Why should we bother with agent-based models? Epstein lists 16 reasons why agent-based models might be worth it; Edmund Chattoe-Brown is a bit more modest with three: combining qualitative and quantitative approaches, transcending linear notions of cause and effect, and building distinctive theories.
Here I want to add another reason: building an agent-based model forces you to think more carefully about the involved mechanisms. Take the following example. A common — usually implicit — model of attitudes towards foreigners assumes that a larger number of foreigners leads to more feelings of threat which lead to more negative attitudes. We can throw this into a regression analysis; it’s all about finding positive correlations. We can ask a few individuals to reflect on the relationship in (in-depth) interviews. But what kind of association should we expect? How much more does more threat actually mean? What is a lot of foreigners? We could go with logical models, as Rein Taagepera suggests, or we could sit down and try to write something in NetLogo — it really forces us to think about what mechanisms are at work, and how we should specify them. Often it also makes us realize how little we know…
Chattoe-Brown, Edmund. 2013. “Why Sociology Should Use Agent Based Modelling.” Sociological Research Online.
Epstein, Joshua M. 2008. “Why Model?” Journal of Artificial Societies and Social Simulation 11(412).
Taagepera, R. 2008. Making Social Sciences More Scientific: The Need for Predictive Models. Oxford: Oxford University Press.