-----Original Message-----
From: ROSEN Forum [mailto:*** Behalf Of Judith
Rosen
Sent: Wednesday, March 10, 2004 7:01 PM
To: ***
Subject: Re: The value of Rosennean Complexity, applied...
I think I can see the writing on the wall.
John K., you may have just come up with the "test" for Rosennean
Complexity
theory. The kind of models that are going to need a Rosennean approach big
time are weather prediction models. As global warming progresses, the
weather is likely to be the single largest threat to humanity because
changing climate impacts everything from where disease organisms
can live on
the planet to drought/flood patterns to where agricultural crops
that need a
long time to become established (vinyards, orchards, coconut palms, other
nut trees, etc) will be squeezed and likely decimated by changing climate
trends... Shipping and off-shore oil drilling, among other
things, are going
to be affected adversely by rising sea levels, wilder storms, and
fluctuating weather patterns. The insurance industry will be just as
vociferous as all the others, because the claims will reflect the damage
that the weather and the changing climate inflicts. It's pretty scary,
actually.
Even if global warming is headed off somehow, better weather prediction
models are already the holy grail of meteorology. There's a lot
of money at
stake. So, what my father's work can do is give a modeler
insights into how
the inaccuracies of current models are being generated. Once you know how
these things are being generated you can come at the problem from two
directions to improve your models: 1.) You can eliminate as many of the
aspects of the model that just don't work with complex systems as possible
and devise new aspects to replace them which will work better. 2.) You can
devise additions to the model that can subtract out as many of the
complexity-caused inaccuracies as possible. Of course, you could create a
third option which is to generate new models from scratch with
complexity in
mind from the beginning.
I think another possibility (perhaps along the lines of the third option) is
based on this approach:
"It must be emphasized that we can still make dynamical models of complex
systems, just as we can formalize fragments of Number Theory. We can
approximate, but only locally and temporarily, to inexact differential forms
with exact ones under certain conditions. But we will have to keep shifting
from model to model, as the causal structure in the complex system outstrips
what is coded into any particular dynamics. The situation is analogous to
trying to use pieces of planar maps to navigate on the surface of a sphere."
[EL 338]
So, to consider modeling something like weather by seeking, from the outset,
*sets* of simple models, rather than attempting to incrementally improve 'a'
single vast model. The sticky problem then becomes one of devising criteria
for when and how to switch fluently between models in the set.
In the analogy of navigating a sphere, we can use the symptom of the
bifurcation of the observables of our latitude/longitude from the range of
those observables encoded in a planar map (that is, we have "run off the
map") to tell us when to switch to another map. Since the coordinate system
latitude/longitude is shared across all maps, we can easily tell which
map(s) we can switch to - in effect, the latitude/longitude system, to which
each planar map can be made commensurate, allows us to build a largest model
out of all the maps.
In a complex system, though, we will not have something analogous to a
overlaying latitude/longitude to watch to tell us which model to switch to -
a coordinate system and observables to which all the models in the set can
be made commensurate - we may only know that we are bifurcating from our
current model in one or more ways. This would be a symptom that it is time
to switch models. Based on Rosen discussion of emergence in FM sec 4.5,
which model we switch to would, at least in part, be determined by what the
current observables are and which model's stateset they naturally fall into.
Regards,
Tim
In any case, models that are geared toward complex models should (if done
right) have a better track record than models that are designed
with a false
supposition of simple systems in mind. The prospect of better,
more accurate
weather prediction is so important to so many different sectors of society
world wide that I should think it would be fairly easy to generate grant
funding. Think of all the countries/cities that are going to be adversely
affected by rising sea-level alone! Some countries are going to become
sister cities to Atlantis if we don't figure out what's going on
and what to
expect. Many of the richest, most densely populated cities of the
world are
near coastlines. Very expensive real estate and some big military
bases are
on the coasts here in the US. Funding for this kind of research is sure to
come from somewhere-- if not US government or academia, then another
country's govt. or what-have-you. I should think that private
industry would
be another biggie (oil companies, insurance companies, two places I'd
start...) or even private individuals. How many billionaires have beloved
estates on seaside real estate? What's a few hundred million?
Judith