[PREV - OUTNICE] [TOP]
Additions: July 7, 2000
Robert Axelrod's "The Evolution of Cooperation":
I think Axelrod essentially
confirms that much of what
we normally think of as He makes the point that
"altruistic" behavior can cooperative behavior can
arise from purely "selfish" and does evolve in
behavior. adversarial situations,
even without any direct
The way I'd put it: communication to assist
What is usually thought of the process (e.g. WWI
as "selfishness" is really trench warfare, where
"shortsightedness". each side of a stand-off
would gradually learn
to avoid any real attempt
at attacking each other.)
The moral may Much is made of
be grounded in the a computerized
real. tournament, an
indefinite series
The idealistic, of prisoner's
derivable from dilemma games,
the pragmatic. where the winning
strategy was
A is A? nearly always
the exceedingly
And simple "Tit for Tat".
D is D.
Tit for Tat:
Always cooperate,
unless your
opposite number
defected on the
immediately
previous move.
Be helpful.
Retaliate immediately
when wronged.
Forgive quickly.
COMPLAIN_COMPLAIN_COMPLAIN
This certainly
isn't my strategy:
I'm inclined to
A subject: the difficulty of the "slow to anger,
applying the idea of but never forget"
"Tit for Tat". style.
e.g. what's an appropriate
tat for a given tit? I think it's notable
that a close
Should you apply a competitor of
sliding scale of Tit-for-Tat
reprisals, where was Tit-for-Two-Tats,
you hit back harder But... attempts i.e. let the first
after a repeated at more complicated one slide, and
offense? strategies with retaliate on the
memory were failures second offence.
How do you manage to in the tournament.
be both "provocable" A variation of
and "forgiving" in a "turn the other
world where things cheek"... given
are not neatly that we only
divided into rounds have two cheeks.
of exchanges of a
limited kind of (or is it
interaction? four?)
And... isn't Tit for Tat
a strategy that can
reduce you to the level
of your opponent?
You can lose when trying to
convince a third party that
you're not as bad as your
opponent. Yelling "He
started it!" isn't going to The question is,
distinguish you very well is it possible that
when both sides are lying, the tournament
cheating and killing. results are not
strictly applicable
to reality, which is
not precisely a simple
extended series of
Prisoner's Dilemma
games.
Maybe Tit-for-two-tats
would win in a tournament
with parameters closer
to reality?
What happens if
you have some
freedom to choose
who you're going
to play with
next?
As I remember it
Axelrod's tournament
involved only
random pairing of
players.
Preserving
reputation
could be
worth the
loss of
a round.
March 7, 2003
Axlerod shows that some of what we consider civilized
behavior ("cooperation") falls out of the mathematics
of strategy. The original nature of the autonomous
agents may not matter very much... if they *are*
autonomous, you can expect that they'll tend to
converge on a certain kind of behavior over time.
-------
[NEXT - COMPLAIN_COMPLAIN_COMPLAIN]