Cork, Ireland: June 22-24, 2005.
To read the paper, click here.
This paper presents an agent-based model of elections in which there are three possible choices. Previous agent-based models of voting typically have examined the simple two-choice election, while previous formal treatments of three-choice problems have necessitated strict axioms that are both theoretically restrictive and empirically unfounded. The model presented here corrects both of these limitations. It illustrates how elections may produce socially undesirable outcomes including frequent reversals of collective decisions and election winners that a majority of voters do not prefer. The model endogenizes two key processes in voter preference formation: the influence of local social networks and the effect of collective choice on an individual’s strength of preferences for subsequent elections. I test these dynamics using a genetic algorithm to explore the model’s parameter space; this “active nonlinear test” (Miller 1998) illustrates the conditions under which preference cycles and Pareto suboptimal outcomes emerge. I use as an empirical referent the well-documented practice of tactical voting in British parliamentary elections. I show how tactical voting favors incumbents, and predict that the Labour Party will enjoy a surprising degree of success in the general elections on May 5, 2005. Click here to try the model.
Segregation in Norfolk: Integrating ABM and GIS technologies to estimate residential patterns
(with Joshua G. Behr)
In this demonstration, we present an agent-based simulation of urban settlement based on Thomas Schelling’s segregation game. By incorporating geographic information systems (GIS) land-use data for Norfolk, Virginia into the model, we illustrate how patterns of segregation may arise over time from three simple independent variables: land-use zoning, market values of residential land, and individual levels of tolerance. The use of a formal dynamic simulative model allows us to test several values of individual tolerance to determine the conditions under which urban segregation arises. We then test the predictive quality of our model using demographic data from Norfolk. The results illustrate that even parsimonious formal models of urban settlement can explain much of the observed patterns of segregation in today’s cities. While the literature on residential segregation is replete with case studies presenting city demographic and land use maps, there remains a paucity of systematic analysis across cities. That is, absent is a standard yardstick to measure residential patterns so meaningful comparisons may be made. Our use of the computer-simulation techniques of agent-based modeling allows us to construct such a standardized yardstick based on an algorithmic model of residential settlement patterns. Click here to try the model.
The metaphor refers to the English practice of the middle ages in which villages provided a common grazing area for livestock. If each member of the village tried to maximize his or her wealth by grazing as much livestock as possible, the commons would quickly be depleted. The village as a whole ends up worse off. But if villagers forego short-term gain from the commons, they can sustain the resource over the long run.
This model replicates the tragedy of the commons by creating a grassy common area and a small population of cattle farmers. Click here to try the model.
This is a simple model of an influenza epidemic. It seeks to investigate how much prior immunization of a population is optimal from a policy standpoint. Given the competing costs of innoculation versus lost productivity, the model allows the user to investigate which policy choices and assumptions about the virus minimize the loss of worker productivity. Click here to try the model.
This is an agent-based model of the log map, a well-known formula for the demonstration of chaotic dynamics in nonlinear systems. It illusrates the "butterfly effect" or how miniscule differences in initial conditions can create large divergences in the path evolution of a nonlinear system. (See Lorenz 1963.) Nonlinear dynamics and the butterfly effect have profound implications for the social sciences for two reasons. One is our traditional use of statistical methods works only for linear and log-linear systems; they are not useful for nonlinear systems. Second, the log map illustrates that even a simple deterministic system may produce dynamics and values that appear "random."
The model adapts Kiel and Elliott's (1997) demonstration of nonlinear dynamics using the log map and a spreadsheet. Click here to try the model.
This is a simple model that illustrates the power of social networks. In Tipping Point: How Little Things Can Make a Big Difference (2001), Malcolm Gladwell discusses how knowledgeable people (mavens) can use individuals with wide social circles (connectors) to disseminate information quickly and effectively. This process can produce "social epidemics," whether they are trends and fads in fashion or moral outrage on the internet. The metaphor also seems applicable to political campaigns that seek to target "opinion leaders." In this simple model, two mavens (red and green) seek to tip the social choice toward their opinion. Using one of three strategies, the mavens move around this virtual society encouraging individuals to adopt their preferred social choice. Notice how when a red maven focuses only on connectors, for example, the social choice quickly tips toward red. The model thus produces something like a social epidemic.