[Martin Taylor 940901 12:40]
Avery.Andrews 940901.1544 with comments on Bill Powers (940829.1700 MDT)
In the case of PIFs, we assume, with some evidence but
perhaps wrongly, that there really is a correct, unique description of
brain activity in terms of perceptual signals, etc, and a correct
decomposition of its structure into ECS's with their PIFs, etc. This
is part of the empirical content of PCT (I think). On the other hand I
don't see how this can be confidently asserted of the external world.
Absolutely. See Bill Powers' posting (Bill Powers (940829.1700 MDT))
with most of which I am in complete agreement, specifically with:
There is no entity in the
environment corresponding to the perceptual signal, nor need there be.
There is, we presume, a physical reality out there, and what happens in
that reality affects our sensor organs. What we do with our effectors
affects that reality, and in some way our effects on the reality "come
back to haunt us." Perhaps they come back in the form of control of
perceptions, but perhaps they come back in ways we wot not of.
There are no "intrinsic" entities out there. We don't know what the
world IS made of in any absolute sense, and we don't know what relationships
exist among those whatever-they-may-bes. What we do "know" is that in
our personal experience certain effects tend to occur in our perception
when we do certain things. Other effects just occur in our perceptions
(like the sun rising every day whether we do something about it or not).
And the ways these things relate seem to have been constant or only
slowly changing over long periods of time. We perceive "things" and
"relations" out there, like "monkeys", "trees", and "monkeys 'in' trees",
but the world knows nothing of monkeys, trees, or "in-ness." We CREATE
those things by having perceptual functions for them. But what we can
USEFULLY create, in the sense of being able to control perceptions that
involve them, is constrained by what the outer world is "really like."
Any other position is probably another way of expressing solipsism.
Once we have created a "thing," it DOES exist "out there." It is different
from all the other infinity of possible "things that might be." And for any
one thing created by a perceptual function, all we can know of the
mysterious "out there" thing we have created is the value that is output
by the perceptual function. Of all the uncountable infinity of possible
"things" we might have created, we created this one, and we are stuck
with it until we change it. It does what the "real" world dictates, not
what we dictate, although by acting on the "real" world we may affect
what it does, by means of which we know nothing, except through other
perceptions we have also created.
What we can do, however, is construct measuring instruments of various
kinds, and stick them various places, and record the results. When we
find results that covary with a PIF, then we're identifying a CEV.
We may be coming close to finding a controlled CEV (CCEV), which is a
short way to say "a CEV that corresponds to a controlled perception."
But we can never know when we have exactly identified a CCEV, because
we might not have even touched most of the variables whose values are
CEVs (controlled or not) corresponding to perceptual signals that are
arguments to the PIF for which we are testing.
As Bill P. says: "It is still a physical system, but how we
understand what is controlled is quite beside the point for the control
system itself. The only "CEV" exists in the mind of another observer
looking at the same environment."
Just so. But I disagree with what Bill says just after this:
The CEV is where interactions in the world occur, even though nobody,
nobody at all, knows at any moment what any particular CEV might be.
Well, I disagree with this reification. All a control system has to do
is look at the same collection of physical variables differently, and
they constitute a different CEV. This has nothing to do with the actual
state of the environment; it is a matter of perception alone.
When a control system "looks at the same collection of physical variables
differently" that is another wording for a control system "with a different
PIF." That, of course, defines a different CEV. It is, as Bill says,
a matter of perception alone. The value output by either PIF is, though,
a function of "the actual state of the environment". (Allowing always
for the possibility that some of the inputs to a PIF result from
Back to Avery:
But different instruments may record covarying results (an electric
current in one place covarying with a light-intensity in another),
or the size of a solid angle in the light array at a point covarying
with the distance from that oint to an object, which is why I want
to speak of multiple CEV's associated with one PIF.
Neither instrument might record anything that corresponds to the actual
PIF that you are looking for, which could be something quite different
that covaries with both. The PIF itself will be quite definite--for
example, the logarithm of the ratio between the light intensity averaged
over one region and that averaged over a surrounding ring. If you knew
it, you could set up your instruments to measure the physical correlates
and make the appropriate calculations to give you an estimate of the value
of the CEV. But you don't know it, and the best way to find it out is
always to apply The Test to various guesses you may have.
The better your guessed CCEV is controlled by your experimental subject,
the closer you are to finding a PIF. But many of your choices
will show SOME level of control in the test. In the example, if you
guessed at the simple ratio between the patch and the ring, you would
find that there is very good control, but the quality of control would
vary with the value of the reference level for the perceptual signal.
If there was a PIF for x^2 + y + z, its output would covary perfectly
with your measurements of y+z, and pretty well with x+y+z if you didn't
vary x too much. Your subject would exhibit perfect control under The Test
when you affected the CCEV by manipulating y+z, and good control when
you manipulated x+y+z, but neither would correspond to THE single PIF
you are Testing for.
Close, but no seegar...
And seegars, don't "exist" in the outer world, so you'll never find a
"physical" controlled CEV that you can be sure of. But if the real
world exists, so do real controlled CEVs. (And I repeat from before:
ANY function of physical variables in the real world is a CEV, which
comes back to the point Bill correctly hammers on, that there are no
intrinsic entities out there--that we could know).
As for layered protocols, so would it be reasonable to regard the
grammar of a language as a consequence of the functioning of a
cluster of protocol nodes, wereby the pn's involved in speaking
French normally call on each other, and likewise for English, but
it is also possible to violate this, as in `code-switching'. For
example a pn for describing events in French might call upon one
normally used for identifying individuals in English. If so, this
would be useful, since there is a reasonable amount of actual
literature on what does and doesn't happen in code-switching.
Simply put: Yes. But I think there would be quite a lot of supporting
nodes that would be common in any case, no matter how completely bilingual
the person might be.
Checking what does and doesn't happen in code switching might be a neat
way to tease out the control systems that are used in normal language.