Part biography (of John von Neumann), part history (of the RAND corporation and landmark moments of the cold war, with cameos from Bertrand Russell and others), and part (about half) pure and applied game theory, this is an entertaining and absorbing book.
The title comes from the best-known and most baffling of two-player, two-choice confrontations (there are three others which formally rank as dilemmas--where individual gain opposes public good--the next best known being "chicken").
Any in-depth treatment of these conceptions normally buries all but the geekiest reader in payoff matrices and jargon such as: symmetric, asymmetric, non-zero/zero sum, co-operate and defect . . . usually too deeply and for too long, before rescuing her with the enlightenment of how relevant any of this actually is. This volume doesn't really get out of the dry maths until after 200 pages either, though it has the biographical and historical angles to help retain engagement. But nobody will read to the end unless she is already something of a game-theory nerd, as this reviewer apparently is.
That is a bit of a pity, since several things would be better if they were at least mentioned (if not covered) near the book's beginning: the universality of 2x2 dilemmas, their extremely common occurrence in nature, and the link to natural selection, evolutionary stability, and the coup-de-grace offering whereby theory hints at a mathematical proof that co-operation not only often does but often should out-compete its alternatives. This all brings relevance to those payoff diagrams and numbers as something that actually matters.
A similar revelation was published in Robert Axelrod's 1984 book "The Evolution of Co-operation" which this reviewer loved. Richard Dawkins' "The Blind Watchmaker" has some of it too. Your reviewer is still waiting for the book that goes further and virally popularises a line of inquiry she is quite passionate about.
The genius of game theory--that strategic interaction is amenable to quantitative analysis, that it is rational and beyond the remit of probability theory--is hinted at as partly von Neumann's claim to fame but also Emile Borel's some years before. The mini-max theorem (exemplified by having one child slice a cake into two, and another child make first choice of slice), and the breakthrough which imbued mathematical credibility to the science, is credited to "Johnny".
But while the maths is robust, theory also fails to predict actual (human) behaviour in several respects. One such is the tendency to beat an opponent even if do doing means walking off with fewer gains. Another is the opposite--sharing spoils where kinship-compatibility is present even if both players could do better with a less-even split. Such quirks--which formally limit the theory--are attributed to non-rational responses. As is usual in this context, this reviewer tends to have a more encompassing view of rationality which captures everything that confers an evolutionary advantage, as the cited behaviours do. She accepts that this steps back outside of mathematics. . .
Two insightful points on the classic PD game: conferring and making a pact before playing (which is usually contrary to the given conditions) doesn't actually help the two players to co-operate because there is still no mechanism to enforce the pact, and ratting happens. Also iterative games--usually understood as quite distinct from the one-shot game in respect of allowing mutual co-operation to germinate--are not rationally any different and here's why: the last iteration is a one-shot deal (players "should" defect), so the penultimate round is the last one where there is a meaningful choice of co-operate/defect, right? Except that that means it isn't either since there is no influencing to be done. Crunching this through, it turns out that in none of the multiple iterations is it ever logical to co-operate. This has the delightful name of backward-induction paradox.
The numerous accounts of the cold war arms race, the "logic" of a preventive war in the late 1940s (another example of no leader actually accepting a mathmatical result), and the strategy, genius or otherwise, of deliberate risk creation during the Cuba missile crisis are insightfully told, and analysed through the lens of game theory. But this book never allows it to get too big for its boots. Supposedly after high school there is no such thing as algebra. Maybe make that: after Princeton?
The title comes from the best-known and most baffling of two-player, two-choice confrontations (there are three others which formally rank as dilemmas--where individual gain opposes public good--the next best known being "chicken").
Any in-depth treatment of these conceptions normally buries all but the geekiest reader in payoff matrices and jargon such as: symmetric, asymmetric, non-zero/zero sum, co-operate and defect . . . usually too deeply and for too long, before rescuing her with the enlightenment of how relevant any of this actually is. This volume doesn't really get out of the dry maths until after 200 pages either, though it has the biographical and historical angles to help retain engagement. But nobody will read to the end unless she is already something of a game-theory nerd, as this reviewer apparently is.
That is a bit of a pity, since several things would be better if they were at least mentioned (if not covered) near the book's beginning: the universality of 2x2 dilemmas, their extremely common occurrence in nature, and the link to natural selection, evolutionary stability, and the coup-de-grace offering whereby theory hints at a mathematical proof that co-operation not only often does but often should out-compete its alternatives. This all brings relevance to those payoff diagrams and numbers as something that actually matters.
A similar revelation was published in Robert Axelrod's 1984 book "The Evolution of Co-operation" which this reviewer loved. Richard Dawkins' "The Blind Watchmaker" has some of it too. Your reviewer is still waiting for the book that goes further and virally popularises a line of inquiry she is quite passionate about.
The genius of game theory--that strategic interaction is amenable to quantitative analysis, that it is rational and beyond the remit of probability theory--is hinted at as partly von Neumann's claim to fame but also Emile Borel's some years before. The mini-max theorem (exemplified by having one child slice a cake into two, and another child make first choice of slice), and the breakthrough which imbued mathematical credibility to the science, is credited to "Johnny".
But while the maths is robust, theory also fails to predict actual (human) behaviour in several respects. One such is the tendency to beat an opponent even if do doing means walking off with fewer gains. Another is the opposite--sharing spoils where kinship-compatibility is present even if both players could do better with a less-even split. Such quirks--which formally limit the theory--are attributed to non-rational responses. As is usual in this context, this reviewer tends to have a more encompassing view of rationality which captures everything that confers an evolutionary advantage, as the cited behaviours do. She accepts that this steps back outside of mathematics. . .
Two insightful points on the classic PD game: conferring and making a pact before playing (which is usually contrary to the given conditions) doesn't actually help the two players to co-operate because there is still no mechanism to enforce the pact, and ratting happens. Also iterative games--usually understood as quite distinct from the one-shot game in respect of allowing mutual co-operation to germinate--are not rationally any different and here's why: the last iteration is a one-shot deal (players "should" defect), so the penultimate round is the last one where there is a meaningful choice of co-operate/defect, right? Except that that means it isn't either since there is no influencing to be done. Crunching this through, it turns out that in none of the multiple iterations is it ever logical to co-operate. This has the delightful name of backward-induction paradox.
The numerous accounts of the cold war arms race, the "logic" of a preventive war in the late 1940s (another example of no leader actually accepting a mathmatical result), and the strategy, genius or otherwise, of deliberate risk creation during the Cuba missile crisis are insightfully told, and analysed through the lens of game theory. But this book never allows it to get too big for its boots. Supposedly after high school there is no such thing as algebra. Maybe make that: after Princeton?
via JREF Forum http://ift.tt/TqSTSJ
Aucun commentaire:
Enregistrer un commentaire