Re: CAMREC: GAs and NNs are just as bad as optimisation!

From: the Campaign for Real Economics (
Date: Tue Mar 02 1999 - 16:42:26 GMT

Date: Tue, 02 Mar 1999 16:42:26 +0000
Message-Id: <>
To: bogus
From: (the Campaign for Real Economics)
Subject: Re: CAMREC: GAs and NNs are just as bad as optimisation!
In-Reply-To: <CAMREC: GAs and NNs are just as bad as optimisation!>

Message-Id: <>
To: CAMREC list members <>
From: the Campaign for Real Economics <>
Date: Tue, 02 Mar 1999 16:42:26 +0000
Subject: Re: CAMREC: GAs and NNs are just as bad as optimisation!

> Bruce Edmonds wrote:
> Dear Gianluca Baldassarre,

> Thank you for your supportive post and information about your thesis.
> Any steps towards greater grounding in actual studies of behaviour (even
> if anecdotal) represents progress.
> However, in my view, shifting from constrained optimization of utility
> to Genetic Algorithms (GA) or neural networks (NN) does not, of itself,
> represent any improvement in modelling real economic processes. The
> basic problem is when the technique takes precedence over the evidence.
> (Obviously, I do not know whether this is true in your case.)
> There are big problems with the descriptive credibility of GAs or NNs as
> representations of what real agents do. Simply put, there is
> (typically) no reason to suppose that they descriptively correspond to
> how the agents actually make their decisions. GAs and NNs are just
> another formalism (a JAF?). Shifting from one JAF to another will (in
> my view) make little difference to economics. It is rather like
> inventing a screwdriver and _then_ looking around for something to do
> with it!

> The only advantage to introducing GAs and NNs is that it marginally
> increases the variety of the stock of JAFs that we have to hand to help
> us in our modelling. In this respect the greater the variety the
> better, preferably no formal tools or methods of representation should
> be disallowed on a priori grounds.
> How do these thoughts square with the content of you thesis?

I think that one of the worst draw-backs of economic mainstream
is that it is "technique driven", in the sense that for historical reasons
chose to adopt the mathematical maximisation techniques as its
way to express/solve its models, and now it tends to
ignore/underevaluate all the problems/phenomena that
cannot be expressed within that framework,
e.g. disequilibrium dynamics and bounded rationality of agents.
(Exactly "inventing a screwdriver and _then_ looking around")

I believe (and you sound to agree) that one thing to do
to push economics out of that limited boundaries is
to enquire other methodologies, as computer simulations,
and other kinds of models, such as the ones based on artificial intelligence

(SOAR architectures, BDI architectures,
Neural Networks, Genetic Algorithms, etc.).
The goal should be to enlarge the "language" through which economics
expresses its models, trying to figure out which typology of models
can be used to better address which typology of economic problems.

On this purpose, in the first (theoretical) part of the thesis I
listed a series of economic phenomena that can be suitably studied
by using GA and NN (and similar techniques that allow us to represent
adaptive cognitive processes of agents),
but not by using the maximisation techniques.
This part of the thesis was due to my reaction of dissatisfaction with
the constraints imposed on the range of addressable problems,
by the techniques/language adopted by the economic mainstream.

A drawback of this part of the thesis, as Herbert Simon himself noticed,
is that I "a priori" and wrongly assign less importance to techniques
based on logic and symbolic artificial intelligence,
compared to the ones related to soft computing.

> What is important about a model is that it corresponds descriptively to
> the process being modelled, limited only by *justified* assumptions and
> the limitations of our modelling technologies. Then it also needs
> validating against the data from the object process.
> Why start with a GA? Why not start with a model of decision making or
> learning that is well validated itself (e.g. from cognitive science or
> marketting practitioners)?

The lack of contact with empirical reality is another main drawback
of economic mainstream that should be avoid.
I completely agree with you that in judging the different models/techniques
we have to assign a central position to the empirical evidence about
economic phenomena.
At the end of the day the utility of this or that technique/typology of
model has to be judged
on the base of its suitability to represent/understand different aspects of
real phenomena.

In the second (simulative) part of the thesis I walked that direction.
I used GA and NN to modelise and account for (even if in an anecdotal way)
some behaviours about the fixation of prise in oligopolistic markets.
In doing so GA and NN showed (to my opinion) to be quite appropriate
for the object processes.
So now (aside the interests of the results about the fixation of prises)
we have piece of work that can help us to delineate the utility of those

By the way, is the mark-up prise-fixing rule of any interest for your
research about
ways of fixation of prises in real markets?
Are you interested to have a copy of the article extracted from the thesis?

Gianluca Baldassarre.

Ph.D. student,
"Artificial Societies:
MAS and ALife simulations of evolution of culture",
Department of computer science,
University of Essex, CO4 3SQ,
Colchester, Essex, United Kingdom.

This archive was generated by hypermail 2b29 : Thu Jan 06 2000 - 17:00:29 GMT