Intro to Computer Modeling for Social Scientists

Majority Game on a Ring To start we have agents living in a one-dimensional world. They have one of two possible values and switch values to the dominant value in either a plurality or a specified majority of the other agents within it's radius.

Majority Game on a Torus We now have the same model as above, but on a two-dimensional grid with wrapping edges.

Majoity Game with Two features Building on the previous model, we now add another feature (also with two possible values). The values of each feature are independent and each still changes according to the same majority/plurality rule described in the first model. The four possible agent types are each represented as different colors.

Majority Game with Multiple features Building on the previous model, we now add up to six different features (each with two possible values). The values of each feature are independent and each still changes according to the same majority/plurality rule described in the first model.

Majority Game with Multiple Values Going back to the original 2D model for the moment, our agents again have just one feature to change. But now the feature can have up to fourteen different values.

Multiple Features and Multiple Values By combining the two previous models, our agents now have up to six independent features, each one capable of taking one of up to eight values.

Features, Values, and Self-Consistency We now add to the previous model by incorporating a self-consistency behavior to the agents. In addition to conforming to neighbors' types, agents now also seek to set all the values of their various features to be equal. In each turn the agents use either the majority/plurality rule or the self-consistency rule.

Learning through Social Communication 1 Agents live a world where they must choose action A or B, and there is a certain probability that either A or B will be the winning choice. Can agents as a community determine which action tends to yield the better result?

Learning through Social Communication 2 Now there are two regions in the world with different probabilities of success. Can agents learn where action A is better and where action B is better?

Learning through Social Communication 3 Only a small section of the world now has the different probability. What factors determine the ability of agents in those spaces to learn that the outcome of their actions is different?

Banks&Fires Banks are either risky or safe. Risky banks get a higher payoff, but they might get busted. Watch the bank busting spread like wildfire through the risky banks. Do banks develop patterns to 'protect' risky banks from these 'fires' or not? Is there a steady ratio of risky:safe banks?

Segregation Model and Enhancements Schelling's segregation model offers a fantastic springboard into the world of agent-based modeling. The model is conceptually rich and the mechanism is simple so it is easy to adapt and extend in many directions. This version of the model is the same as the one in the Netlogo Model Library, but with a lot more comments describing code and a more detailed information tab describing what's going on and some recommended extensions. It was put together for the EITM workshop at the University of Michigan and presented (by me) on July 8th, 2009.