Re: a baysian solution

From: Nick Bostrom <>
Date: Sat, 18 Apr 1998 14:56:18 +0000

Wei Dai <> wrote:

> On Thu, Apr 16, 1998 at 08:23:33PM +0000, Nick Bostrom wrote:
> > But A and B can stand for *classes* of possible universes. In every
> > univese of class A, the coin landed heads (say), and in every
> > universe of type B, the coin landed tail. Each of these possible
> > universes contains a person called Nick. Suppose I know that exactly
> > one of these universes is the real one, and that I am called Nick.
> > Then I can formulate the question: "What is the probability that the
> > person called Nick should have a birth rank <=100, given that the
> > real universe is one of type A, i.e. one where there are 200 people?"
> > Now, this question cannot be ruled out on the grounds that
> > probability assertions including personal pronouns are meaningless --
> > for it contains no personal pronouns. Yet, this is enough to set the
> > Doomsday argument rolling.
> How do you know that every possible universes contains exactly one person
> called Nick? Suppose you didn't, then the percentage of type B universes
> that contains a person called Nick would be about twice the percentage of
> type A universes that contains a person called Nick (since a type B
> universe has twice as many people as a type A universe). Since these are
> exactly the universes that would survive as possible universes after you
> learn your name, this effect would cancel out the DA.

I think that is wrong. It presupposes that I regard the name "Nick"
as being randomly choosen from the set of all possible names. When I
discover that I am called Nick, I would then get reason to believe
that many people are called Nick. --But "Nick" is not randomly
choosen from among all possible names. The only reason why "Nick" was
choosen as a sample was that there existed somebody with that name
who took his own name as a sample. And if you choose the name from
the set of all actual names (names of living persons), then you don't
get any information from finding that that name is instanciated.

Say that the percetage of B universes containing somebody called
"Nick" is twice that of A universes. But the same would hold for any
other name. Since I know that I have one name or another, then, if
the above reasoning were correct, I could infer that the B universe
was twice as likely as the A universe, even before I knew anything
about my name. This is equivalent to accepting the Self-Indication
Axiom. I show in my Doomsday-paper that there are serious problems
with accepting the SIA.

(The difference between finding your name and finding your position
in time is that you could have found that you position in time had
been incompatible with there being 100 people, but you couldn't find
a name that would have been incomplatible with either hypotesis.)

So I don't see how suspending probability judgements over indexical
propositions can obviate the Doomsday Argument.
Nick Bostrom
Department of Philosophy, Logic and Scientific Method
London School of Economics
Received on Sat Apr 18 1998 - 07:03:43 PDT

This archive was generated by hypermail 2.3.0 : Fri Feb 16 2018 - 13:20:06 PST