RE: PhD-thesis on Observational Selection Effects

From: Bostrom,N (pg) (
Date: Sat Jun 17 2000 - 08:30:38 MDT

Eliezer wrote:

> Thus, for example, the Doomsday Argument is out because it would have
> predicted the end of the world shortly after the Industrial Revolution.
(This is answered in DA-FAQ at the site.) It is in
the nature of probabilistic reasoning that in untypical circumstances people
who use it will be misled. If the human species survives for a very long
time, then the people living shortly after the Industrial Revolution will
have been in a highly untypical situation; they will have been misled by the
Doomsday argument, but they will not have been irrational to accept it (if
there were nothing else wrong with it - which, however, I argue that there
is). In order to evaluate something like DA you'd have to look at its net
effect, not just its effect on a few highly atypical observers. Now, most
observers (nine tenths) will be among the middle 90% of all observers who
have ever existed. If every observer thinks, in accordance with DA, that
there will probably not be hugely many more observers after her than before
her, then the vast majority of all observers will turn out to have been

> The Everett-Wheeler many-worlds theory is actually simpler than
> contemporary theory - it omits state-vector reduction - yet still
> predicts the existence of our world. The Modified Razor strikes
> Many-Worlds down because it predicts our world much less uniquely than
> the modern theory of physics.
This relates to the problem of measure in many-worlds interpretation. The
MWI could have a good fit to data, if it could be showed that it assigns a
high probability to a relatively large fraction of all observers who are in
the same reference class as ourselves observing what we observe. The case is
very similar to multiverse cosmologies - e.g. chaotic inflation, where lots
of real universes exist with different boundary conditions and physical
constants. How our evidence favours or disfavours such models is something I
discuss in detail in the thesis, and the same arguments can be applied to

> Use of the Anthropic Principle is permitted, but you must demonstrate
> that a sufficiently large number of samples exist - otherwise the theory
> would not predict the existence of the exceptional sample
That's correct: in order to explain the fine-tuning of our universe by
appealing to an anthropic principle, you need a suitable ensemble of actual
universes out there ("possible worlds" won't do). But evidence for
fine-tuning is a reason to think that there is a multiverse.

> Nick Bostrom
> Dept. Philosophy, Logic and Scientific Method
> London School of Economics
> Email:
> Homepage

This archive was generated by hypermail 2b29 : Thu Jul 27 2000 - 14:13:27 MDT