> From: "Kevin S. Van Horn" <ksvhsoft@xmission.com>
> I don't know that Cox's views are really relevant. Ideas must stand or fall on
> their own merits, not on whether they conform to some ancestral viewpoint. The
> only relevant question is whether Cox's Theorem, or some derivative argument for
> "the obligatory use of single-distribution probabilistic belief models," is in
> fact a compelling argument for this position. I myself find Jaynes's arguments,
> derived from Cox's and Polya's work, to be very convincing.
I agree completely with Kevin that Cox's views are irrelevant here.
I don't find Jayne's arguments any more convincing than Cox's though.
Take a look at the note I referred to in my message
(http://www.cs.cornell.edu/home/halpern/papers/cox1.ps or .../cox1.pdf).
There I list a number of sets of hypotheses that are strong enough to
force a probabilistic belief model. I don't find any of these
particularly compelling (although I'll admit that "compelling" is in the
eye of the beholder). I'm not saying that these are the
only possible sets of assumptions. However, anyone who subscribes to
arguments like this needs to list carefully exactly what the assumptions
are. Very few people seem to have done that (and most who have have
omitted some important assumptions). (As an aside, I'm a fan of
probability, even though I don't find these particular arguments convincing).
-- Joe