RE: PhD-thesis on Observational Selection Effects

From: Bostrom,N (pg) <"Bostrom,N>
Date: Mon, 24 Jul 2000 17:35:13 +0100

Jacques Mallah wrote (some time back):
                
>It has applications in cosmology, evolutionary biology,
thermodynamics and
>the problem of time's arrow, game theoretic problems with
imperfect recall,
>the philosophical evaluation of the many-worlds and
many-minds
>interpretations of quantum mechanics and David Lewis' modal
realism, and
>even for traffic planning.

                    I would have liked to see some discussion of each
application. I guess
                that's for the book.

I suppose so. I haven't at all resolved the target-audience problem yet. For
instance, you complain about the wordiness, but many philosophers I talk to
think that it is much to brief in many places. Different cultures; hard to
satisfy them all. Also remember that few people will have spent as much time
as you have thinking about these issues prior to reading the text.

                Also, you could have included the stuff you and I
                discussed about how the using the MWI prevents some of the
counterintuitive
                consequences (e.g. regarding Adam.)

If I remember correctly, I was not convinced that MWI prevents these
consequences. Even if it did, I don't think it would be very relevant since
we are concerned with what is a sound methodology, something which, one may
think, should be independent of specific assumptions about the world. In
particular, If one is advocating a general principle (such as SSSA-R) then a
single genuine counterexample, even if it is hypothetical, would suffice to
refute it or force us to restrict its scope.


>However, we discover new consequences of SSA that are more
counterintuitive
>than the Doomsday argument. Using these results, we
construct a version of
>SSA that avoids the paradoxes and does not lead to the
Doomsday argument
>but caters to legitimate methodological needs.

                    Here's the part I really object to. Your SSA was vague
(you didn't even
                discuss observer-moments up to that point), but was OK as
far as it went.
                When you introduced the SSSA, you went off track. Your SSSA
gives obviously
                wrong results, such as not leading to the Doomsday argument

On the contrary, SSSA leads to DA just as surely as SSA. It is only when
going to the relativized version of SSSA, i.e. SSSA-R, that one can avoid
DA. Funny, btw, that you think that not leading to DA should be considered a
fault! Most people seem to regard DA as a reductio ad absurdum (which I
think is unfair, although DA may be rather counterintuitive).


                and in the
                "god's coin toss" where it gives the wrong Bayesian
probability for the
                coin.

Of course, what is the "right" and "wrong" probability is precisely what is
at stake here. (I presume you are referring to SSSA-R?)

                Nor is there any theoretical justification for not regarding
all
                observer-moments to be in the same reference class. (As you
can tell I am
                partial to the ASSA.)

The main motivation is that it avoids certain consequences that most people
regard as highly counterintuitive and unacceptable.

                    You might then ask how I would avoid the Adam-style
problem. The answer
                is already somewhere in the paper: if Adam is a freak, we
don't need to
                worry much about him. It's in the nature of Bayesian
probability that for
                some people the method must backfire.

I do say that, yes, but if we have a nice alternative then this problem can
become decisive. The thing is not so much that Adam would have been mistaken
- who cares, except himself? - but rather that we don't seem willing to
follow the rule that we prescribe for Adam. If we are not willing to do that
- in all cases where the rule says it should be applied - then we don't
really accept the rule. That is, we don't really accept the unrestricted
applicability of SSA. The question then arises, what rule do we accept? And
I suggest that the answer is SSSA-R.


Nick Bostrom
Dept. Philosophy, Logic, and Scientific method
London School of Economics
Received on Mon Jul 24 2000 - 09:47:43 PDT