>But doesn't the prisoner's dilemma depend on the >prisoner's (a) not
>communicating and (b) a fortiori, not going through a >process of
>struggle and self-criticism to work out the principles of >collaboration?
>That seems to rather separate from the world of actual >human experience,
>either today or 60,000 BP.
Game theory is very useful for things like, for example, designing auctions aimed at making people pay top dollar for mobile telephone licenses. Most of the practical uses are in this sort of area. (thinking about this, it relates back to an earlier conversation about the usefulness of maths in economics; auction design is probably the only area of economics where it's absolutely impossible to understand it at all if you don't understand it mathematically).
More generally, there are real-world situations where the prisoners' dilemma results are valid. Also note that the PD doesn't really need the non-communication stipulation. you can let the two prisoners communicate as much as you like right up until the instant that they have to decide; they can make as many promises as they like that they'll co-operate, but these promises are not credible if you assume that the prisoners are moderately self-interested (ie, self-interested enough that there is some number of years in jail which they would sell out a buddy to avoid, which isn't all that restrictive or unrealistic).
The argument is the basis of the concept of a Nash equilibrium. A Nash equilibrium is defined (loosely) as a set of strategies under which each player is doing as well as he can given the other players' strategies. The analogy to a Pareto optimum is close; you can't be at a Nash equilibrium if any player could do better for himself by changing his (taking other players' strategies as given). So (co-operate, co-operate) is not a Nash equilbrium because either player can improve on it by moving their strategy to "defect", while (defect, defect) is a Nash equilibrium because at that point, neither player can improve on it by moving their strategy to "co-operate".
I'd also note that Nash equilibrium is a much less compelling and interesting property of strategy sets than it is commonly conceived to be (for example, by the Nobel committee). There is no force pulling you toward a Nash equilibrium in the way that there is toward a competitive equilibrium or a no-arbitrage condition. It's just the mathematical tractability of the Nash condition in a few cases of interest, plus its superficial resemblance to the Pareto condition, that makes people try to identify Nash equilibria as necessary rational choice outcomes, which they aren't.
dd
Get Your Free Email at http://www.al-islam.com