Sunday, July 22, 2007

Tinkered Predictions

I wanted to add that I have some problems with game theory as it applies to predicting animal and especially human behavior - and the fault may just be in my own lack of understanding. But if what I understand is correct, game theorists will deliberately ignore aspects of a situation they don't see as relevant to determining the relative payoffs that result from the choice of available strategies, AND they seem to choose only those strategies that offer alternative routes to the end or ultimate "payoff."

But in doing so, they will have assumed that the end payoff is the determinant of all behavior from the outset. And yet that is most often not the case when considering the purposes for which animal (and human) intelligence have actually evolved

For example, they would seem to be completely ignoring long term versus short term payoffs, and the necessity to survive for the short term before concerning oneself at all with long term goals.

Instead, they appear to see short term solutions as ultimately controlled by long term consequences, when in fact the reverse may be more often true.

So why then does game theory seem to work in spite of this apparent error in its general concept? It may be because when the predicted results are inconsistent with other observations, some tinkering occurs - sometimes by adding what are referred to as "equilibrium refinements." Or as at least one adherent has said, one of the most valuable features of game theory is that additional relevant features of the interaction may be integrated into the game-theoretic models.

But when we tinker with the theory of the game in this fashion, is it still the same game? Or are we just applying logical corrections to what must inevitably result from flaws in the initial premise of game theory?