Bruce Buchanan wrote:
> I have read this with interest, albeit from the perspective of a relative
> amateur. And I realize that the question is one of *scientific* modelling.
> But if the objective is also to identify the features of a more general
> framework, I have a couple of (perhaps naive) questions, particularly from
> the point of view of cybernetics.
>
> What about modelling with dynamic features related to a possibly changing
> environment? Are there any additional requirements for models which might
> handle the less formal challenges of real world situations? These may be
> part of what is usually meant by complexity.
What is "usually" meant by complexity is a difficult area (for example
Don means something different from me by the term). One would have to
answer "by whom?" (a landscape painter, an electrical engineer, ...). I
am certainly of the view that although dynamic elements can contribute
immensely to the "complexity" of something it is not the only cause.
That aside, there are at least two different situations:
1. Consider an idealised situation where the amount of data and/or the
speed it arrives is low compared to the resources available for
modelling it. Then the dynamics of the situation can be attempted to be
captured in a model just like any other feature (with the roughly the
same difficulties and chances of success).
2. The case where the data to be modelled (or its arrival rate) is
great compared to the resources available for modelling it (for example
this may be due to interation with the source of the data). This is the
usual situation, the one I think you are talking about. In this case
there are sharp pragmatic decisions to be made about modelling strategy
- these will depend heavily on your goals etc. Of course, if this is
the situation even a static situation may force a similar approach.
It is true that a change from a sufficiently static object of modelling
and a sufficiently dynamic one may mean the difference between a type
(1) and (2) situation.
> I do not mean to criticize a model on grounds that are not within the
> objectives of the proponent and may not apply.
Yes my particular goals in that paper were to deal with (a) approaching
a meaningful and more general characterisation of noise and (b) dealing
with the "Myth of Simplicity" as Bunge would put it (i.e. that "simpler"
theories are a priori more likely to be true). I am however very
interested in understanding situation (2) (see another paper of mine
"Modelling Boundedly Rational Agents ..." at a link from my home page).
> Yet there are many
> situations where abstract formal models alone might be misleading. The
> requirement may be for systems models, in which specificity relates to the
> purposes or goals, and to the criteria which inform the feedback. In terms
> of such a model, error relates not only to predications of the model but
> also to the interactions and possible needs for adjustments among input
> factors and outputs, and perhaps the re- assessement of goals and criteria,
> etc.
1. Systems models are no different from any other "abstract formal
model" if they are formalised.
2. If they are not completely formal (or formalisable) then their
intended interpretation is important (due to their linguistic nature).
3. As with any model one needs to be aware of its limitations, its
testability etc.
I think I must have a wider conception of "model" than you. A model can
include goals, criteria etc. as an inherent part (whether it is formal
or not). A "systems model" is just a particular kind of model, with a
particular viewpoint, area of applicability, convention of
interpretation, style of modelling etc. The fact that it can be used in
reflective modelling is far from unique and is shared with many other
ways of modelling.
> My question, in summary: would the features required of dynamic feedback
> models which change over time and provide a guide for action (e.g. models
> useful for understanding and dealing with economic and/or environmental
> variables) be included within the capabilities of the *scientific*
> modelling described?
Yes, mainly because I did not specificy the situation at all tightly and
only came to very weak (philosophical type) conclusions. In fact the
paper can (and I would argue should) be read as an attempt to verbally
model some aspects of modelling. As such, it is ameanable to the usual
weaknesses and difficulties inherent in such an endeavour and should be
judged by usual criteria for informal linguistic modelling.
> As I see it, the question is important since assumptions about possible
> models which qualify as *scientific* may influence possibilities of
> perception and responsible action.
I don't *think* I make any such assumptions in the paper apart from
those inherent in taking a general extrinsic viewpoint (due to its
intended goals listed above and its intended audience of philosophers).
---------------------------------------------------
Bruce Edmonds,
Centre for Policy Modelling,
Manchester Metropolitan University, Aytoun Bldg.,
Aytoun St., Manchester, M1 3GH. UK.
Tel: +44 161 247 6479 Fax: +44 161 247 6802
http://bruce.edmonds.name