This book is based on the idea that complex organized behavior can emerge out of simple atomic behavior. In physics, simple atoms interact with each other and generate complex behavior like temperature. Such models were never thought to be applicable to humans because people are too complicated and have free will. Buchanan collects work being done by “social physicists” to show that macroscopic human behavior, with hundreds of people interacting, can be modeled by “social atoms” following simple behaviors.
One aspect which made it difficult to construct such models of human behavior is that people learn. Atoms may blindly have the same response every time when presented with a certain situation, but if a response doesn’t work, people will try a different one. We can now use computers to model people’s learning behavior, and see the dynamic equilibrium that results.
For instance, he cites a paper by Brian Arthur discussing the “bar problem”. Arthur liked to go to the El Farol bar near the Santa Fe Institute on Thursday evenings when the Irish musician Gerry Carty played there. But about half the time that he went, the bar was too crowded for him to enjoy the music. So he started guessing when other people wouldn’t be going to try to plan his visits. He realized that others must be doing the same thing.
Being an economist, he decided to model the behavior of attendees. Some attendees would guess that if it were crowded the week before, it wouldn’t be crowded the next week. Others would feel that it would stay crowded. In either case, though, if a strategy weren’t working, people would abandon it and try a different strategy. And when he plotted the results of what bar attendance would look like, it fluctuated wildly around the maximum bar capacity. Why? If somebody hit on a strategy that works for letting them attend when the bar was uncrowded, other people would eventually try that strategy, which would make the bar crowded on those nights, making that winning strategy now a poor strategy. There is no stable equilibrium with “atoms” that learn. Unsurprisingly, the same model approximates the behavior of the stock market, another domain where any successful strategy is immediately copied to the point of no longer being useful.
Another “social atom” model that I really liked was one of a world where there are atoms of four colors: red, blue, green and yellow. When those atoms interact, they can adopt one of four strategies: cooperate with everybody, cooperate with nobody, cooperate with atoms of the same color, cooperate with atoms of different colors. The colors provide no evolutionary advantage whatsoever – they are just an easily identifiable way to tell atoms apart. What happens in a population of randomly colored atoms each starting with a randomly chosen strategy?
Surprisingly, atoms that only cooperate with their own color soon grow to dominate the population. The atoms self-segregate into neighborhoods of like color, and atoms that cooperate with all colors are taken advantage of by all sides before being squeezed out. Axelrod and Hammond call it The Evolution of Ethnocentric Behavior. The atoms in the model have very simple behaviors, and yet generate a result that looks suspiciously like racism. Coincidence?
One last model that I liked took a look at wealth distribution. It started with a population of atoms that each had the same amount of money. The atoms had to preserve a certain amount of money (equivalent to rent and food), and were allowed to invest the rest. Like all investments, the investment was risky, with a chance of losing the money invested, but provided significant returns if successful.
The population ended up with the distribution of wealth we see in the world where a small minority controlled the vast majority of the wealth. The model did not include skill in picking investments or any sort of ability by the wealthy to protect their investments by lobbying, so those weren’t factors. Buchanan explains it as being a result of the power of compound interest, or as the paper authors Bouchaud and Mezard called it, wealth condensation. An atom gets lucky and is successful in their first few investments. Now they can risk a much greater proportion of their wealth in investment because their basic needs are satisfied. Because they risk more money, they gain even more when they are successful, and in a large enough population, somebody gets lucky enough times to end up with all the wealth.
I highly recommend this book. It’s a quick read, but it’s thought provoking. These simple models fascinate me with their ability to generate results that look like people in the aggregate. The models don’t “prove” anything, but it’s instructive to realize that such results as racism or wealth inequality may have nothing to do with people being “evil” or “greedy” – they may simply be a byproduct of people learning successful adaptations within the system. And if such models can be used to predict the behavior of people in the aggregate, is social physics the beginning of Asimov’s psychohistory?