[From Bruce Buchanan 950222.15:20 EST)]
I know it is said that Fools rush in where angels fear to tread, but most of
the angels on the csgNet are of the 2nd or 3rd degree so I am going to take
my chances ;-). Here are a few ideas which I think may be relevant to
ongoing discussion, intended to elicit corrective feedback if possible.
This is all in the spirit of Martin's comment that Niels Bohr was on the
right track when he said: "Take everything I say as a question."
<[Bill Leach 950218.22:39 EST(EDT)] writes:
I keep thinking about Martin's assertion "that there is information about
the disturbance in the perception". In the sense that one could look
ONLY at the perception; without knowing anything else other than the
subject signal IS a perception for a control system, then I posit that NO
there is no information about the disturbance in the perceptual signal.OTOH, as soon as you begin adding specific knowledge about the control
system itself, the possibility of discerning "information about the
disturbance" increases above zero.
[Martin Taylor 950220 13:30]
In past discussions, I have taken the position that it is wrong to say
information theory is inapplicable. The interesting question is whether
application is useful, and to that question I have an open mind. I tend
to believe it may be, and Rick's latest message may provide a route toward
an answer. The problem has been that (as I have throughout argued) in a
single-loop control system, especially a linear one, there are much better
analytic tools, and information theory is unlikely to tell you anything
you didn't know already. But it might...
[From Bill Powers (950221.2210 MST)]
Brief comment on information in perception. In order to detect the
information in the perceptual signal that corresponds to variations in
the disturbing variable, it seems to me that it would be necessary to
know the behavior of the disturbing variable by some means other than
through the perceptual signal. Something like synchronous detection
would be needed -- otherwise there would be no way to detect a component
correlated with variations in the disturbing variable.This means that while the _analyst_ might be able to calculate what
proportion of the perceptual signal's variations are due to the
variations in the disturbing variable, the control system itself could
not do that because it has no separate channel that tells it what the
disturbance variations actually are. So even if the information exists,
somehow, in the perceptual signal, there is no way the control system -
can find out what it is.
In reflecting upon this discussion, as well as other comments by Rick, I
make the following observations:
1. While control systems can be considered hypothetically in isolation they
cannot be so in fact. At the least there are agencies which have created
and continue to maintain e.g. neural systems i.e. the metabolism of
neurons, etc., and a matrix of body functions and other information
sources.
2. In principle it may be indifferent whether the control system is being
observed externally by an experimenter, or from within the organism itself
by a coordinate part of the hierarchical control systems. (We do not need
to imply or consider consciousness for this.)
3. A clear distinction should be made (it probably has been many times and
I have missed it in csg) between (a) Information Theory per se i.e. in
terms of statistical probabilities and negentropy, as in communications
engineering, and (b) information with semantic or context-derived
discriminatory content, of specific significance to the organisms in
various ways, including information about appropriate criteria i.e.
values, preferences.
4. In addition to observing that the map is not the territory, Kozybski
centered out what he called "time-binding" as a crucial aspect of human
thought and organization. Now, there is an inherent lag or hyteresis in the
basic feedback loop, which provides the structure and functional
relationships, through reiteration in associated loops, for the
possibilities of extensive connections with other complex loops, short-term
memory, and thus the possibility of sustained patterns of reverberating
circuitry which provide the neural grounds for higher order representations
(a la Warren McCullcoch and Pitts). This provides the neural basis also
for consistency of behavior via controlled variables.
5. While the nature of the disturbance cannot be known in an isolated
control loop at the level in question, it is the business of higher levels
to put together perceptual data from whatever sources are available to
identify the causes of disturbances, in semantic terms, to assist adaptive
response. These other control loops are there for the same evolutionary
reasons as any single lower level control system.
6. An implication of all these interacting hierarchical control systems is
that the behavior of part functions are in this way integrated to serve the
needs of the whole organism. This, to my mind, is the explanation of how it
can be that the whole, in organized systems (not just collections of
items), is greater than the simple sum of the parts.
7. All this has further implication, e.g. in relation to the nature and
role of higher values, and I could go on, but is there anything basically
at fault in the foregoing views? I can hardly be said that they are vague,
for they are perfectly clear and exact at the level of organization being
addressed, as I see it anyway.
Cheers and best wishes.
Bruce B.
