Re: Decision theory

From: Gale <wmgale.domain.name.hidden>
Date: Tue, 05 Jan 1999 11:56:56 -0500

Jacques Bailhache wrote:

> I wonder if there couldn't be another possibility, neither determinism, nor
> randomness, but free will, which I agree to be something rather mysterious.

> This idea of description levels seems very important to me. This is perhaps
> the key idea for understanding the spirit and its properties, consciousness,
> free will and creative intelligence.
> My idea is that the universe, and the living beings in particular, could be
> made of an infinity of material organization levels : body, cell, molecule,
> atom, particle, quark, ..., with different physical laws at each level.

What is needed is an operational definition of "free will".
I want some way to tell from *external* observations of
a system whether it has free will. Can one come up with
a definition such that humans show free will but electrons
do not? I want to be able to apply the definition to a
mechanical intelligence.

It seems to me that this is likely to involve some notion of
unpredictability. Since I do not want to include electrons
as having free will, the unpredictability has to go beyond
a statistical predictability.

I agree that the levels notion is important, but would emphasize the
emergent properties aspect rather than the compositional aspect.
Thus molecules display huge variety which is barely present at the
atomic level. Cells display self-reproduction which is barely present
among single molecules. Multi-cell beings have such emergent properties
as self awareness, intelligence, and values.

A point of an operational definition of free will would seem to me to be
that the values and intelligence of a person are not directly
observable,
but are inferred from their past actions. Having some notion of their
values, some predictability is possible, but lack of precision in
measuring values leaves a large amount of unpredictability. Since we
are
including autonomous agents for which establishing precise starting
conditions is not possible or is ethically inadmissable, then this could
only be tested on ensembles of beings.

Problem: suicide would appear to be statistically predictable
for ensembles of people, yet surely this is an act (the ultimate act??)
of free will. Or is it? Well, it has to depend on age, culture,
status within the culture, ... Age is rather clearly not a direct
determinant, but an indicator for other variables not easily measured.
So if there is only a residual variance after taking into account
explanatory variables, then one might want to argue that
not all of the suicides *had* exercised free will, even if they
could have. But one would want to appeal to the residual variance
to argue that free will was shown by some of the suicides
(and non-suicides), and potentially available to all of them.
Thus free will is shown if we cannot build a model that accounts
for both the mean and variance from the single prediction of a
probability for an ensemble.

I believe this will exclude electrons (prepared spin up in x direction,
spin measured in y direction), but I'm not at all sure it excludes
amoeba. And it seems a rather weak concept if it is manifested by
single cell creatures, so there are likely some basic ideas missing.

Cordially,
Gale
Received on Tue Jan 05 1999 - 09:00:14 PST