http://plato.stanford.edu/entries/game-theory/#Behav

Although this article is long, I will focus on the section titled, Game Theory and Behavioral Evidence. One of the major criticisms of game theory is that it does not make specific enough predictions and that it does not adequately predict behavior. The assumption that the players be perfectly rational is argued by some to be unrealistic, and while it works well for the mathematical model of game theory, it is not based on behavioral evidence. However, it has been shown empirically that game theory can be used to predict the behavior of lower order animals with great accuracy. Why the discrepancy between low order animals and higher order ones, including humans? It is believed that since low order animals lack the ability to learn and create societies, traditional game theory models them very well. In light of this, other variations on game theory such as evolutionary game theory have been developed. Evolutionary game theory allows the players to develop preferences as they play the game. This implies that the game is being played multiple times, as opposed to traditional game theory in which the game is played once, and each time is considered independently of the other times. Unfortunately, even evolutionary game theory has a problem. Humans tend to, sometimes arbitrarily, change preference. When this happens, the model breaks down.

In researching game theory, I came across various people who believe that game theory is dead. They believe that game theory’s vague predictions and assumptions are fatally flawed, and therefore, game theory should be replaced with other theories. Clearly there is a place for game theory in this world as it models low order animal behavior well. How difficult would it be to extend game theory to include higher order animals? During our lectures on game theory, we were instructed that there is no reason why a player’s goal cannot include things about other players. Is it possible to add some irrationality to a player? Game theory is a model, we must therefore take into account the simplicity of the model when evaluating it. If another model makes predictions as good as game theory but it is simpler, then that model is better. The question then becomes, will changing game theory to more adequately predict human behavior destroy its appeal as a relatively simple model? Are there models that more accurately predict human behavior, and are they simple enough to be useful?