[om-list] Limits of Prediction in the Natural and Social Sciences

[Thomas L. Packer at home]

Interesting.

    I think agent-based models have a chance at improving this kind of
modelling.

tomp

~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Víðar sum quem nihil obstat.
Omnia apud me Mathesis fiunt.
www.Ontolog.Com
~~~~~~~~~~~~~~~~~~~~~~~~~~~~


----- Original Message ----- 
From: "Mark Butler" <butlerm at middle.net>
To: "One Model List" <om-list at onemodel.org>
Sent: Saturday, 04 October, 2003 08:47
Subject: [om-list] Limits of Prediction in the Natural and Social Sciences



Most physical models treat the idea of intelligence as an irrelevant
emergent property of deterministic systems at best.  However, the
presence of intelligent free agents in a physical system must be
something far more subtle more than quantum mechanical randomness, for
two reasons, first fundamental randomness in an intelligence is
completely incompatible with the concept of moral responsibility, and
second, fundamental quantum mechanical randomness merely accelerates the
apparent heat death of the universe, an eventuality we can hardly take
seriously.

Free will in man (let alone God) cannot be a random or arbitrary
fluctuation inside a deterministic machine, as many neurophysiologists
would have it.   Nothing incapable of increasing the specified
complexity of a system can be considered truly intelligent.
Deterministic mechanisms, at best, can only preserve specified
complexity.  Chaotic mechanisms, albeit productive of strikingly
beautiful patterns (those created by cellular automata, for example),
never increase specified complexity, either.

The metaphysical issue concerns the limits of predictive knowledge about
the future.  Leaving the questions of divine knowledge and power for
theology, what are the limitations of predictive calculations about the
future based on partial knowledge of the way it is in the present?  The
limits of predictive projection of macro-level social behavior years
into the future is a serious policy issue for government and private
parties alike. The market or natural price for all forward looking
financial instruments - public or private - including bonds, equities,
insurance, taxes and so on is based on a long term projection of future
conditions by the parties concerned.

Current econometric models assume social stability and rational behavior
- no major wars, no cultural decline, no moral collapse.  But a brief
examination of history abundantly demonstrates that moral considerations
are primary in the rise and fall of both enterprises and civilizations.
Rational behavior is a key assumption of the Black-Scholes equation for
the market price of futures - under adverse conditions, however, market
participants do not behave rationally (famously bankrupting the
centuries old Barings Bank during the Asian currency crisis a few years
back).

But might we make much better predictions if we thought of people as
imperfect moral agents, disregarding moral, prudential, and legal
constraints on their behavior in proportion to the weakness of
character, rather than assuming rational self interest and perfect
obedience to law?    If we included an estimates of the level of
morality of the traders, investors, regulators, etc, we ought to get
much better econometric models - ones that can predict market behavior
under conditions like the Great Crash of 1929 much better than we do
now.  More importantly we ought to be able to make much more reliable
predictions about the fate of nations and civilizations in moral decline.

We do this now, of course.  Nations that neglect moral considerations in
the behavior of other states and countries do so at their own peril.
The behavior of a single prince or potentate is hard to predict, but
analytical models could tell us about the behavior of a market economy
under laws and moral traditions different from our own.  Surely a
generalized banking failure has a highly different character in Japan
than it does in the United States.  One of the major challenges for the
Japanese is that most economic research has been done in the context of
the Western tradition.  Japan, despite all appearances in matters of
commerce, has an entirely different one.

Utility theory is a well developed analytical theory of behavior that is
easy to implement. If you want to predict behavbior in context of
philosophical and political concerns, however, you need to model thought
processes, not individual thought processes, necessarily (much too
hard), but rather collective thought processes.  You cannot predict
model international economic behavior under war conditions very well,
for example, unless you keep track of the basic moral reasoning of the
people in each country.  Political concerns flowing from those moral
considerations are paramount in wartime.

This flows into game theory, where strategy can be modelled, but long
term prediction is difficult.  Fundamental questions of diplomacy
revolve around projections of victory in any international dispute, so
anything would be helpful.  I don't mean such mundane things as
preserving your foreign investments, either, but rather fundamental
questions of life and liberty for people throughout the world.

Virtually every serious public policy question revolves around these
considerations.  How big should the defense budget be? How much internal
security is necessary?  How much regulation of private commerce is too
much?  What about banking regulations? securities, foreign trade,
monetary, non-profit, education, and the environment?  How much does a
minor decline in average morality affect the stability of a complex
civilization (sometimes catastrophically, apparently)?  How much would
an increase in morality reduce the need for regulation?  What are the
likely unintended consequences of any proposed regulation or
deregulation on something as complex as the electrical power grid?

The most critical aspect of any model is how you keep track of what you
don't know and the degree to which you do not know it. A proper
treatment of entropy and ignorance using information theory,
probability, statistics, fuzzy logic, etc. is fundamental to any
scientific prediction about the future.  That and natural law are
usually sufficient for natural science, but by adding an analytical
model of basic moral considerations to now common models of economic
self interest, ought to greatly increase the utility of analytical
models in the social sciences.

Comments?

- Mark



_______________________________________________
om-list mailing list
om-list at onemodel.org
http://www.pairlist.net/mailman/listinfo/om-list