After a few kind words by Pete Vincent,
> >For my part, I found your post excellent ...
Ed Weick replied:
> I, on the other hand, do not. I have seen little evidence that you really
> know anything about the global economy that you hope to model.
I don't, really -- you're right about that, Ed. Who does, though?
But I can write a simulator, put data into it, and see what happens.
And I can make it all public, so people like yourself can make
constructive criticism.
If I had to do everything myself and had to find all the mistakes
myself, it would be a hopeless task, but I think I can count on a lot
input by very sceptical people such as the ones on this mailing list,
and perhaps a bit more concrete help too.
Later in a more sober mood Ed wrote:
> Nevertheless, I do feel that the questions I have raised about the
> simulation that Douglas Wilson is proposing are valid: Is there really
> something to be simulated? If so, what? Will the proposed simulation lead
> to a better understanding of economic phenomena? And, do we not already
> have a considerable understanding of the global impacts of megaphenomena
> such as population growth and energy resource depletion? ...
The only real response I can make to each of these questions is that a
good simulation should answer them. I don't know the answers, but I
think I can get some answeres, even with only the prototype.
Ed wrote:
> On whether there is something to be simulated, I pointed out in a previous
> posting that, despite headlines and hype to the contrary, economic activity
> is still overwhelmingly domestic, not international. This makes me wonder
> how a "global economy" might be defined for the purposes of simulation. I
> feel too that, in a global simulation, broad political realities would have
> to play a central role. How might they be factored in?
The first dataset will consist almost entirely of numbers from a
variety of sources, and from that I expect to produce what discrete
mathematicians call a weighted graph -- not a chart or picture, but a
network of nodes or vertices linked by edges. We can then apply
connectivity and cutting algorithms to see how well connected the
graph is -- I expect more connectivity than Ed would, but that remains
to be seen.
If broad political realities play a central role, that should be
visible in the results. For example, a program called Metis,
originally written for balancing the load amongst several processors
in a supercomputer can try to divide the graph (or network) into two
(or more) parts of approximately equal size making few cuts or cuts
of low weight only. In cold war days this probably would probably
partition the graph into the well-known East and West blocs, but
now, well, who knows?
> I would suggest that, in a simulation of the kind being proposed,
> it matters a great deal what kind of overall global world is being
> assumed, since the nature of that world would determine who provides
> economic support to whom, who is willing to sell strategic resources
> to whom, who provides weapons to whom, and other such things.
I want to make as few assumptions as possible. I don't want to assume
any "kind of overall global world". That should be a result, not
an assumption.
> Which leads me to the issue of whether a model of the "global economy" would
> really be helpful. In a previous posting I asked what it might tell about
> whether China might devalue the yuan, knowing full well that it couldn't
> tell us much. But perhaps it could tell us quite a lot about the
> consequences of yuan devaluation. ... [much omitted] ...
> But in doing this, it is probable that we would get down to a level too
> micro for a global model -- or the global model would have to terribly
> comprehensive.
Initially the model will indeed have to be "terribly comprehensive",
and should err on the side of containing too much data, rather than
too little. I suggest that we just don't know enough to make a
smaller, less comprehensive model. Later, having some results from
a simulation based on a lot of data -- tens of thousands of numbers --
it may be possible to simplify, based on what we have learned.
> Now to the megaphenom stuff - the end of cheap energy, population growth,
> the concentration of population in unsustainable cities, pollution, the
> effects of climate change. Here there is both a very great need for
> simulation and the possibility of doing something useful. But you are no
> longer dealing with the global economy. You are dealing with something much
> bigger and more comprehensive, a little "like the grand theory of
> everything" in cosmology.
I use economy in a very loose sense, including not only economic
matters but other things that affect economic matters -- human
reproduction, pollution, climate change, and so forth. That's one
reason the mode has to be comprehensive. It's also one reason why
this may have to be a distributed simulator that farms out some
sub-problems to other simulations on other machines with other data.
Again, once we have learned enough it may be possible to simplify, but
not until then.
This is a big project, Ed, and there's no way around it. I suspect
it is too big a project for most corporations or governments to
undertake by themselves, and I certainly wouldn't try it if I thought
I had to do it all myself. But the "Linux model" of cooperative
development by volunteers can make it possible. I'll do what I can
with initial systems engineering (documentation) and I'll do a
prototype. After that, I'll stay around as just one person amongst
many. If I do a good enough job with the documentation and
prototyping I am sure others will come on board to share the work.
dpw
Douglas P. Wilson [EMAIL PROTECTED]
http://www.island.net/~dpwilson/index.html
