[From Rick Marken (960703. 0900)]
Martin Taylor (960702 12:10) --
There is "information" the measure, and "information" as in "what I got from
this morning's paper." The distinction is like the difference between
"perception" the numeric value of a neural current and "perception" the
way I see the world.
I agree with this distinction. Information theory is about "information the
measure". The information measure was developed to be intuitively consistent
with "information the perception". I perceive "informativeness" when I read
the paper because it tells me about events that, had I not read the paper, I
would not know about. Of course, the formal information measure is based on
assumptions that make no sense in terms of how perception works; for example,
I don't start reading the paper wondering which of the many possible things
that might have happened last night actually happened. But the formal measure
of information must assume a set of a prioi possibilities, the occurance of
one of which provides "information" proportional to the a priori possibility
of the event.
So "information" exists as a measure of relationship between variables; it
also exists as a perceptual variable; I can perceive "informativeness".
But I argue that information does NOT exist as an actual entity; something
that can be "carried" from one place to another.
Me:
and that no evidence has yet been presented that shows how information
contributes to our technical understanding of control system operation;
Martin:
I'd amend that to say no evidence that you accept.
No. Actually I meant _no evidence_ at all. But refresh my memory, what
evidence did you present, that I did not accept, that shows how information
contributes to our technical understanding of control system operation?
Me:
indeed, we have shown that information theory -- using the reinfied concept
of information as a basis for explaining the behavior of control systems --
leads to mistaken >conclusions about how control systems actually work.
Martin:
That, you have most definitely NOT done. The "mistaken conclusions" are ones
that are arrived at by people who think that information theory deals only
with the input-output relations in a signal channel.
No. The mistaken conclusion is arrived at by people, like yourself, who
claim to understand information theory correctly. The mistaken conclusion is
that a control system must have information (the measure, the perception or
the entity -- whatever) about disturbing influences to the controlled
variable in order to control that variable.
Bruce Gregory (960702.1440 EDT)
I don't see why we could not say that the bi-metalic strips in
the thermostat provide the system with information about the
temperature in the room
I agree. This has always seemed to me like the only reasonable application of
information theory to control. Information theory could be used to measure
how well the perceptual signal (position of the bi-metallic strip, for
example) represents (communicates information about) the environmental
variable to be controlled. The more information about the controlled
variable that exists ion the perceptual signal, the better the system
controls what the _engineer_ wants it to control.
Note that this use of the information measure says nothing about the
factors that influence variations in the controlled variable; there is no
need to measure (and nothing to be gained from measuring) how much
information about any particular influence on the controlled variable
(disturbance) is carried by the perceptual signal.
Chris Cherpas (960702.1541 PT)
Rick mentioned naive realism's relation to PCT (or lack thereof) and I'm
curious as to whether anything has been written about, or would be written
about, where PCT might be positioned relative to philosophy.
Bill has some papers on this. There is a paper on "applied Epistimology" in
LCS and I think there are some other philsophical things in LCS II
In general, are there any philosophies "closest" to PCT
I don't know which philosophy is closest to PCT; I don't know that much about
philosophy, really. I think the basic PCT epistimology comes out of the PCT
model itself. The model shows that all any person can know is their own
perceptions; we have no direct access to the presumed cause of those
perceptions which, in the model, is called "the environment" but it is also
called "boss reality" when we wax philosophical. This sounds like British
empiricism to me.
Of course, if all we know is perception then how do we know that there IS a
boss reality. Didn't Hume say "we don't know"? And didn't Berkeley say that
it's all taken care of by God in the Quad? In PCT, we say that there IS
evidence of a boss reality; the evidence is the fact that there are real,
consistent constraints on how we can control our perceptions. We can't open a
door just any old way; in order to produce the perception of an opened door
only certain actions will work -- turning the handle with the appropriate
force in the approapriate direction, for example. These constraints on how we
can control perceptions don't tell us _what_ boss reality is; but they
strongly suggest that we are, indeed, operating in the context of a
consistent boss reality.
Science has systematically examined the constraints on how we can control our
perceptions -- that's what scientific experiments are; tests of what we have
to do to get certain perception, or tests of what perceptions we get when we
do certain things - - and invented models to explain these constraints. These
models (according to PCT) are our current best guess about the nature of boss
reality.
There is also a little Kant-ishness in PCT (as I understand Kant). The PCT
model assumes that the brain is designed to perceive the world in certain
ways; there are _classes_ of perceptual experience (intensities, sensations,
configurations, etc). These classes of perception seem to bear a family
resemblence to Kant's "categories" of knowledge. It is this aspect of
PCT epistimology that seems to conflict with "naive realism". While PCT does
assume an external reality (without committing to a particular model of that
reality) it suggests that it is naive to think that what we perceive
_corresponds_ to what is out there in that reality. This doesn't mean that
perception is a hallucination; it means that perception is a _mapping_ of
whatever is "out there" into what we experience as being "out there". We see
the world in terms of certain colors for example; but now we know that what
we experience is a rather non-linear mapping of what we think is out there --
eletromagnetic energy; what we see as the same color experience can be the
result of many different physical realities (different combinations of
wavelength). So the perception of "green" is based on something that's really
out there; but "green" -- the unitary perceptual experience -- is not what's
really out there.
Best
Rick