Skepticism about Game Theory (was IRAQ sort of: Torching the oil)

From: Lee Corbin (lcorbin@tsoft.com)
Date: Mon Feb 24 2003 - 19:31:03 MST

  • Next message: Michael M. Butler: "ISTAS versus ITES, was Re: MEDIA: Satellite imagery in news coverage"

    At least a couple of posters have implicitly assumed the usefulness
    of a game-theoretic approach to pursuing international questions,
    and, I think that I once implied as much myself.

    But I am very skeptical that any new insights will obtain
    in this fashion, at least at the amateur level of everyone
    here, and the extremely likely oversimplifications of the
    formalism. But the main reason is that evolution has
    already equipped us with an awesome ability in this area.

    (Ever notice the language substitution God --> nature --> evolution?)

    We extremely easily imagine "if I do this, then he (they) could
    do that, and then... there would be three cases, and in case one..."
    and so on. People who play games such as chess and Go do this
    almost exclusively (although, in addition, much of their learning
    comes from something akin to language acquisition). Yet they
    almost never find it of much use to actually diagram tree
    decisions, even if they were allowed to do so; one's natural
    abilities are admirably well-suited, as I say. (It is true: is not
    permitted to "make notes" while playing OTB (over-the-board), and
    I shall have to consult people I know who do a lot of correspondence
    chess whether or not they resort to that. I doubt it.)

    Now "Game Theory" usually means something rather different. It
    mostly connotes the situation in which one's plays
    are not laid out so neatly as in chess, tic-tac-toe, or Go.
    Instead, one has moves whose consequences (and the resulting
    part of the game tree) are hidden from one. The bulk of the
    work in game theory concerns moves constituting "mixed strategies",
    where randomness is used to select a move; GT then allows one in
    many cases to compute a probability density over one's possible
    choices.

    I know specifically of two or three historical situations in
    which the generals could very well have used a game-theoretic
    approach in this way. One was an ancient battle in the Mideast
    (Meggido), in which the Egyptian army had to choose between
    attacking their opponent head-on, or from behind through a
    mountain pass. Neither possibility would be optimal if
    the opposing general guessed their strategy. Likewise, Luce
    and Raiffa give the example, in their classic "Games and
    Decisions", of the naval game between Admiral Kenney and
    the Japanese in their struggle for New Guinea, "The Battle
    of the Bismarck Sea". Normally, though, as in the D-Day
    campaign, good old deception is used if you can get away
    with it.

    Now sometimes GT could be useful, it seems to me, in helping
    some people see over their ideological blinders. But even
    there, I would really like to know of any actual cases. People
    who simply never could grasp the Cold War in abstract terms
    would nonetheless excel in understanding a precisely
    equivalent contest between, say, union and management.

    Lee



    This archive was generated by hypermail 2.1.5 : Mon Feb 24 2003 - 19:27:11 MST