[Martin Taylor 991020 17:50]
(Mark Lazare 991020 1145 am) in response to Martin
Mark: agree with you thus far, but this next part......there is a change in
your post from "satiety" as the controlled perception, to "hunger" being the
controlled perception.
No. "hunger" is a value of satiety from "low" to "none". "Replete" is
a value of satiety that we could call "high".
Taylor:
Often the reference level for it is "no hunger" so that when hunger is
perceived, there is indeed an error signal. But the error signal is
not itself perceived.
Mark:
the reference level for it (satiety) is "no hunger." Satiety, I see as the
Reference and therefore the lack thereof would result in error -- the feeling
of hunger.
You can't have a "lack of satiety" unless you are missing some kinds of
sensor systems. You can have low satiety (= hunger) moderate satiety
(= something for which we don't have a name, perhaps "comfortable
fullness"), or high satiety ("being stuffed"). But you can't have a
non-value for it. You can have a reference to feel high satiety, low
satiety, or something in between. You can perceive satiety to be high,
low or in between. If the value perceived is different from the
reference value, you have error in the satiety control system, which,
in standard PCT, feeds into the output function, _not_ up the perceptual
control hierarchy.
Which, yes can be perceived, but not controlled for at that same
level. ERROR is bi-directional -- most commonly and most profoundly it
effects output. However just like in MOL error goes up the hierarchy as
residual cumulative error -- the cumulative error is sensed as input
perception by the re-organizational system -- but the is separate from the
controlled perception.
Yes, error signals are hypothesized to be sensed by the reorganizational
system. But it seems reasonable to believe that if one has survived to
the ripe old age of a few months or a few years, one has managed to
reorganize at least to the extent of knowing a few ways to influence
the perceived level of satiety. I don't think reorganization really
comes into the issue at hand--which is how it comes about that a low-level
error is corrected through the action of high-level control units.
Taylor:
A perception is not (in standard HPCT) an error signal. There is nothing
in general HPCT that prohibits the perception of an error signal, but
standard HPCT doesn't incorporate perceptions of error signals.
Mark:
Reorganization does not happen often, but when it does.... In my experience
of PCT and what I have seen work on models is most effective at changing
lower level control process, it is most often a change the Gain of the error
signal. The integration Factor, in PCT is the basically the "acceptable
level of error" a higher level control system is willing to take, from a
lower level system.
Huh? I think you are mixing up two or three different phenomena here.
There are 12 different possibilities for what reorganization might do
(according to my occasionally reposted taxonomy that Bill approved).
But all of them can be subsumed under the general label of "learning
how." The gain control mechanisms Tom studied are different, involving
adaptation to local situations. I grant that reorganization can also
affect gain, in one of its modes, so there is an overlap. But in one
situation there is direct local control in which the output of one
control system influences the gain of another (which isn't "standard
HPCT, by the way, however true it may be of real organisms). In the
other there long-lasting change.
This is "standard HPCT" it is nothing new, I studied in
my master's program in the early 90's under Bourbon. Error is sensed but
only by a higher level control system, and it's residual cumulative error is
sensed by the reorganizational system.
Error, in what I know of Tom's work, is sensed by a control system
_outside_ the ordinary perceptual control hierarchy. I's not sure
what you mean by "It is residual cumulative error is sensed by the
reorganizational system." I think you must have edited something there
and missed a "that". Assuming that you meant "It is residual cumulative
error _that_ is sensed by the reorganizational system," the statement is
rather sloppy, distinctly irrelevant to the issue at hand, and is in
effect a restatement of a _hypothesis_ that absolute error in the
perceptual control hierarchy is one of the intrinsic variables controlled
by the reorganizing system.
All of which seems unrelated to the question at hand.
Taylor:
So if it is true (as you assert) that hunger is an error signal, and yet we
perceive it, then we have to go beyond standard HPCT into some extended
version in which error signals can be perceptions (and presumably can
be controlled--by some means other than the operation of the control
system in which the error occurs). I think that's weird.
Mark:
That is basic PCT and HPCT. It is what you see when you are doing modeling.
Error goes up the hierarchy,
Never!!! At least not in the standard model.
without that in PCT reorganization could not
occur.
Why do you say that? And why do you put reorganization into the same
hierarchy as perceptual control? That most certainly is NOT stanrd
HPCT. Even if it were, why do you say that "error" is the _only_
intrinsic variable? That sounds extremely unlikely for organisms that
whose biological ancestors managed to maintain their bodily and
chemical integrity long enough to have descendants.
No new extended version is needed just yet, however I would encourage
much more modeling of what we have now -- testing the breath and scope of the
current version PCT should be more than enough work for everyone.
A good sentiment. But it helps to have some understanding of what the
"current version" actually looks like (assuming that one can talk of
a "current version" when there seem to be about three of them for every
five contributors to this list:-)
As I understand it, the "standard version" has one perceptual control
hierarchy, in which perceptual signals are reported upward, transformed
and recombined (necessarily nonlinearly) level by level; the perceptual
signal in any individual control unit is compared with a reference value
for that unit; the difference constitutes an error signal that is fed
to an output function, and the outputs of the output functions at any
level are fed down, recombined and transformed to form reference values
for the control units at the level below. The "standard version" also
has one reorganizing system that manipulates the perceptual control
hierarchy in such a way that the side-effects of perceptual control
serve to sustain an array of "intrinsic variables" at life-sustaining
values (or at least in life-sustaining dynamic orbits).
Your version of the "standard version of HPCT" seems to differ radically
from this. The differences need clarifying, and if there is a way to
make differential predictions of the result of a possible experiment,
such an experiment should be designed and made the subject of your Ph.D.
studies, given that the other model was part of your Master's.
Anyway, could you clarify wherein your "standard HPCT model" differs
from what I described above?
Martin