ontological levels

["Roberto Poli" <***>]

Tim,

Knowing that you defend a "moderate" form of subjectivism, I may subscribe
almost all of what you have said. Our difference is more a problem of stress
than of substance: I prefer to maintain in the focus of (my) attention the
fact that our models are models of something (the objectual side, so to
say). You prefer to focus your attention on the side of the model (the subjective
side). No problem.

Let me only add a couple of remarks. Ask yourself why models may change.
To cut short a long story, two main reasons can be found: (1) a model may
change for internal reasons (i.e., because we invent new algorithms or find
more efficient ways of dealing with some kind of information); (2) a model
may change for external reasons (because we would like or need to include
more data/information (new "meters") into our models. This latter move is
meaningful only if the new data (the new meter) makes some difference (in
the Rosennean jargon: if it introduces some difference into an equivalence
class). What does it mean? I take it as claiming that the new meter makes
more detailed our knowledge of the "object" (we know more of our target).
If the target depends on our decisions only, how can we arrive at knowing
more of it?

A different (and begging the question) answer is: there is no need to look
at the object. You may simply take your equivalence classes, partition them
in all the possible ways, and calculate all the possible answers. Nothing
wrong if not for the simple fact that this latter strategy cannot work (for
reasons of complexity -- in the traditional formal sense; RR would have
said, for reasons of complication).

Put it paradoxically: The objectual side is needed in order to simplify
and give order to our modelling capacities.

The objectual side is needed for making our models models of something.
The internal machinery alone is not enough. The strategy based on subjectivity
does not require anything more than internal (formal) machinery.

One more observation is needed. Our (scientific) models should respect a
basic constraint: the results arising from model-based inferences should
concord with the results arising from objective causes.

We may keep under control our inference engines. It seems much more difficult
to legiferate on the nets of real causes. They resist our efforts and more
than once they make us fool (this is why we call them objective :-).

This said, it is obvious that the greatest part of any scientist's work
deals with models. But here and there they have to check whether the model
went astray (i.e. whether the inferences internal to the model (say, the
laboratorium) are or are not in agreement with what is happening in the
real world).

cheers, r