Suchergebnisse

This paper is about the theoretical implications of agent-based modeling exercises. Construction of an agent-based model challenges a social scientist to formalize many concepts and relationships that would have remained implicit or unrecognized. While formalizing these "unimportant" assumptions can be a nuisance, it can also have substantial theoretical payoffs. In order to fill the gaps of the model, the researcher is forced to confront the gaps in the theory that motivated the model in the first place. Using examples drawn from several large political science simulation models, the paper argues that frailty, defined as unpredictability in the behavior of agents, is often required in order to bring closure to the modeling exercise. It is difficult (or impossible) to square the dynamic or aggregate implications of the agent-based model with observations without placing a substantial amount of emphasis on frailty. Hence, the component in behavior that we often treat as "error" in empirical analysis is actually a vital part of the glue that makes the many different moving parts of a social system interact in coherent ways. The example models were developed with the Swarm simulation system (http://www.swarm.org) during the last decade.

This is the publisher's version, also available electronically from https://sites.google.com/a/fspub.unibuc.ro/european-quarterly-of-political-attitudes-and-mentalities/Home ; This paper is about the theoretical implications of agent-based modeling exercises. Construction of an agent-based model challenges a social scientist to formalize many concepts and relationships that would have remained implicit or unrecognized. While formalizing these "unimportant" assumptions can be a nuisance, it can also have substantial theoretical payoffs. In order to fill the gaps of the model, the researcher is forced to confront the gaps in the theory that motivated the model in the first place. Using examples drawn from several large political science simulation models, the paper argues that frailty, defined as unpredictability in the behavior of agents, is often required in order to bring closure to the modeling exercise. It is difficult (or impossible) to square the dynamic or aggregate implications of the agent-based model with observations without placing a substantial amount of emphasis on frailty. Hence, the component in behavior that we often treat as "error" in empirical analysis is actually a vital part of the glue that makes the many different moving parts of a social system interact in coherent ways. The example models were developed with the Swarm simulation system (http://www.swarm.org) during the last decade.

This paper is about the theoretical implications of agent-based modeling exercises. Construction of
an agent-based model challenges a social scientist to formalize many concepts and relationships
that would have remained implicit or unrecognized. While formalizing these "unimportant"
assumptions can be a nuisance, it can also have substantial theoretical payoffs. In order to fill the
gaps of the model, the researcher is forced to confront the gaps in the theory that motivated the
model in the first place. Using examples drawn from several large political science simulation
models, the paper argues that frailty, defined as unpredictability in the behavior of agents, is often
required in order to bring closure to the modeling exercise. It is difficult (or impossible) to square
the dynamic or aggregate implications of the agent-based model with observations without placing
a substantial amount of emphasis on frailty. Hence, the component in behavior that we often treat
as "error" in empirical analysis is actually a vital part of the glue that makes the many different
moving parts of a social system interact in coherent ways. The example models were developed
with the Swarm simulation system (http://www.swarm.org) during the last decade.

Simulation research has made some notable contributions in political science. This article describes a variety of simulation projects and points out the strengths and weaknesses of simulation in contrast with other methods of research. The strength of simulation research is that it is naturally suited to modeling projects that include a large number of autonomous, interacting agents. Statistical and formal methods of analysis have made contributions in these areas, but there is good reason to believe that simulation can go further. This point is explored in several contexts, including social choice theory, individual-level simulation models, international relations, the prisoner's dilemma game, and more general agent-based models of multiperson interaction.