Tuesday, October 2, 2012

Settlers of Catan is a Great Social Science Model

Brad Pummer at Wonkblog recently had a great article describing the potential of MMO computer games to serve as laboratories for economics researchers (epidemiologists too!). While video games can test theories by facilitating controlled experiments, other leisure activities like board games can help us interpret social science theories.

A theory (or model) is basically just a simplified version of the world, one that sorts details that are important for explaining a phenomenon from unimportant details. Among social scientists, there's a big cleavage between those who believe social phenomena is best explained by structural and contextual forces, and those who believe agency-based explanations are the way to go. This cleavage, which exists on a continuum, is roughly analogous to the philosophical question concerning the degree to which humans posses free will. A great example of this rift that's been in the news recently is the media coverage of the presidency: are presidents successful because they happen to be in office when the context is favorable (i.e. economic cycles, historical patterns, demographics, etc.), or do they succeed because they made the right decisions and possess good leadership qualities.

Settlers of Catan is an intellectually fascinating board game because it demonstrates a broad range of heady concepts from economics, political science, psychology, game theory, and probability theory. The basic game consists of 3-6 players competing to grow their small economies by constructing roads, settlements, cities, and various random-payoff and positional goods. Players collect tradeable resources based on the geographic location of their infrastructure: the game board consists of randomly-distributed tiles corresponding to resource-producing biomes. Each tile's productivity also varies randomly using a probability distribution.

Settlers is great because it shows how incredibly difficult it is to develop an adequate social science theory that actually has predictive accuracy. Explaining the outcome of any particular match is tough: did the victor win because of favorable initial geographic conditions? Did they have perfect strategy and bargaining logic? Perhaps they won because their opponents were irrational? Maybe they would have lost 9 times out of 10, but they just got lucky? Because it's impossible to re-play games with identical initial conditions and dice rolls, controlling for variables is impossible, meaning we'll never know for sure. Sometimes a specific variable will be so extreme--godly spot placement, miraculous dice rolling, brain-dead opponents--that predicting the winner is possible. But those situations only happen occasionally; most of the time variables are moderate and so isolating any one cause is tricky. The difficulty in establishing causation in a simple German board game with a limited number of variables shows how hard social science is, which studies the messy real world.

No comments:

Post a Comment