Thus spake Steve Smith circa 11/27/2009 01:01 PM:
In other words, a model isn't a model until it is _used_ rhetorically.
This is where we still differ. In 30 years of modeling and visualization work,
I have found that very few models, simulations or visualizations of the results
(visualizations often embedding their own models and simulation to fill in for
the incompleteness of the models/simulations they are (re)presenting) get by
without being used _rhetorically_ but more often than not they were conceived or
built with a much more _analytic_ intent and their major *value* often turns out
to be _synthetic_. That is to say, a well conceived/designed/built model
(especially those with a strong visual representation component) often end up
exposing knew knowledge to the creators of the model. New relationships are
discovered that were obscured by the obscurity of the system being modeled or
formal language of the _expression_ of the model.
Well said and I agree except that I believe _none_ of those models have
gotten by without being used rhetorically. My guess is that if you
identify the ones you think have gotten by without being used
rhetorically, we could trace their lifetimes and find that they have at
some point.
I'm still not clear that all models *must be* rhetorical devices, though I do
concede (again) that they are generally useful (and therefore used) for that.
Can you make the case that analysis and synthesis are also forms of rhetoric?
No. Analysis and synthesis are generic methods. They can be _used_ in
any number of ways. And they happen spontaneously, as well as
intentionally. So, while they are common tools in rhetoric, they are
not always rhetorical.
I don't know that I'm justified in this, but I think of rhetoric as being an
intentional act of persuasion by a sentient being. If we expand rhetoric to
include the case where the _model_ persuades the _modeler_ to believe something
formerly not understood (or believed) about the system _modeles_ then I would
concede.
First, we do NOT need to expand the definition of "rhetoric". Yours
suffices: an intentional act of a sentient being to persuade a sentient
being.
Second, let me restate my previous criteria: 1) models are extant
objects with purposes and/or causes and effects of their own,
independent of any referent and 2) they have a referent.
So, given that, (1) is trivial to deal with because if we just assume
realism, all "things" satisfy that. (2) is more difficult.
_How_ does "referral" come about? Can an extant object, without any
sentient beings involved, have a referent? Can one thing, objectively,
"out there", _refer_ to something else?
My claim is No. In order for something to be a model, there must be an
intentional, sentient being that assigns the object to its referent.
Now, why would a sentient being assign meaning (referent) to a symbol
(model)? My answer is "to reason", to parse and understand the world.
Here is where your methods of analysis and synthesis enter. But what
does it really mean "to reason"? If you're a solipsist, then you can
get away with saying something like: "to satisfy one's own sentient
self". But if you admit that other sentient beings exist and, to an
overwhelming degree, form and maintain your sentience, then it all boils
down to communication and a kind of shared understanding of the world.
I.e. you can't understand some thing unless other sentient beings also
understand that thing (obviously to varying degrees).
Hence, "to reason" requires sharing your understanding with your fellow
sentient beings. You got your mind from others and they got their mind
from you. This is true even if you're launched out into space right
after you're born, because your life-support capsule has our mental
constructs inscribed into it. Every object in the life-support capsule
has a meaning, an intentionally designed in purpose, put there by us to
keep you alive.
Now, if "to reason" is an act of communication (where modeling is a
specific type of communication), then all we need to establish is that
_all_ communication is a type of rhetoric.
What does it mean "to communicate"? When one sentient being tries to
transduce the contents of their mind into signals and send them across
various media so that they can be received and understood by another
sentient being, what is the purpose? What is the intention of the sender?
Is the sender's purpose to totally _brainwash_ the receiver so that the
receiver can think no other thoughts than those thought by the sender?
My claim is No. The sender's purpose is to do a good enough job
explaining their thoughts so that the receiver can, at least, ...
somewhat sympathize or empathize with the sender. The receiver need not
agree in total, but just enough for the sender to feel like the message
got through.
So, if you buy that, then we have to talk about "persuasion". When you
persuade someone, is it your intention to _brainwash_ the other person?
Do you intend to take over their mind and make them think precisely the
same way you do? Well, my answer is No. When we persuade, we attempt
to use the receiver's mental conditions to find some common ground and
just get them to move a little bit toward our mental conditions. We
don't engage in mind control. We engage in persuasion. It's not
coercion, but persuasion.
That is rhetoric. And we do it, intentionally, every time we
communicate. Hence, we do it every time we model. And, in fact, you
cannot avoid rhetoric without also avoiding modeling.
But otherwise, I think claiming that modeling is always a form of
rhetoric hides that fact that most of (real) science is not about persuasion but
rather discovery. Persuasion comes after discovery in *good science* methinks
Hm. This would require a new thread that I don't want to start. But in
short, I can say that science isn't about discovery, it's about the
elimination of _false_ rhetoric. That means, that science is about a)
generating rhetoric and then b) falsifying as much of it as possible.
The idea is, then, that what's left over is somewhat true.
Can one write a simulation without a
model?
Yes. Simulations can come into being in all sorts of ways, including
randomly.
I do think that evolutionary programming could be claimed to fit this somewhat,
but in the interest of splitting hairs, I would suggest that there is a "meta
model" involved... a model of what a "generic" simulation is, including some
kind of MOE to help guide the selection of the simulation.
