Peter's paper -- representational issues

From: Jay Lemke (jllbc@cunyvm.cuny.edu)
Date: Sat Jun 10 2000 - 20:53:40 PDT


The second part of Peter's paper takes off from arguments by Smith and
Thelen (whose 2 vol. collection on self-organization and psychology I once
reviewed) about a different kind of 'embodiment' for cognition -- its
material production in human bodies in interaction with their environments.

Theirs is a more general argument than Lakoff's, but where Lakoff's is
limited by the extent to which we believe that the semantic primitives of
language arise only from a Newtonian view of the body-in-the-world and that
those primitives loom larger than all the higher order resources that
discourse has for making sense of the world, Smith and Thelen's argument is
limited by the fact that it's only been convincingly worked out in practice
for studies of the simplest forms of human behavior (like locomotion).

But what Peter is trying to come to grips with in relation to Ilyenkov's
notion of the role of the ideal in intelligent behavior is a broader claim
that Smith and Thelen are trying to explain with the very unfortunately
chosen example of the steam governor.

Watt’s steam turbine governor is a really BAD analogy not just for
intelligent human behavior, but for what Smith and Thelen are trying to
describe: a self-organizing system. Yes, machines can exhibit something
like ‘emergent behavior’ where the sum of the parts does qualitatively new
things not anticipable from knowning all about the parts without also
knowing all about the contexts in which the whole is going to function.
Smith and Thelen are borrowing this type of example from the new paradigm
in robotics (e.g. Rodney Brooks, P. Maes), where you try to make a machine
that interacts with its environment in complex and (to us humans) adaptive
way by just programming it with simple local rules and letting its material
interactions with the environment more or less determine how those rules
apply in each instant. You don’t program it with a ‘schema’ of general
rules and then another set of rules for how to operate with those general
rules under various conditions; i.e. you don’t build a
representation-driven machine, you build a functional-interactivity driven
machine. The surpise is that when a lot of simple rules interact in a real
environment, you get unpredicted, complex, and sometimes adaptive-looking
behavior.

This new robotics of representation-less artificial intelligence has a lot
in common with genuine dynamical-developmental open systems like cells,
organisms, ecosystems, or cities. For a theory of cognition it suggests
that (1) brains don’t contain, compute, or embody rules, schemas,
representations, or symbols; they just interact with their environments in
complex ways that some people like to describe in those terms; (2) you
don’t need a notion of mind separate from that of functional brain
processes; (3) human symbol-mediated behavior is a very elaborate special
case of co-self-organization with an environment; (4) human semiosis is
itself describable as an emergent phenomenon (really an epi-phenomenon) of
a particularly twisted (heterarchical) multi-level
neurological-organismic-social-environmental system of material processes.

There is a much more general philosophical critique of representational
cognitivism (e.g Bickhard & Terveen) that offers the AI community an
alternative paradigm for building as-if intelligent machines, and the rest
of us an alternative account of the phylogeny of, and the minimal system
conditions for, getting ‘cognition’ in Peter’s special sense: adaptive
learning and perhaps a number of other interesting phenomena like
planning, anticipating, imagining, etc. This approach has a lot in common
with my own dynamical systems approach, which starts with the evolution of
living systems and asks what are the simplest conditions needed to have a
system that develops in interaction with its environment and that moreover
‘does semiosis’ (i.e. responds adaptively or functionally in context to
interactions with the environment on some higher scale of organization,
where the response appears to have taken the lower-level interaction as a
‘sign of’ something else that is relevant at the higher level; if it
sounds confusing, that’s the ‘twist’ .. see my longer accounts for a more
roundabout, less dizzying version, e.g.
http://academic.brooklyn.cuny.edu/education/jlemke/papers/gent.htm ).

So, I’m certainly with Peter when he analyzes the behavior of Watt’s
machine in the context of the human design and use processes that define
its functional context and what counts as adaptive functionality (i.e.
successful design). So would Latour also agree with such an analysis. The
machine is part of a network that only makes sense with us, and our uses
for steam turbines, in it; and a network with Watt’s governor in it is
going to be different all around as a result … history!

Certainly part of this network is Watt’s talk and his engineering drawings
and written descriptions and theories, and these have the key Ilyenkov
property of the ‘ideal’ that they are meaningful or useful or functional
only in relation to a larger social-material system (my ecosocial system,
Latour’s actant-network) where different people have a way of using them
that keeps the network happy (i.e. turbines get built that can be used and
create a market for Watt’s governors so more of them get made, etc.).

We are however caught in a logical loop here if we continue to think of the
‘ideal’ or the ‘semiotic’ in terms of symbolic representations of
material phenomena. The only progress in the whole argument was to get rid
of the notion of representation, and with it of the separation of the
phenomenon from our ways of socially and practically engaging with it. The
dynamical systems (me), or self-organization (Smith & Thelen), or
interactivist (Bickhard, Brooks) alternative to representationalism is much
closer in spirit to Ilyenkov’s account of craft activity, or what we may
tend to think of more as tool-mediated than symbol-mediated activity. That
is, it is much closer to a-theoretical practice, non-reflexive practice, to
action-in-the-moment as in Bourdieu’s accounts (or Polanyi’s).

But there is not such a great distance either from Ilyenkov’s basic view of
the ideal as a materialization of possible activity, as in the form of a
text or engineering drawing. That materialization is the product of social
activity, not individual activity; it has a social history, it arises from
and affords social collaboration, and all the rest. It is just that in the
non-representationalist lingo we would not say that the symbolic artifact
has a direct correspondence to anything, that it represents or symbolizes
an action or a thing; we would say that a chain of correspondence is built
by human and nonhuman ‘actors’ (the drawing, the steel) materially (i.e.
eye-, brain-, and hand- wise) interacting through time in ways that are
at root (i.e. by an iteration of this argument) a-theoretical and not
symbol-mediated. That is, we are saying that a symbol is not something
different from a tool, it is a special case of a tool, and not
metaphorically either … it is not the idea of the symbol that mediates
behavior, it is the material form itself. The twist, and the miracle, the
new emergent phenomenon that makes it seem like symbols have a reality of
their own (whether Ideal as in Plato, or mental as in Descartes, or even
social-ideal as it seems for Ilyenkov) arises from what can happen in a
complex dynamical material system that has multiple levels of organization
and an awesome number of possible ways of connecting events on one level
(receptor-photon interactions) with events on much higher levels (turning
the drawing right-side up). Which of these connectionist possibilities
emerges in development depends on still higher levels of organization of
the same system: social interactions, social conventions, social history,
species and ecosystem co-evolution.

So we 'interactionists' (me, Bickhard, Smith & Thelen, and our ilk) are
really trying to account for the role of the ideal in a very small-m
materialist way, while still preserving the essential role of social
practice, community, and history in the account.

JAY.

---------------------------
JAY L. LEMKE
PROFESSOR OF EDUCATION
CITY UNIVERSITY OF NEW YORK
JLLBC@CUNYVM.CUNY.EDU
<http://academic.brooklyn.cuny.edu/education/jlemke/index.htm>
---------------------------



This archive was generated by hypermail 2b29 : Sat Jul 01 2000 - 01:00:31 PDT