Re: Raymond Smullyan

From: Marchal <marchal.domain.name.hidden>
Date: Thu Aug 5 03:03:43 1999

Wei Dai wrote:

>On Thu, Jul 15, 1999 at 04:38:05PM +0100, Marchal wrote:
>> Wei Dai, you should read cautiously SMULLYAN'S "An Epistemological
>> Nightmare".
>> You will understand at once that, with comp, there is no hope to use any
>> "scientific theory" to build a "decision theory".
>
>Ok, I just read it, but I don't see how it rules out a decision theory. The
>point of the article seems to be that a person can be wrong about his own
>subjective experiences, i.e. think or believe that he experienced something
>when in fact he experienced something else. Current decision theory doesn't
>take this possibility into account, just like it doesn't take into account
>the possibility that a person makes a mistake while trying to compute
>probabilities. I see no reason why both won't eventually be part of a more
>realistic decision theory.


I guess you are right. My point was a little bit more metaphysical.
What I still not understand is the relation you seem to infer between
a "realistic decision theory" and the "search for a TOE".
Having a TOE, a decision theory can be useful, but having any knowledge
entails that. If I 'discover' that the earth is round, it can affect my
decisions concerning the way to explore the world. But the roundness
of the earth is independant of the 'decision' theory.
My feeling is that a 'good' decision theory should be independant of the
philosophical framework, and cannot really constitute a philosophical
framework by itself.
As a matter of fact I belong to a AI-laboratory where a lot of people
works on automated decisions in the physician's frame. What to do with
incomplete information ? How to handle ignorance. What to decide having
different diagnostics by equivalent expert ? Most of them have
completely abandon "probability" theory. They prefer the Dempster-Shafer
theory of evidence also known as belief function theory. The handling of
ignorance is rather special and different from probability. The very
nature of a belief (unlike knowledge) is that a belief can be wrong.
I think that any practical decision theory must take that into account
as something basic.
I think that too for any 'theoretical psychology'. It cannot be normative.
My intuition (coming from computer science and provability logic) is
that decisions are essentially personal, and never completely justifiable.
Probably I misunderstand the relationship you make between "decision
theory"
and "philosophical framework". If you can elaborate on this or refer me
in the archive (or elsewhere), that would help me.
(BTW my philosophical framework is the computationalist hypothesis).

Bruno
Received on Thu Aug 05 1999 - 03:03:43 PDT