Fixed link

Prisoner's Dilemma

William Poundstone

posted on 04 July 2003

reviewed by Joe McCauley

This is a first rate book for both greenhorns and experts. It starts
with a mini-biography of both von Neumann and the RAND Corporation,
and covers Nash's idea of equilibrium in context. Symmetric and asymmetric games, pure and mixed strategies, zero and nonzero sum games are defined and discussed. More than ten separate games are presented very readably as examples, and the famous Flood-Dresher experiment is also presented, indicating that repeated plays do not lead to (Nash) equilibria (optimization ala neo-classical economic theory is not the answer). We are told about von Neumann's awful advice to the US Government to "nuke" the USSR immediately in the early fifties, and the reader will not be blamed if she calls to mind the "preemptive war in Iraq" based on the Bushie aim/claim to "find weapons of mass destruction". The latter is courtesy of Nietzschean-Bloomian neo-cons like Rumsfeld, Perle, and Wolfowitz.

Poundstone (a former physics student) informs us that war games started with the Prussian military, who invented the mother of all war games, Kriegspiel. Von Neumann played Kriegspiel as a kid, and the game was still popular among Princeton mathematicians in Nash's time. Psychologically interesting is that vN also experieinced a sequence of Cartesian-like dreams in an attempt to prove the conjecture by Hilbert that Turing later showed to be wrong. vN's belief that neo-classical economic theory is completely wrong is also mentioned. There, as opposed to his advice about "preventive war", history has shown him to be dead right.

In the context of the prisoner's dilemma, political conservatives are defined as "defectors" while liberals tend to play "cooperate". The Bush administration provides a working, daily example of playing defect against the rest of the world. See, e.g., Kagan's book "Of Paradise and Power: America vs. Europe in the New World Order" and Kristol's "The War Over Iraq" for details of how the neo-cons always advise playing defect. The book ends with a discussion of the ways in which game theory fails to answer basic socio-economic questions That, in spite of the fact that game theory has been promoted as a basis for mathematizing Darwinian ideas (see the mathematical book "Evolutionary Games and Population Dynamics" by Hofbauer and Sigmund, where Eigen's replicator dynamics is given a game theoretic interpretation).

We learn from the iterated prisoner's dilemma experiment that people tend to cooperate, that Nash equilibria are not realized in practice. To the question: "Why do people cooperate at all?" Poundstone asserts that most situations in real life are more like iterated prisoner's dilemmas. The reader is left to ponder whether this is true or not.

This is an excellent book. I enjoyed it thoroughly and (as greenhorn) benefited from it enormously.