From newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!cs.utexas.edu!qt.cs.utexas.edu!yale.edu!nigel.msen.com!caen!umeecs!sparky.eecs.umich.edu!carnes Sun May 31 19:04:42 EDT 1992
Article 5967 of comp.ai.philosophy:
Newsgroups: comp.ai.philosophy
Path: newshub.ccs.yorku.ca!ists!helios.physics.utoronto.ca!news-server.ecf!utgpu!cs.utexas.edu!qt.cs.utexas.edu!yale.edu!nigel.msen.com!caen!umeecs!sparky.eecs.umich.edu!carnes
>From: carnes@sparky.eecs.umich.edu (Richard Carnes)
Subject: Re: AI failures
Message-ID: <1992May29.002128.12705@zip.eecs.umich.edu>
Sender: news@zip.eecs.umich.edu (Mr. News)
Organization: University of Michigan EECS Dept., Ann Arbor
References: <usm7vINN7np@early-bird.think.com>
Date: Fri, 29 May 1992 00:21:28 GMT
Lines: 143

This posting is a response to some recent comments made by Hans
Moravec on the subject of sociobiology.  It bears indirectly, but only
indirectly, on the usual comp.ai.philosophy concerns.  

In article <usm7vINN7np@early-bird.think.com> Moravec writes:

>[...] There is a very strong political motivation in the opposition to
>sociobiology - as it undercuts most of the Marxist social engineering
>premises.

And the *support* for sociobiology has no strong political
motivations??  Obviously, there are ideological passions influencing
both sides of the debate.  Equally obvious, scientific issues should
not be decided by ideological criteria.  But when high stakes ride on
the answers to scientific questions, we should exercise great care in
answering them.  Some sociobiologists have claimed that their
investigations show that social class structures are inevitable, that
aggressive impulses toward strangers are ineradicable, and that "even
with identical education and equal access to all professions, men are
likely to continue to play a disproportionate role in political life,
business and science" (E.O. Wilson).  If we act and construct policy
on the basis of these claims, the aspirations of many people for a
less class-stratified society, for a more peaceful world, and for
gender equality will be stifled.  I take it these aspirations are what
Moravec means by "Marxist social engineering".

We must distinguish two kinds of sociobiology (after checking under
the bed carefully for Marxists).  Philip Kitcher puts it this way:

  Sociobiology has two faces.  One looks toward the social behavior of
  nonhuman animals.  The eyes are carefully focused, the lips pursed
  judiciously.  Utterances are made only with caution.  The other face
  is almost hidden behind a megaphone.  With great excitement,
  pronouncements about human nature blare forth.
  
  I have attempted to identify the two faces: to make it clear that
  there have been great advances in our theoretical understanding of
  evolution by natural selection, that some of the techniques developed
  have been carefully applied to the study of nonhuman animal behavior,
  that there are interesting results about the social lives of insects, 
  birds, and mammals; and yet to show how the building of grand
  conclusions about ourselves is premature and dangerous.  We have seen
  again and again how the assertions about human nature begin with
  unrigorous analyses of fitness, how they deal loosely with data about
  animal and human behavior, how they employ problematic concepts, how
  they rely on dubious connections between optimality and selection, how
  they offer spurious arguments for the inflexibility of the phenotype.
  (VAULTING AMBITION, p. 435)

The distinction is between sober sociobiology, which consists in the
systematic study of the biological basis of all social behavior, and
what Kitcher terms "pop sociobiology", which "consists in appealing to
recent ideas about the evolution of animal behavior in order to
advance grand claims about human nature and human social institutions"
and is "deliberately designed to command popular attention".  It is
pop sociobiology that has received and deserved much criticism.  It is
not primarily the subject matter that distinguishes pop from sober
sociobio; rather, it is the style of argument: rigorous and careful
versus shoddy and flimsy.  

An illustration of the pitfalls to which pop sociobiology is subject
is provided by Moravec:

>   > I think a strong argument could be made that the early civilizations
>   > depended on slavery the way ours depends on mechanization, and wouldn't
>   > have worked otherwise.
>|> So therefore it was right?
>
>I don't believe in a universally absolute right and wrong as you seem
>to (based on your unqualified question).  I believe it was an
>appropriate social institution for its time.  Slaves were the
>large-civilization institutionalization of the guy in the tribe that
>everybody picks on, who gets to do all the dirty work, who gets fed
>last.  There's a pecking order among chickens, wolves, chimpanzees and
>humans.  It seems to be a necessary, or at least stabilizing, solution
>to the problem of group decision making.  Instead of fighting over
>every decision, we'll establish in advance who has to follow who's
>orders.

To begin with, what exactly is it that Moravec is asserting that
chicken, wolf, chimp, and human societies have in common?  That there
are power relations among the members of these societies?  But that is
trivial.  If we call the alleged common characteristic a dominance
hierarchy, what is its precise operational definition?  It may be true
in the case of a particular species that a hierarchy reduces
conflicts, but how is it determined which particular conflicts are
settled by "dominance" and which are settled by a fight or other
means?  Is there a single linear scale of "dominance" in animal
societies, as was formerly thought with respect to human
"intelligence"?

Assuming one has precisely defined the "pecking order" that allegedly
characterizes human societies, how does one know that it is the
product of natural selection?  It is not sufficient merely to show
that this is *possible*.  Perhaps the most characteristic error of pop
sociobiology is the equation of current utility with historical
origin.  The term "adaptation" ambiguously refers to both a state
(good design) and a process, but they are distinct concepts.  It is
often true that an adaptive feature or behavior became prevalent and
remains prevalent through natural selection, but it is not
automatically true.  The human brain, for example, has many possibly
adaptive capabilities (such as belief in a god or the ability to
create AI) that it is difficult to believe were selected for in our
evolutionary history.

Further, how does one know that there are genetic differences between
those humans who exhibit the appropriate pecking order behavior and
those who do (or did) not?  It is not a trivial question.  Behavior
that works need not have a genetic basis.

Finally, assuming that the supposed human dominance hierarchy
behaviors are genetically grounded, how does one know that they cannot
be modified (or only at unacceptable cost) by altering the social
environment?  Another difficult question.  All we have assumed is that
"given a common background, there is an allele whose presence or
absence makes a critical difference to the manifestation of the
behavior in the typical environments encountered" in a group of humans
and/or animals (Kitcher).  

Nevertheless, the pop sociobiologist leaps out of this muddle to his
grand conclusion about human nature: all human societies are ordained
by Darwin or Nature to have a dominance hierarchy, with Top Dogs,
Bottom Dogs, and those in between, and the Bottom Dogs are destined by
genetic imperative to take orders from those above and to do the slave
work of society.  Try as we might, there's nothing we can do about,
unless perhaps at unacceptable cost.  A cheering thought indeed to Top
Dogs everywhere.

>[...]
>Results from this kind of thinking that have excited me in recent
>years include Robert Axelrod's famous computer tournaments that
>demonstrate that being nice (cooperate, then tit-for-tat), rather
>than cheating whenever possible, is the strategy with the highest
>payoff in most long-term relationships.  

Correct me if I'm wrong, but my understanding is that what Axelrod
showed is not that "being nice ... is the strategy with the highest
payoff in most long-term relationships", but that TIT FOR TAT is the
dominant strategy for pairwise iterated Prisoner's Dilemmas in which
the number of iterations is unknown.

Richard Carnes
Licensed Leftist Humanist


