[extropy-chat] Role of Observer is not Relevant

Jef Allbright jef at jefallbright.net
Fri Apr 6 22:22:17 UTC 2007

On 4/6/07, Rafal Smigrodzki <rafal.smigrodzki at gmail.com> wrote:
> On 4/6/07, Jef Allbright <jef at jefallbright.net> wrote:
>   If I may ask another calibrating question:  Do you
> > have an opinion on the validity of subjective Bayesian probability?
> >
> ### Can you expand on the question?
The philosophical foundation of Bayesian probability highlights the
ultimate subjectivity of any observer (although embedded in a
consistent, but only indirectly knowable reality.)  That is the view I
am promoting.  I figured that it would provide another highly relevant
data point to get your view on that topic.  I thought there was also a
chance that it might trigger in you a better understanding of my
position, because it's clear to me that I'm not making myself clear to

> > My POV is that *every* agent is necessarily at the "conceptual center"
> > of their universe, and by recognizing this one forms a more accurate
> > model of "the way things work", formerly known as "reality."
> ### Is there any thing else than the "model"?
That depends very much on the context of your question, which isn't
clear to me.  In the view I am promoting here there is an agent, with
a model of its "reality" and its interactions with that local reality.
 It seems that your view has much the same, with the ontological
addition of a "platonic plenum" containing a hyperinfinite set of
somehow "real" entities.

> > Since understanding is essentially modeling, at various levels of
> > abstraction, it seems obvious to me that a model gains nothing (and
> > necessarily loses by misallocating its probability mass which must sum
> > to unity) by positing entities for which there is no evidence.  This
> > is not the same as denying the possibility of other entities, (indeed,
> > acknowledgment of the inherent incompleteness of any model implies the
> > existence of entities outside the model) but only saying there is
> > nothing to say about them, so for *all* practical purposes, they don't
> > exist.
> ### Do you think that there are stars that are too far from us to ever
> reach the Earth, given the expansion of the universe? If yes, what are
> their spectral characteristics? Are they the same as the
> characteristics of local stars? Different? In principle unknowable?

I would certainly infer that there are stars beyond our observation,
and (if Gordon gts wasn't watching) I would apply the Principle of
Indifference and infer that their properties are distributed similar
to the stars that we do observe.  But this gives me absolutely zero
new information about any of those hypothetical stars, and this is

But it appears that you take this even further, and believe not only
that those hypothetical stars exist, but their specific mathematical
coordinates exist in some sense independent of any observer.

If there were a row of boxes, and inside each box I found a ball, then
(all else being equal) with increasing number of boxes I would develop
increasing confidence that a subsequent box held a ball, but it could
easily be empty. My inference doesn't provided any information about
the actual existence of a ball in the next box.  I would claim a high
probability that there is a ball in the box, but I wouldn't be
justified in claiming that there IS a ball in the box independent from
some observation, no matter how indirect.

> If you really believe that your location is not special, then you have
> to ascribe the same spectral characteristics to stars in your vicinity
> and stars that are too far be seen, even in principle. And if they
> have spectral characteristics, they exist.

As argued above.

> You cannot say "there is nothing to say" about entities that are
> entailed by the existence of known entities.

Inductive inference does not entail entailment.

Is this perhaps the crux of our disagreement?

> > The logical incoherence in your rendition of platonism may be more
> > apparent if we point out that by positing the "existence" of
> > unobservable entities, we must admit that there's nothing to
> > distinguish between highly probable unobservable entities and highly
> > improbable unobservable entities.  Therefore, it seems to me, the
> > "platonic plenum" amounts to a meaningless mush.
> ### Do you think the likelihood that unobservable stars have the same
> spectral characteristics as observable ones is identical to their
> likelihood of having any other arbitrary characteristics?
> Knowledge about the observable universe informs you about the
> unobservable parts as well. You can and should have an opinion about
> the probabilities of various unobservable entities (i.e. measures of
> the relative sizes of the parts of the plenum that make up these
> entities).

It appears that the crux of our disagreement is your implication that
inference adds information.  Did you ever review our earlier
disagreement about Occam's Razor?  It might be related.

> I can suggest to you "The End of Time" by the theoretical physicist,
> Julian Barbour for a discussion of the plenum.

I read this several years ago, and a similar concept in _Time's Arrow
and Archimedes' Point_ by Huw Price a few years earlier.  The "block
universe" concept is intriguing, mainly because it aims to reduce the
number of ontological entities by eliminating time, but I remember Huw
Price in particular, explicitly backing away from what this might mean
in subjective terms.

- Jef

More information about the extropy-chat mailing list