Applying Agent-based Simulation Software and Sociological Theory to Decision Making


The programming of the simulation is one of the core elements of the project: Translating the social scientific models into a simulation software allows their predictions, based on data inputted by the area experts, to be retrospectively tested. Furthermore, as the simulation software is provided to users, developing an intuitive and powerful interface to run the simulation and easily interpret its results, the programming of the interface is another core aspect.

The programming of the simulation consists of two main parts:

Design of Simulation System Architecture
One is to translate the theoretical models provided by the social scientists into computer algorithms. A constant dialogue between the social and computer scientists ensures that the models are actually implementable, with modifications and simplifications being suggested, while at the same time faithfully representing the core ideas underlying the theories. For example, the implementation of game-theoretical coalition models requires to develop and adapt algorithms to compute equilibria for the specified games used in the simulation. When exact analytical solutions are not possible, likely outcomes need to be computed. Another aspect of this task is the cooperation with the area experts to clarify the data needs as well as the structure the data is inputted into the simulation (also see here).

The final version of the simulation will incorporate four major models - coherence, coalition, action and state of the world - that are interlinked as illustrated below:

While the coalition model describes how the individuals align themselves to form ethnic groups based on ascriptive characteristics, the action model is concerned with how the groups form relate to each other - with different degrees of conflict or cooperation. The coherence model (also see here) is the major innovation in comparison to other game-theoretical models: By modeling explicitly how individuals change their beliefs and preferences in light of prior experience (to make it appear rational in retrospect) and belief systems present in the society, cultural change becomes endogenous - allowing to make predictions how coalitions and behavior will change over time. Finally, the state of the world model represents the structural constraints on the actors and how they change due to the outcomes of their interactions (the events).

Design, Testing, and Enhancement of User Interface
Making the simulation easily usable as well as flexible is the second major task. Expert knowledge of human-computer interaction is applied to ensure that the system provides simple and convenient ways for users to specify the political scenarios they wish to examine, use the models they find most appropriate and obtain the depth of information that they require, from simple deterministic predictions to complex contingency diagrams. The data output allows users to trace the impact of different factors on the final predictions, giving them an insight into what variables to influence to affect outcomes in the real worlds. Using different of the available models, furthermore, allows users to test the robustness of these recommendations when making different assumptions about how various variables are interrelated. To allow each user to use the software optimally according to his/her needs, techniques for modeling classes of users to customize interactions will be implemented.



Faculty: Gürdal Arlsan, Sun-Ki Chai, David Chin, Keisuke Nakao, Scott Robertson

Research Assistants: Pantaleo DeCandia, Evan Yazawa, Robert Puckett