Abstract

This work is about some of the philosophical problems that arise in the context of agent-based modeling. Agent-based models differ from other modeling techniques primarily in terms of the way they represent their target system. They are a comparatively novel kind of computer simulation and are used to investigate complex systems. In recent decades, agent-based models have been used in disciplines such as economics, archaeology, sociology, and ecology. Typical target systems of agent-based models include markets, foragers, opinion dynamics, and predator-prey interactions. This thesis contributes to the research on agent-based modeling by situating agent-based models in the broader context of scientific modeling and by resolving several important philosophical problems that arise in relation to agent-based modeling research. Those problems mainly concern the character of agent-based models and their capacity to enhance our understanding of social phenomena. The first part of this thesis contributes to ongoing discussions in the philosophy of science about the status of scientific models by providing new perspectives on the function of models, the difference between computer simulations and other modeling techniques, and by analyzing the use of models across several disciplines. The second part of the thesis is concerned with the use of agent-based models in scientific research practice. Several case studies of agent-based modeling projects in archaeology and economics are used to illustrate the role of agent-based models in research practice. These case studies include large-scale projects like the Village Ecodynamics Project and Eurace@Unibi project and smaller models, such as Jeffrey Brantingham’s neutral model. The relationships between agent-based models and complexity science and other social sciences are discussed in order to understand how this modeling type can be used to understand and explain phenomena. Although agent-based models are highly idealized and often do not provide correct explanations, I argue that they can, nonetheless, play a useful role in explanatory research, such as when they are used to identify the difference-makers of phenomena or to rule out potential explanations.