Re: game theory

["Devine, James" <jdevine@xxxxxxx>]

JKS writes:>I think it is bizarre to ask whether game theory is
evil, or (as Jim Divine suggests) whether it makes you
crazy, or comes from paranoid schizophrenia, or
something like that.<

I never made such a suggestion and never would. Please quote me where I said anything like that. (I thought I was giving a measured defense of game theory, which was meant to be light-hearted. If you don't like my jokes, please tell me.) and it's dEvine. 

excuse my bad mood. I _hate_ being misquoted and/or misunderstood. Let me repeat my conclusions: game theory (GT) doesn't have to assume that people are inhuman machines or ignore the role of culture. One has to be very careful applying it. Nash equilibrium is useless except as an ideal standard to compare the world to. 

>  Nash went mad, but you can't
argue with his maths.<

you can easily argue about the applicability of the math. Math doesn't correspond to reality; it only represents the abstract dimension.

By the way, Nash is currently deemed sane. And his sanity or insanity has nothing to do with the validity of the Nash equilibrium concept or of game theory.

> The prisoner's dilemma and the
Nash equilibrium are two of the very greatest -- maybe
the very greatest -- results in social theory in the
20th century.<

these are assertions, not proof, counselor. But of course, when I was a juror the judge told us that the opening statement was not evidence. 

I would agree that the PD was an important innovation, though the related collective ("public") goods problem is more important empirically (since it corresponds to a many-firmed market, etc.) I'm not going to prove that, since I lack the time, so I'll leave it on the level of advertising sloganeering. There's a chapter or two in Hargreaves Heap on it. 

On the other hand, Nash equilibrium is an idealized state similar to Rational Expectations equilibrium (as I said). Both are akin to the Platonic forms in terms of empirical validity.

>GT is an extremely powerful and beautiful set of
mathematical tools that has a wide application in
thinking about society, particularly in competitive
situations, which of course is really important if you
are an economist whose job it is to understand
capitalism, or a political scientist who wants to
understand international/world politics. <

some would agree. some not. BTW, I don't think "beauty" is a very important criterion in an effort to understand the world. After all, the world is pretty damned ugly.

> It involves
abstractions and idaelizations, of course,a nd people
are really like that -- duh -- in this theory, as
someone who used an important precursor of it once
said, men are mere bearers of social relations.< 

I don't understand this. Are you saying that Marx was a precursor of the GTists? He understood market competition pretty well, following Adam Smith, but that's different. His emphasis was empirical, not with the building of abstract models. Though he too abstract, it's clear in his theoretical work that he wanted to be as concrete (empirical) as possible.  

The idea that people are bearers of social relations should be understood as saying that individual goals, ideologies, etc. and the effects of their actions are shaped by their positions in the social structure. As I noted, GT doesn't (usually?) take individual tastes, ideologies, etc. as endogenously determined by the social structure or game. I do think that one can use GT to understand some stuff in Marx. But it would be a mistake to reduce his though to GT, making him a minor pre-von Neumannian. Is that what you're doing? 

>People
are working are the complsxifications [???]  with, in my
area, e.g., behavioral law and economics. But that it
still poorly understood and litle developed, and will
never have the elegant simplicity of game theory.<

part of the problem with GT for many is its elegant simplicity. The world ain't elegant or simple, alas. How can one apply such a theory to the world? very carefully. Unfortunately, care is scarce among social scientists at a theoretical level above the technical, mathematical, details. Social scientists are pretty careful about using a theory correctly in the purely logical sense but often not so about when it should be applied in the empirical world. That's why you see very smart people like von Neumann using GT to advocate preemptive attacks on the USSR during the 1950s. 

>The theory, like most theories, rests on assumptions
taht are technically false. But it is powerful and
predictive theory, and <

it's not predictive if there are multiple equilibria, which is true of many if not most games. Even if a game has a unique equilibrium, that is a result of the assumption, including the equilibrium concept used. (Nash equilib. isn't the only one.) So one can easily tailor one's model to fit the pre-existing data, so that there's no real prediction. 

Most GT is really a matter of _understanding_, not predicting. There's nothing wrong with that. Most social science is about understanding. 

>the problem with it from the
left is just that it should be allowed to become
ideology, that is assumed to be about invariant human
nature in all times and places regardless of
circumstances.<

are you agreeing with "the left"? 

>Besides the theory is useful to the left in lots of
ways. For example, the PD is a real kick in the teeth
to the Panglossian assumptions of Gen Equil Theory,
which says that rational self interested actors will
give us The Best Of All Possible Worlds....<

The PD isn't needed for that. A simple theory of externalities will do. Or imperfect information.

And the PD knife cuts both ways.  Ignoring externalities and information problems such as adverse selection, the advocates of free markets can say that it's like the capitalist firms are the prisoners and the consumers are the jailer. The capitalist firms defect (as in one standard PD result) and the consumers benefit. Hail capitalism! there is only one God and He is the Invisible Hand! there is only one Prophet and he is Milton Friedman! 

BTW, as I said before, this standard PD result doesn't always happen. It's amazing how often the prisoners end up cooperating with each other. In the collective goods problem (with a large number of "players") this cooperation is much more difficult. This suggests the problem of competition isn't the PD as much as the number of players. Even with a large number of players, cooperation can be pulled off, as with the persistence of many traditions. 

On the PD, I wish economists would spend more time watching cop shows (or better, engaging in empirical research). In addition to separating the perps to get them to talk, there are other ways to achieve the goal (even ruling out Rumsfeld's methods). For example, there's the Mutt & Jeff (good cop/bad cop) routine that does the mindf*ck on the perp, trying to confuse him or her. Alas, most economists don't study the internal conflicts in human's souls. Jon Elster has a good book on it, but it's rare outside the Freudians.

Jim Devine