[From Bill Powers (2003.11.22.1130 MST)]
Marc Abrams (2003.11.22.0822) –
There are aspects to the theory and
model
that are currently speculative and untested. That is, they have not
been
experimentally validated and little data exists to support some of
the
claims.
“Little data” is a relative term. Compared to what we will have
some day, yes it’s pretty scarce. Compared with none, it’s a useful
amount.
_That is not to say that
these assumptions and speculations are
wrong_. It simply means that it remains an open question as to
whether
aspects of the current theory and model provide the best fit.
I don’t
think I would get an argument from either Bill or Rick on this
point.
You’re right. But this observation will remain valid forever. You can
always look at any theory and point out that it is incomplete, that more
data are needed, that revisions and improvements will undoubtedly be
made. However, since this can always be said, it’s not a very useful
observation unless someone gets a swelled head and starts claiming that
he or she has the last word on the subject. Have you noticed people doing
that very much around here?
There’s a reason, to look at the other side of this, for not spending too
much time talking about general shortcomings of PCT or other theories. It
is all too easy to use this lack of perfection as a reason for not
learning and understanding the theoretical structure as it is. A person
can easily decide that since the theory is incomplete, lacks data, and
might even be false in some regards, there is no point in putting out any
serious effort to understand it (Considering your ambitious efforts to go
to school and learn what you’re talking about, you need not fear that
this is directed at you).
I think you will find that most of the people on CSGnet have spent
considerable time (in some cases, large parts of their lives) learning
what other theories propose about human and animal organization. Many
have even become experts in more conventional theories, earning PhD’s in
various fields of psychology, sociology, education, inguistics, and other
fields. To earn a PhD you must do some original research, but beyond
that, you must demonstrate a deep familiarity with what is going on
currently in your field, as well as the history of your field. This means
that for a person with such a background to become aligned with PCT, some
pretty serious comparisons must have been made, and the principles and
logic of PCT must have been quite thoroughly scrutinized. And in many
cases, the conventional ideas must have been evaluated as lacking
something that PCT supplied.
It may not be obvious, but the first people to express skepticism about
scientific speculations are normally the people who propose them. The
first inclination, when a new possibility comes to mind, is to think of
some way of testing it, chasllenging it to see if it will hold up in real
life. The Test for the Controlled Variable isn’t just a formality or a
classroom exercise; it’s a method for challenging any proposal about what
people control and how they control it.
Consider the speculations behind the method of levels, the MOL. It is
speculated that reorganization follows awareness. It is speculated that
awareness can become associated with different control systems and
different levels of control, so that reorganization doesn’t happen
everywhere at once. It is speculated that if the locus of awareness can
be moved to a higher level or order, reorganization at a higher level
will or can if necessary take place. So this suggests a method of therapy
under which the sole purpose is to direct a person’s attention to higher
and higher levels of organization.
The method of levels is a direct test of these speculations. The fact
that people can do it at all indicates that there is something to the
notion of mobility of awareness. The kinds of things people report being
conscious of while using this method support the idea of lower and higher
levels, and the changes that take place support the idea that
reorganization is going on. Of course this is only a beginning, and there
is much left to learn. But that is what we are doing: learning. We are
not saying, “Oh there is so little data, there is so much fuzziness,
there are so many mistakes we could make – we might as well give up
right now.” We are simply trying to see which ideas work and which
don’t. We’re starting from where we are and going on, one step at a time.
And of course at every step we are testing our ideas, not simply
assuming they’re right and using them, or even worse, selecting evidence
to prove they’re right while suppressing or ignoring contrary
evidence.
So the question I have is this; Are
the speculative and untested aspects of
the theory and model open for discussion and alternative ideas on
CSGnet?
Of course. If you have any examples, by all means bring them
up.
Are plausible alternatives
discussable here? Or are you simply trying to
validate the theory and model you proposed 30 years ago? Bill, how
interested are you in discovery?
Hmm. Aren’t you a teeny bit embarrassed to ask that question of me?
Of course plausible alternatives are considered here. Plausibility is
enough to get any idea considered. But it’s not enough to get it
accepted: there is also the matter of testability. An idea that can’t be
tested in some way isn’t worth wasting time on, no matter how plausible
it is. Lots of plausible ideas have turned out to be wrong. Of course, I
reserve the right to ask similar questions about other people’s
ideas.
My interest in PCT is in the
physiological/biological substrate of the
model. I believe all purposeful behavior, from the folding of DNA
to
catching a baseball is based on Bill’s PCT control model. I am
finishing
up my first semester back in school and with any luck I will go on and
study
Biophysics with an emphasis in the regulatory processes involved at
the
cellular and sub-cellular levels in neuroscience. PCT is, and will be,
my
guiding light.
If PCT is any good, you won’t have to believe in it. All you have to do
is remember the questions it raises, and see if what you’re learning
offers equally persuasive answers – and tests. I am full of admiration
for your re-entry into academia, and I expect important things from you
in the future of PCT.
