From: Rafal Smigrodzki (rafal@smigrodzki.org)
Date: Sat Mar 22 2003 - 20:16:11 MST
On Sat, 2003-03-22 at 15:12, Eliezer S. Yudkowsky wrote:
>
> I oppose the punishment of non-punishers; it may be an ESS but I don't
> think it's a good thing.
>
### I wonder if this is a rational position, or an emotionally based
one. The following two links point to some evidence that altruistic
punishment is a part of an evolutionary stable strategy, built into the
minds of most humans:
http://bostonreview.mit.edu/BR23.6/bowles.html
http://www.nature.com/cgi-taf/DynaPage.taf?file=/nature/journal/v415/n6868/full/415137a_r.html
(the second link might require a subscription to Nature. I can send a
copy to those who request it)
In the very long run, maintenance of stable societies made of
potentially immortal persons is probably possible using strictly
rational reasoning. If your attention-span is capable of encompassing
very long time spans, you will be able to act according to your
interests, and this will include cooperation, as well as discouragement
of non-cooperation. Since a failure to discourage non-cooperation can in
some cases bring about significant losses, it has to be discouraged as
well, as a second-order consideration. Shirking a duty to punish a
cheater puts a burden on other participants in the game, and is
second-order cheating, calling for second-order punishment. A concept
easy to grasp for a mind looking far forward, without the need for the
emotional crutches that we rely on.
On the other hand, the emphemeral creatures that we are, such sustained
rational actions were not possible in the EEA. Instead, common
hard-wired responses developed, probably through group selection, with
an innate tendency to seek revenge, and, punish and exclude those who
fail to seek revenge (also derided as "walk-overs", doves, peaceniks,
etc.).
I agree that the urge to seek revenge and the scorn for those who don't
share this thirst, are dark feelings, but I wouldn't unequivocally
condemn them. On one hand, hopefully humans will be able to reconstruct
themselves to survive without them, reducing the risk of useless
vendettas, chauvinism, and other forms of tribalism, caused by obsolete
calibration of the hard-wired systems. On the other hand, we will need
to use some parts of these hardwired strategies in a reasoned way, as we
will still need to act against opposing forces, dispassionately,
efficiently, relentlessly, to improve long-term cooperation and survival
outcomes of cooperators. A calm justice.
It's interesting to think about the programming parameters that will
need to be entered into our decision systems. To achieve maximum
survival, you might need to adjust the tendency to reciprocate (the
trigger-happiness) according to the reliability of information available
(the more reliable information over many iterations, the less need to
shoot at the first defection), general level of risk (low-risk, low
response), and certainly many other elements. Very, very interesting.
Rafal
This archive was generated by hypermail 2.1.5 : Sat Mar 22 2003 - 20:24:45 MST