Re: (OPE-L) Re: how simple is too simple?

From: Ian Wright (iwright@GMAIL.COM)
Date: Fri Sep 03 2004 - 14:21:47 EDT


Hi Jerry

I hope you enjoyed the sailing -- it sounds very adventurous.

> if one is presenting a model or illustration that purports to
> say something meaningful about the capitalist mode of production
> "how simple is too simple"?

This is a difficult question to answer in the abstract, and would
require an extended discussion on the philosophy of scientific
method.

If a model is too simple then it should be possible to show that more
concrete determinations invalidate the conclusions that have been draw
from the simple model. The classic example is the simple commodity
economy in which prices are proportional to labour values. It is too
simple, because if we include the extra determination of
profit-equalising prices then there is an apparent contradiction.

But this is part of the process of inquiry. The advantage of starting
with a simple model is that it is, after all, simple.

I agree with the dialectical rule of thumb that all theories are too
simple in the sense that they are incomplete: various totalities can
be ideally considered in isolation, but in reality they are connected
and affect each other.

But taken to an extreme this point of view can prevent the detection
and isolation of separable mechanisms that constitute a complex
totality. It can degenerate into a pseudo-theoretical assemblage of
concrete empirical facts, with lots of special case explanations, in
place of a deep understanding of the underlying unity of the phenomena
under investigation.

There are two rules of thumb I generally follow: (i) make theories as
simple as possible, and (ii) make them explain as much as
possible. Clearly these requirements are contradictory.

A justification for (i) and (ii) can be found in information theory,
via a modern form of Occam's razor. Basically there is a smaller
probability that a simpler model will fool us into thinking we have a
good model for the phenomena than a more complex model.

Explorations of simple computational rules have shown that they can
generate apparently very complex, disorderly and patternless
phenomena, which on the surface would seem to require a
correspondingly complex explanation. But in reality it is simply a
difficult problem to deduce what simple rules could have generated the
complex phenomena. This is shown very clearly in Wolfram's "New Kind
of Science". So apparent complexity does not imply that a complex
explanation is required. It might be better to assume that apparent
complexity implies that there are one or two simple mechanisms that
are interacting with each other.

Also, models can be far too complex. Although I agree that in an
important sense the Arrow-Debreu model of a competitive economy is far
too simple and ignores essential features of capitalism, it is also in
an important sense far too complex because it has a huge number of
parameters, such as those that specify in detail the subjective
preferences of all individuals. It turns out that the model can be
configured to generate any dynamic behaviour we wish (SMD theorem),
and hence its explanatory value is low because it does not rule out
any possibilities. It certainly does not explain the invisible hand;
if anything, it is an argument against it. As you say: "even if an
isolated individual situation can be posited we can not legitimately
infer that what is true for that individual situation is also true in
the aggregate". For this overly-complex model we can infer that
anything can happen in the aggregate, in other words we cannot infer
much at all. As Duncan Foley says somewhere, "it is "methodologically
too ambitious"; it attempts to model too many things in too much
complex detail.

ATB,
-Ian.


This archive was generated by hypermail 2.1.5 : Sun Sep 05 2004 - 00:00:00 EDT