>what we manufacture is a natural system. all real natural systems are complex.
>however, we percieve the system through putting a formal system in its place.
>that formal system need not be complex, it can be a simple mechanism. Indeed
>when we make TV sets we work hard to make sure their complexity does not get in
>the way of our design of them as a simple mechanism.
I agree with this. When we make a TV we intend to make a mechanism.
But instead we end up with manufacturing a complex system which
does not always behave as we thought. As I understand, you propose
to start building TV sets without having a mechanism in mind. For
example, we can make some self-organizing network that under some
conditions may develop into a functioning TV-set. In other words, we
will let the system to have more freedom. This is all fine.
But my point was that even if we intend to make a mechanism, we
still allow our product to have some freedom (unintentionally). For
example, we do not specify the path for every electron in a circuit.
Thus, the difference between bilding mechanisms and building
non-mechanisms is more in perception of what we do and less in
methods of manufacturing. But I agree that even methods of manufacturing
can be improved if people realize that they never can manufacture
a mechanism.
I suggest the following summary of this discussion:
1. All real systems are complex, and there are no mechanisms
among real systems
2. Models of real systems may be mechanistic (attempt to
describe and control every detail) and non-mechanistic (that
have some internal freedom, e.g., neural networks,
genetic algorithms).
3. Modelling strategies can be:
a) using mechanistic models and believing that they are true
representations of real systems
b) using mechanistic models as metaphors keeping in mind that
real systems may be quite different
c) using non-mechanistic models and believing that they are true
representations of real systems
d) using non-mechanistic models as metaphors keeping in mind that
real systems may be quite different
I vote for strategies "b" and "d" and against "a" and "c". Using
non-mechanistic models may give substantial advantages in some
cases compared to known mechanistic models. But I would not
consider non-mechanistic models as a panacea. In many cases
mechanistic models give a very accurate (but metaphoric)
description.
-Alexei
-------------------------------------------------
Alexei Sharov Research Scientist
Dept. of Entomology, Virginia Tech, Blacksburg, VA 24061
Tel. (540) 231-7316; FAX (540) 231-9131; e-mail sharov@vt.edu
Home page: http://www.gypsymoth.ento.vt.edu/~sharov/alexei.html