Re: [asa] prisoner's dilemma

I had a look at the chapter in the 1989 edition of Dawkins' "The Selfish
Gene" in which he discusses the Prisoner's Dilemma at some length, and I'm
definitely less than convinced that it has much to do with morality.

In a chapter headed "Nice guys finish first", he discusses the game of
"Iterated Prisoner's Dilemma", where you just do the prisoner's dilemma
repeatedly and accumulate the scores at each turn. There was a competition
run by a guy called Axelrod in which computer programs were submitted to
play the game in an all-play-all tournament. It turned out that the program
that won the tournament was one of the simplest, called "tit for tat". The
strategy it adopted was to cooperate on the first move, and then for all
successive moves, merely copied the last move made by the other player. So
if the other player defected on one turn, it would "retaliate" by defecting
on the next turn etc. This was deemed a "nice" strategy on the basis that
it was never the first one to defect. Two "Tit for Tats" playing each other
will always both cooperate. A strategy that could be the first one to
defect was deemed a "nasty" strategy. The tournament result showed that all
the "nice" strategies finished in the top half (except one), and
correspondingly all the "nasty" strategies ended in the bottom half.
Various other "moral" features were seemingly demonstrated. A slight
improvement on Tit for Tat was "Tit for Tat with forgiveness", where with a
certain probability (about 1 in 4) a Tit for Tat would refrain from
retaliating, but would cooperate, which could sometimes break the cycle of
mutual defections.

It seems to me that the labels assigned for "morality" bore little
resemblance to genuine altruism in the Christian sense of the word. In
order to succeed well against all-comers, a program HAD to have retaliation
built in to it. The concept of turning the other cheek (a strategy of
always cooperating) is a dead loss and would always fall prey to a "nasty"
strategy because it always gets suckered when the other player defects.

So I think the terms "nice" and "forgiveness" are just convenient labels to
put on these abstract strategies, but bear little resemblance to genuine
morality. "niceness" and "forgiveness" help you to win, but only if you
don't take them to excess - always turning the other cheek in such games
doesn't work.

Iain

On 8/15/06, David Opderbeck <dopderbeck@gmail.com> wrote:
>
> *So is our morality really altruistic? Or just selfish at a higher
> community
> level? A wise pastor of mine once referred to it as "enlightened
> self-interest".*
>
> This is one problem with using game theory to assess "morality." Perhaps
> none of the students are acting "morally." There needs to be a way to
> separate "moral" actions from the "natural" or "typical" exercise of human
> nature, which our Christian faith tells us is depraved. Game theory can't
> supply that.
>
> On 8/13/06, mrb22667@kansas.net <mrb22667@kansas.net> wrote:
> >
> >
> > Here is the original version as I remember it.
> >
> > Two suspects are apprehended for a crime. The state has a weak case and
> > the
> > prosecuting attorney is therefore hoping to get the suspects to testify
> > against
> > each other. They are taken to separate cells, and each are given this
> > identical message: If you testify against your partner, but he
> > refuses to
> > testify against you, then you will go free and he will get the maximum
> > prison
> > sentence. However if you and he both testify, you will both serve long
> > (but
> > shorter than the maximum) sentences in prison for your crime. If both
> > of you
> > refuse to testify then you will both serv short prison sentences and
> > then go
> > free. Your partner is being given this same offer.
> >
> > Summary: both are tempted to go for the best outcome for
> > themselves (to go
> > free if they testify and the other doesn't). But since they will
> > probably both
> > do that, they will implicate each other and get the second WORST option
> > (the
> > long prison sentence). However if they could both just trust each
> > other to
> > hold out, they could both get the SECOND BEST option of serving the
> > short
> > sentence. But, unable to communicate, they just have to trust each
> > other to try
> > this option. What would you do?
> >
> > I like to modify this to a "rewards" scenario that I can have students
> > pair off
> > and do in the classroom. Each student in a pair can choose between two
> > options, say "TRUST" or "GO FOR MORE". If one trusts & the other
> > chooses to
> > trample them, then the Go for more person gets 3 m&ms and the trusting
> > person
> > gets zilch. But if they both go for more, they both only get one.
> > However if
> > they both choose to trust, then they both get two. Do this scenario
> > several
> > times, and it is interesting to see the class dynamics. The totally
> > trusting
> > pairs will net more m&ms between them than any other option. But if
> > your
> > partner is trusting you, why not go for 3 m&ms instead of 2? And on it
> > goes.
> > So is our morality really altruistic? Or just selfish at a higher
> > community
> > level? A wise pastor of mine once referred to it as "enlightened
> > self-interest".
> >
> > --merv
> >
> > To unsubscribe, send a message to majordomo@calvin.edu with
> > "unsubscribe asa" (no quotes) as the body of the message.
> >
>
>

-- 
-----------
After the game, the King and the pawn go back in the same box.
- Italian Proverb
-----------
To unsubscribe, send a message to majordomo@calvin.edu with
"unsubscribe asa" (no quotes) as the body of the message.