Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Posting-Version: version B 2.10.3 4.3bsd-beta 6/6/85; site dg_rtp.UUCP Path: utzoo!linus!decvax!ittatc!dcdwest!sdcsvax!ncr-sd!ncrcae!ncsu!mcnc!rti-sel!dg_rtp!throopw From: throopw@dg_rtp.UUCP (Wayne Throop) Newsgroups: net.sci,net.philosophy Subject: Re: Prisoner's Dilemma Message-ID: <274@dg_rtp.UUCP> Date: Thu, 3-Apr-86 16:19:04 EST Article-I.D.: dg_rtp.274 Posted: Thu Apr 3 16:19:04 1986 Date-Received: Sat, 5-Apr-86 07:26:53 EST References: <12481@ucbvax.BERKEL <2007@brl-smoke.ARPA> <538@umich.UUCP> <2202@brl-smoke.ARPA> Lines: 55 Xref: linus net.sci:406 net.philosophy:4536 >>>, > gwyn@brl.ARPA Doug Gwyn >> torek@umich.UUCP Paul Torek >>>Many years ago, Anatol Rapoport had a Scientific American article >>>on games such as The Prisoner's Dilemma, in which he too tried to >>>argue against the logical conclusion of basic game theory, >>What conclusion is that, and what alternative did he propose? > [...] > It is clear that IF cooperation was allowed (it was prohibited) > and IF there was a means of enforcing the cooperation, THEN the > best mutual combined strategy would be to select the matrix > element that Rapoport wanted. Hmmmmm. Normally, cooperation in such setups is "allowed", it is simply not enforced. In fact, other than one play option often being called "cooperation", most setups don't mention cooperation explicitly at all... they merely state that the goal is to get the highest score (not the highest score for the group, mind you, the highest score for *yourself*). Is this what you mean by "prohibiting" cooperation? Now, if I'm remembering the article in question correctly (it has been *quite* a while), Rappaport was in essence trying to apply results gotten from an "expanded matrix", which looked sort of like he was talking about iterated trials, to the single-shot dilemma. This indeed seems bogus and I agree with Doug on that point, but I have to disagree when he says: > [...] to me, it was a blatant > attempt to pretend that there is a rational basis for socialism. (Not that I'm disagreeing that it seemed that way to you... :-) The iterated variants of PD, given concrete populations of players, *do* give a "rational basis for socialism". I should clarify that I'm taking "socialism" here to mean "choosing the 'cooperate' option in a PD-like scenario"... I'm not talking politics here. More concretely, what iterated PD *really* gives rational basis to is the notion that co-operation can evolve among players of iterated PD-like games, even when each player is *only* trying to maximize the player's own payoff. The point is that iterated PD is much more like the real world than single-shot PD, and thus the "always defect" that results from modeling real-world interaction as single-shot PD is often inappropriate. The bottom line is, Rappaport's analysis of single-shot PD seemed flawed to me also, but that doesn't mean *all* analyses of PD-like scenarios that find problems with the games-theoretic strategy are likewise flawed. In PD-like scenarios, players of the simplistic games-theoretic strategy can, in fact, be reliably and repeated outperformed. -- Wayne Throop at Data General, RTP, NC !mcnc!rti-sel!dg_rtp!throopw