[ExI] against Many Worlds QT

[Lee Corbin]

Stathis writes

> The author seems to be arguing that many worlds should allow the
> possibility of using all sorts of strange utility functions not
> available to single world theories, and that this damages the case for
> using decision theory to derive the Born rule in MWI, since one such
> attempt to do so requires the (unjustified, it is claimed) adoption of
> mean utilitarianism.

Thanks for the summary (on top of Giulio's, of course).
I don't have time to read the paper, so I appreciate it.

AS for that last sentence, we should go for kind
utilitarianism instead, of course.

> The author also seems to have a problem with how
> probabilities are deduced by observers in the MWI (or any multiverse
> model). He points out that any observer following a sequence of random
> events like coin tosses will conclude that the pattern he observes is
> indicative of a particular bias that might not be there: someone will
> definitely observe a sequence of 1000 heads, and that person will
> conclude the coin is heavily biased, even though it isn't.

More seriously, I agree with the author, at
least narrowly, but I guess I don't see what
he is complaining about here. Whether or not
you believe in a single strand universe, or
MWI, an unlikely string will give rise to
peculiar and probably inappropriate conjectures.

> But I don't see how this is different to
 > single world probability if you have a
> large number of non-communicating experimenters
 > tossing coins: a small number might conclude
 > that the coin is biased, but most will get
> approximately 50/50 heads/tails, just as in the MWI.

Oh. Right. You took the words right out of my mouth.

> The author discusses model multiverses which we could create with a
> branching computer program, and uses these to support his contention
> that probability doesn't work in the MWI. But *something* has to
> happen in these models, if they are physically possible. That is, the
> observers would come up with some sort of physical theory and account
> of probability. It doesn't do to throw up one's hands and say that
> probability becomes meaningless in these cases.

Well, not having looked into it with near as much
patience and effort as you, FWIW I concur.

The major looming problem in my own ontology here
is that Bayesian statistics doesn't seem to jibe
with MWI no matter how many times I've reread how
Jaynes and others try to explain it.

In MWI, there is a definite fraction of outcomes
to any situation: there really is (in MWI) a group
of identical universes (a phrase David Deutsch uses
over and over in "The Fabric of Reality" and which
I abbreviated as GIU long before knowing our friend
GIULIO).

Then if a photon has a 50% amplitude to go straight
or to go up, it results in an objective 50/50 split.
(More likely, of course, in a 49.37118561917009...
split.) Thus probability on MWI seems to me to be
completely objective---which flies in the face of
the received Bayesian wisdom, which also on other
grounds seems correct.


As for whether accepting MWI makes any difference in
life, I think that it does. But not much.

One interesting result is that a human being tends
to obtain a more mature outlook on practical realities
sooner under MWI. For example, a teenager who commits
a very dangerous act (that his mother happens to see)
will respond to her tongue-lashing: "well, Ma, look,
nothing happened!" She understands that that isn't
all there is to it.

But in the multiverse, "you can't beat the odds" [1].
The young man is forced to acknowledge that the
dangerous activity resulted in a possibly very
great diminution of his measure throughout the
many worlds, and to realize that what he did was
unwise after all---just as his mother said.

Lee

[1] The exact phrase "In MWI, you can't beat the odds"
I owe to Robin Hanson, who was visiting me twenty
some years ago and who captured in this inimitable
motto the notion that I was trying to explain .