🚀 go-pugleaf

In article 21160@unocal.com, stgprao@xing.unocal.com (Richard Ottolini) writes:
> There are serious mathematical flaws with this model.
> It was the fad in the 60s and 70s to use "systems theory" where first-order
> difference equations link state variables.  The solution to most of these
> equations are exponentials in time.  Therefore, given enough time each state
> variable takes one of three values: a constant, zero, or infinity.
> That is why Rome models always spells doom, because the simplistic mathematics
> goes to zero or inifinity.


The flaws are not "mathematical" -- the systems dynamics models are internally
consistent logically and computationally. The flaws lie in their consistency with
the "real world".

The problem pointed out by Forrester and Meadows and their associates is that,
although their models have shortcomings, which they themselves acknowledged,
so do the mental models used by human beings to make decisions about the subject
of the models. And the mental models of most people are even worse than systems
models like their first efforts. Human beings have a lot of trouble tracking the
complex feedback loops that operate in real systems. The result is that policy
deicsions guided only by mental models are often "counterintuitive" -- advocating
policies that will have the exact opposite effects their intuitive models expect.

The Club of Rome sponsored models were an attempt to begin a long process of
developing models that might overcome the limitations of our mental models and
improve the quality of our decisionmaking.

Obviously, they did not catch all the minor feedback loops that buffer real
trends -- or their model would not have yielded some of the wild swings that it
did. However, none of the factors modeled lacked countervailing factors that,
if modeled accurately, could have been expected to result in exponential growth
or decline.

It is not an answer to a model or a modeling effort that its correspondence to
reality is imperfect. The challenge is to come up with better models -- models
that incorporate all the knowledge on which the previous ones were based, and
add it that knowledge. Hopefully, with time and better analysis, we may arrive
at models that have some predictive value with better than decimal-order-of-
magnitude accuracy. However, such accuracy may not be necessary to guide the
decisions we make. It makes little difference whether the model predicts a
collapse in forty years or in two hundred if what we have to do to avoid collapse
is the same in either case.

The challenge to readers is to come up with better models, and let them speak
for themselves.

---

starflt@uunet.uu.net, jdr@starflight.corp.sun.com
Jon Roland
Starflight Corporation, 1755 E Bayshore Rd #9A,
Redwood City, CA 94063-4142, 415/361-8141