Although I sympathize with your approach, there are 2 strong arguments
against this:
1) You've set things up so that you have an infinite progression (and,
hence, regression). In order to use evolutionary programming so that a
model arises with less intention, one needs meta-intention. Then, of
course, we could construct a system that didn't create models, but
creates modelers. Then we could construct modeler-constructors rather
than modelers. Etc.
It's better not to define words like "write" in this recursive way if we
want to be able to distinguish between things like naturally occurring
versus synthetic or artificially constructed.
2) One can accidentally write a simulation for something totally
unrelated to what they intended to write. Or, which is more common, one
can intend the simulation to be used for one thing; but it turns out to
be much more useful to simulate another thing. Neither of these cases
adhere to the "meta model" concept in the intentional sense you describe
above.
It's better to allow that any thing can be used to simulate any other
thing, as long as we can acceptably ascribe (and circumscribe) the same
attributes to both, regardless of the origins of either object.
When real world systems appear to mimic eachother, I would not call one a
simulation of the other but rather more like convergent or parallel evolution or
more often, systems whose underlying dynamics are constrained/informed (whatever
that means) by the same mathematics. It is *we*, the sentient beings who
impose on the two systems some kind of model that we then attribute according to
our points of view, etc. Don't you think?
Yes, most definitely. We agree completely, here. The sentient beings
do the modeling using simulations.
A model can be a theory or a thesis because a model can contain theorems
and sentences. (And the way we use the term "hypothesis" in science, a
model can also be a hypothesis... In fact, the way both words model and
hypothesis are used in science, all models are hypotheses because some
parts of every model are _always_ unjustified.)
And all models are always only partially validated? So there is some imaginary
threshold of validation where you would call a model "mere rhetoric"? Much
rhetoric is grounded in anecdotal evidence and can even have *scads* of
anecdotal evidence (high quantity, very low quality). D
Absolutely! Rhetoric is persuasion, not mind control. Models, as
rhetorical devices, require _some_ validation; but it is context
dependent as to how much and what type of validation is required to
persuade. Perhaps when we have two physicists trying to persuade each
other, the degree of validation is higher than when we have, say, two
theologists trying to persuade each other. But just because the extent
and type of validation required is context sensitive, doesn't mean the
threshold is "imaginary".
And in my experience, the evidence that the client is more interested in
rhetoric than in the data or a valid model comes out near the end of the
project, not at the beginning. With suitable cynicism, it is easy to
anticipate this, but hard to anticipate the opposite (recognize a righteous
client when you see one).
That's wise, and probably a lesson I haven't mastered.
It's perfectly reasonable to build models that support a client's
rhetoric as long as the client is willing to change their rhetoric when
all the models built in support of the old rhetoric are falsified.
Yeah... imagine that. It's a good idea and I suppose I've seen it happen now
and again, but usually something almost *more* nefarious happens... instead of
abandoning the old tired (and clearly misbegotten) rhetoric for something more
well justified in the light of the data and the models (and.. and... and...) I
often have found that the client simply adopts an even more bizzarre rhetoric
that isn't contradicted by the data/model but isn't necessarily well supported
by it either.
I have seen this to some extent. But with a clear research method, it's
usually easy to minimize the bizarreness of the new rhetoric. In
industry, we have the universal metric of money. If your bizarre
rhetoric can't save you money, then albeit true, it's useless. And in
science, we have the publication gauntlet. If you can't persuade at
least a few people that your bizarre rhetoric is worth propagating, then
albeit true, it's useless.
So, bizarre rhetoric can only survive if it can be reified at least well
enough to pass some fairly stringent testing.
I agree that *a* goal of modeling can be to reify the rhetoric. I suppose I'm
coming around (a little) to your use of the term rhetoric... in that I
appreciate that well accepted and highly supported (by data and models) theories
start out as pretty unsupported theories which start out as almost completely
unsupported hypothesis which are pretty much "rhetoric" even if the only one
being persuaded is the person creating the hypothesis and seeking a model that
helps fit the data to it which in turn helps too shape where one should seek
more data.
Yes. That's the gist, all hair splitting aside.
Usually, during the process,
what actually happens is that reality is used to measure the model.
Then the model (and the rhetoric) is changed so that it matches up with
reality. At that point, you have a good enough basis flip it around and
start using the model to measure reality.
I'll have to think on this more, as I want to argue that the act of using
reality to measure the model involves inserting "yet another model" into the
game which is roughly what validation is all about anyway... stacking a series
of more and more sophisticated models up from ones that seem to be so brutally
simple that they cannot be argued.
If you free yourself from the concept of "levels" and "stacking", then
you'll have a much easier time. [grin] It's not stacking model upon
model. It's an endless rhizomic bath where models (including
sophisticated hierarchies, but not requiring them) form and dissolve,
are compared and contrasted. Some cohere and stabilize for long times
and wide extent. Most dissolve quickly. "Brutal simplicity" is an
illusion, I think. I read what you're saying as "long times and wide
extent". Some models are so widely applicable, and seem to always be
true, that we just can't puncture them no matter what we try.
We should spawn a separate thread on this (or not).
I choose "not". [grin] I've waxed filosofickle too much lately.
