Methodological Investigations in Agent-Based Modelling

• Main Author: Silverman, Eric Stephen
• Published: University of Leeds 2007
• Subjects: 004
• Online Access: http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.485596

Agent-based modelling as a methodology has grown increasingly popular across it number of disciplines in recent decades. In particular the social sciences have begun to use agent-based modelling as a means to examine the development an~ evolution of human society. While agent-based models have been used extensively to model elements ofother highly complex phenomena, such as biological evolution, the study of social phenomena introduces other significant methodological difficulties for the prospective social simula- .. tion researcher. Beginning with a survey of agent-based modelling and its theoretical foundations, this thesis examines the application ofsuch models to the social sciences and the methodological and philosophical problems that can result. This discussion proceeds by first discussing the relatively young field ofArtificial Life, and the philosophical implications - . of the use of models in this discipline, with particular emphasis on methodological issues illuminated by discussion in the broader population biology literature. Secondly, a similar analysis ofmodelling for the social sciences demonstrates the unique problems facing the modeller who seeks to use simulation to understand social phenomena, and examines the applicability of theoretical frameworks from Artificial Life to this discipline. Finally, by developing a contrast between agent-based models in artificial life and social simulation, and by discussing Schelling's well-known residential segregation model as a case-study, a methodological framework for modelling in the social sciences is developed. As this detailed three-part analysis shall demonstrate, while some difficult problems remain despite this developed framework, particularly in the validation of social simulations, agent-based models do indeed allow for innovative and potentially enlightening .examinations of social phenomena. Great care must be taken in particular to ensure that these models remain tractable, and that the inherent theory-dependence of such models is both discussed and dealt with appropriately in the model's design. Such models can also provide a means to study the evolutionary roots of society itself in a manner distinct from conventional empirical study, providing a unique opportunity to develop novel social theory.