In both cases the system pushes itself toward change. I would think that
dialectics is perhaps a special case of general self-organization dynamics,
the case where binary labelling makes sense. This happens, for instance,
when a dynamical system reaches a point of bifurcation, where two different
dynamical paths are both possible. Characteristically each of these paths
implies the existence of the other. But the usual view is that the system
chooses one or the other each time, and there is no push to the dynamics.
This view is sometimes criticized, leading to a notion of dynamic
attractors, i.e. changes with time in the most basic possibilities open to
the system -- a truly developmental view. What could drive such basic
changes? particularly how could they be driven in a way that is internal to
the system?
The usual answer seems to be that one has to consider system-environment
interaction. The system enters a state, and this has consequences in its
environment. It then adapts to these changes because its very existence is
coupled to its transactions with that environment. These adaptations may be
self-stabilizing (the picture of autopoiesis developed by Maturana and
Varela), or they may be self-destabilizing -- but potentially lead to a new
dynamical possibility for the system-in-context. This overall process
(really a meta-process) certainly sounds rather dialectical to me, and that
is how I characterized it in early work in the 70s that is recapped in the
Postscript to _Textual Politics_.
If someone wants to model dialectical processes in self-organizing
computational systems, they would be well advised to build in a sort of
life-and-death contradiction to the coupling of more complex systems to
less complex environments. The only way to live is to do things that
threaten to change the environment in ways that will kill you, forcing you
to change radically or die. A lot of systems will just die; those selected
by a genetic algorithm (evolution) will adapt. Biological life has a double
strategy of both development and evolution, with the former often recapping
the latter. At the least you would have to aim for a computational system
in which complex developmental sequences evolved from simple non-developing
(or minimally developing) systems. While this idea fits the a-life paradigm
more obviously than the neural-net paradigm, Edelman's 'neuronal Darwinism'
and some of his learning simulations would seem to bridge the gap.
In purely formal terms, dialectic relationships do not operate in terms of
contradictions on only one level of logical organization; they must operate
across at least two (some say three) such levels (of logical typing, cf.
types and types-of-types, or rules and rules-for-rules). Bateson's
discussions of meta-learning are a good source for ideas of how this might
be implemented, I think. In the a-life ecology perspective,
system-environment transactions form one level, and the overall dynamics
that determines what transactions are possible is the meta-level.
System-environment interactions lead to changes in what sorts of sys-env
interactions can take place; the total supersystem of sys-plus-env changes
in such a way that new possibilities of transaction come into being,
emergently and dialectically.
Too bad the human brain tends to blow a fuse when it tries to follow
meta-logic across more than one pair of simultaneous levels!
Note in passing that such meta-dynamic, or dialectical-emergent systems
also generate time, at least in the sense that they produce phenomena like
development and on-going evolution, or dialectical meta-instability, that
require extended trajectories through time for adequate description, and
such that they cannot be reduced to static descriptions at any level of
abstraction. In a single time-slice every living system is just dead. It is
only through-time that life can be defined -- and perhaps through-time on
many time-scales at once. Time does not pre-exist the dynamics of systems.
It is simply a way of describing that dynamics. To understand time, you
need to understand what kinds of dynamics are essentially temporally
extended processes that cannot be reduced to static representations.
(For those not familiar with physics, most of it is built on clever ways to
substitute static representations for dynamical systems. Time is introduced
into the theories only at the beginning and sometimes again at the end, but
the mathematical representations themselves are generally chosen to be
timeless and universal, standing outside of time, which is taken to be a
matter of mere accident, the 'initial conditions' of when some universal
process happens to be taking place on some particular occasion. Of course
the small footnotes in the textbooks all tell you that these are
mathematical shortcuts and that they severely limit physics to dealing with
extremely simple systems. Physics today gives almost complete accounts of a
tiny -- but ubiquitous -- subclass of real systems. Not including complex
time-generating ones like you and me.)
JAY.
---------------------------
JAY L. LEMKE
CITY UNIVERSITY OF NEW YORK
JLLBC who-is-at CUNYVM.CUNY.EDU
---------------------------