|
One person's "model" is another person's "simulation".
If the question is; "When is a model not a model?"... then, is the
answer; "When it's a simulation"?
It seems to me that these two concepts (model and
simulation) are simply different facets of the same idea. A simulation is a
kind of model. Mimicking some system in appearance only is still modeling the
appearance, is it not? Some models may refer to underlying causality and other's
refer to external appearance but they're both models. The problem is that some
people think if you create a convincing enough simulation, you have achieved the
creation of a system exactly like the one being simulated. This is not the case
and I doubt I need to spell out why, for this crowd. The fact is that it is much
easier to model the appearance or external behaviors of some system than to
model the underlying causality that creates the appearance or behavior in that
system. The reasons are because the appearance and external behaviors are much
easier to study in detail, whereas the underlying causality is often difficult
or impossible to fully sort out. Sometimes what makes a system's causality
impossible to sort out is purely the fault of the approach taken. This was the
case with using "laws" of physics to try to answer biological questions,
as my father has described in detail.
A bad model is still a model, though. Anything that is used to
represent some aspect of something else is "a model", by definition. It's about
surrogacy. We can argue about whether any given model is a good model or a bad
model, based on a list of criteria (we can argue about the criteria, too!) and
the Hertzian Condition... but anything can be used as a model if it is
representing some aspect of/about another system.
Judith
----- Original Message -----
Sent: Friday, December 17, 2004 5:53
PM
Subject: Re: [ROSEN] Modeling vs.
Simulation [from "fundamental problems in physics"]
John M.
Sorry to not be explicit, but i think it is not
clearly definable in
practice. I don't have a theoretical definition for
it as Rosen does, or
appears to.
In practice, it seems to be any
representation of a natural system over
time that is not constructed on
sub-models of causalities, ie. generally
thought of as the processes
(efficient causes). However, it is quite
subjective to decide when a model
has taken into account enough
causalities to qualify as a prediction. Most
climate models and
hurricane models are referred to as simulations if they
depict a
possible event. For example, we have a weather simulator based on
a
database of historical storms. If one specifies a set of conditions and
a location, it will construct a typical storm based on prior storms and
then show it to you in the desired location. You can then examine how
such a storm would affect that location, were it to occur. It is a
hypothetical model run, in that case. Such simulations could be causally
based or based on statistical estimates - either one. It is not a model
because it does not refer to actual or expected conditions, only
hypothetical ones. Another example, the game SimEarth actually had
dynamical equations to deal with global climate and ecosystem change. It
was simplified for game purposes, but it employed some knowledge of
causalities. One could then construct hypothetical Earth systems and
watch them evolve as one altered conditions, added factories or carbon
sequestering units, etc. The idea was to see if you could stablize
the
system.
the distinction Rosen makes about simulation seems to
more refer to how
good one's model is - whether or not it is based on
reality or is just
meant to mimic or look like reality. But the problem I
have with that is
that all scientific models are mimics to some degree,
because we don't
know all the causalities or if we've formalized them
adequately.
Best,
John K
|