Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!sdd.hp.com!decwrl!shelby!russell!burke
From: burke@russell.Stanford.EDU (Tom Burke)
Newsgroups: comp.ai.philosophy
Subject: Pragmatist thoughts on emergence
Message-ID: <15705@russell.Stanford.EDU>
Date: 8 Oct 90 01:34:59 GMT
References: <62500@iuvax.cs.indiana.edu> <1990Oct7.012523.3828@watdragon.waterloo.edu>
Organization: Center for the Study of Language and Information, Stanford U.
Lines: 86

The following is a quote within a quote that will hopefully add an
interesting slant on the discussion about emergence, both as regards
what it is and what it's good for:

    Dewey and Mead used the category of emergence [to cope with the
    question of how to interpret the situation in which mind has arisen
    during an evolutionary process whose earliest stages contained no
    mental phenomena of any sort, not even "potentially"].  Its negative
    function was to resist the classic thesis that (1) since mental
    phenomena now exist, they must have been implicitly or potentially
    present in evolution from the very start; and (2) their potential
    presence played an active part in their later realization, and was not
    merely an abstract possibility.  Both constituents of the classical
    thesis, according to Dewey, are gratuitous assumptions which can be
    avoided if we take seriously the continuous, cumulative nature of any
    developmental process, for example, growth from infancy to maturity.
	"The reality *is* the growth-process itself; childhood and
	adulthood are phases of a continuity, in which just because it
	is a history, the later cannot exist until the earlier
	exists...and in which the later makes use of the registered
	and cumulative outcome of the earlier--or, more strictly, *is*
	its utilization.... The real existence is the history in its
	entirety....  Substitute for such growth a more extensive
	history of nature and call it the evolution of mind from
	matter, and the conclusion is not different.... When mind is
	said to be implicit, involved, latent, or potential in matter,
	and subsequent change is asserted to be an affair of making it
	explicit, evolved, manifest, actual, what happens is that a
	natural history is first cut arbitrarily and unconsciously in
	two, and then the severance is consciously and arbitrarily
	canceled.  It is simpler not to start by engaging in such
	manoeuvers."
	[J.Dewey: _Experience and Nature_, pp275-6]
    The "evolution of mind from matter" epitomizes not the sudden
    irruption of the mind into an otherwise material process, but a
    prolonged, successive emergence of certain "qualities and efficacies"
    of the behavior of organisms.  These qualities and efficacies are
    from the point of their first appearance genuine additions to the
    behavioral repertoire of their possessors.
    [T.A.Goudge: "Pragmatism's Contribution to an Evolutionary View of
    Mind", _Monist_ 57.2(1973), p142]

Here the term 'emergence' labels a concrete evolutionary mechanism.
It's not that mind or consciousness "emerged" from matter so much as
that behaviors emerged from other behaviors, the whole shabang
eventually taking on a character we now think of as being "mental,"
"intelligent", etc.  In this sense, mind is not something over and
above matter, but a character of the overall organization of material
processes.  AI (robotics, cognitive science, etc) should hope to
discover the nature of this organization which is otherwise a product
of natural evolution.

So what is "emergence" if it is simply a matter of behaviors emerging
out of behaviors?  As I see it, an emergent behavior is what you get
when a complex of behaviors is mastered as a single behavior (and which
will take on a different functional character as a single behavior
than as a disjoint complex of behaviors).  Perhaps this isn't a
definition so much as an observation.

Issues concerning "complexity" are important here.  But emergence is
to be understood not in terms of complex behaviors being generated
(emerging?) from a bunch of simple unit behaviors, but rather of
complexes of behaviors being mastered (habituated, routinized,
"learned") in such a way that they can function as simple units of
behavior.

For example, when a child learns how to draw a 'W' by drawing four
straight lines in the correct position, what "emerges" is the making
of four simple actions into one simple action.  Drawing 'W's, like
drawing straight lines before, becomes a simple action that can be
employed in still more complex behaviors, such as drawing words.

Following Dewey, this idea should be generalizable to explain
evolution at large, not just individual growth and development.  Each
such achievement of mastery of complexes into units is just one more
sashay onward in evolutionary development, at whichever level of
organization it happens to happen.  (In agreement with S.J.Gould, we
can hardly expect this to generate a "linear" view of evolution, but
that's another story.)

What this seems to suggest for AI research is to try to develop
robotic-computational mechanisms that can manage to turn complexes of
actions into units of action.  This may not be sufficient (who's to
say at this point?), but it does seem necessary if you want to go the
route of mimicking natural evolution in an attempt to design smart
machines that can learn and adapt to changing environments.