In the Theory for Anthropology Group, we have started a discussion of model thinking, i.e., using models, which, as statistician George E.P. Box put it "are always wrong but sometimes useful" to think about social and cultural processes. The discussion was stimulated by my discovery of Scott Page's free University of Michigan online course on models. And now, thanks to Page, I have made another discovery. Discussing the classic Schelling Spatial Segregation model, Page uses Netlogo, free, open-source software developed at Northwestern University. Yesterday, I downloaded Netlogo for Mac OSX (it is also available for Windows and Linux) and began playing with the models in its model library. I was blown away. This post is a shameless plug to joining me in exploring what Netlogo and model thinking have to offer.
Consider, for example, the Schelling Spatial Segregation model mentioned above. As Scott Page notes, it seems obvious to most of us that segregation results from racist attitudes, including in particular, discomfort at living in the same neighborhood as people unlike ourselves. Developed by economist Thomas C. Schelling, the Schelling Spatial Segregation Model suggests that things are not so simple. The basic idea is to start with actors on a grid (a checkerboard is a good example). Each actor occupies a space surrounded by eight other spaces. If actors belong to two types (red and black checkers, for instance), the proportion of other spaces occupied by members of the actors of a given type ranges from 0/8=0.0% to 8/8=100%. The question is what happens if you start with a large, randomly mixed population and select a threshold below which actors become uncomfortable and move away? What the model indicates is that spatial segregation appears at around 30% (actors who are comfortable with 70% of their neighbors being different from themselves). It crystallizes in large, specially segregated blocks when actors insist that they belong to the majority (51%). Nearer to the extremes, however, spatial segregation disappears. When I played with the model, I found that at 10%, only a small amount of sorting led to an equilibrium at which all of the actors were happy. Above 90%, there was no equilibrium and no spatial segregation either. With almost all of the actors unhappy, the result was an endless chaos. Actors moved all the time but could find no spot to settle among actors like themselves.
It is one thing to read a discussion of this model (I own and have read Schelling's Micromotives and Macrobehavior), it is another thing entirely to have the model to play with and see what happens when you tweak its settings.