From my current perspective the
first question I ask is; Is it
physiologically reasonable? Does it make physiological sense? Can it
work
physiologically?
You may be surprised to know that the same questions have been in my mind
throughout the development of PCT. My primary concern with any model I
have proposed is whether it is consistent with what we know about
physiology and neurology. I had one advantage over many people when I
started these explorations. I had become a pretty proficient circuit
designer. If you could define something you wanted done, I could come up
with a circuit that would do it. So I was very familiar with
signal-handling devices, transducers, and actuators, and I knew the
difference between a pie-in-the-sky design and a design that would
actually work.
It was not a very large step to go from electronic circuits to neural
circuits. The principles are a little different, but in the 1960s it was
no harder to go from vacuum tubes to transistors to integrated circuits
than it was to go from transistors to neurons. I don’t mean that I came
up with workable circuits for brain functions – that’s a bit beyond the
state of the art today. But the basic principles remain the same: the
ideas that signals carry information from one place to another and that
computing functions can be implemented that will convert sets of signals
into other signals applies just as well to neural systems as electronic
ones. And, I have found in the last 10 or 20 years, to organ systems,
cellular biochemical systems, and the workings of DNA.
Now on to Bill’s
post.
I’m interested in learning and
exploring PCT ‘chemistry’. Primarily the
Input function. How would you advise me to proceed?
At the first two levels I don’t think there is much of a problem: simple
sensors at the first level, and weighted summation at the second level,
probably supply the basic architecture that’s needed. We understand how
color sensations can be built out of long and short wavelength receptor
signals, how tastes and smells can be build from elementary types of
chemosensor signals, and so forth. But at the third level we hit a
barrier. Through some very involved methods, certain simple shape
recognizers have been constructed, and it is possible for artificial
devices to recognize sound configurations, too, but the results are very
limited and require enormous amounts of computation. Basically we’re
lacking some essential idea at this level. If you want fame and glory,
work out how human perceptual systems manage to perceive configurations
of sound, taste, smell, vision, and touch, under all the varying
conditions in a natural environment.
As to functions higher still in the hierarchy, until we crack the third
level we won’t even know for sure what is left to explain. I think there
is a very long way to go for explaining perceptions in neurophysiological
terms. My own preference, since I’m not in neurology, is to try to define
the classes of controllable variables behaviorally and experientially,
because that is what the physiological models will have to explain,
eventually. We can at least define the targets of neurological
research.
I think that distinctions need to
be made between ‘perceptions’ (Cognitive
entitles) and sensory inputs.
I do make distinctions of this sort, except that I divide afferent
signals into more than two classes. Intensity signals come directly from
sensory endings, so they carry what traditionally have been called
“sense data.” They are combined, both in the brainstem and in
the motor cortex, to produce signals standing for sets of intensities –
sensations. For example, the same sensation of “warmth” can be
derived from intensity signals in different parts of the body, showing
that there has been at least one level of abstraction. Intensity signals
are localized; sensation signals are not.
I have also defined, tentatively, nine more classes of sensory or
perceptual information, each derived from the level below it, and
contributing to the level above it if any. At some point you can start
calling the resulting signals “cognitive,” but the boundary is
fuzzy because many perceptions seem both cognitive and sensory – a
signal indicating “size” for example.
When stages of input processing are analyzed to this level of detail, the
simple distinction between sensory and cognitive functions becomes
inadequate.
Some sensory inputs may
become perceptions and
others never will. Both can be, and are, controlled. There are
many
chemical and electric ‘signals’ that ‘communicate’ within our body and
very
few are perceivable and some only at certain times. Most, if not all
are
regulated and ALL affect behavior.
Certainly. I have said this many times.
As to “why” people
have different tastes, I don’t know what kind of answer
you’re looking for.
I was asking how the current configuration of the hierarchy accounts
for
various ‘tastes’. Since reference conditions are set by error signals
from
the level above I was asking for an example of how the current
hierarchy
would ‘explain’ not the ‘differences’ in taste but the acquisition of
them.
Are you asking about reorganization?
Yes, the current hierarchy has the
‘little man’ problem.
No it doesn’t.
I don’t know what you mean by
“psychologizing” or “psychanalytic”. What is the
hierarchy supposed to explain?
It is supposed to explain various levels of behavioral organization that
are commonly observed in human and animal behavior, from what have been
called reflexes to skills to reasoning and so on – the links from
thought to action, if you will. It is also supposed to explain why we
experience the world as we do, in categories like configuration,
motion, events, relationships, and so forth. It is not intended to
“explain” behavior by saying that if you get traumatized at the
age of 6, you will rob convenience stores at the age of 26, or that if
you catch mommy and daddy in the act, you will hate people of the
opposite (or the same) sex for the rest of your life…
What is it supposed to predict? It
doesn’t explain how
‘perceptions’ are formed, since you currently do not distinguish between
a
‘perception’ and a simple afferent signal. It can’t explain how
‘goals’ are
formed because you have the ‘little man’ problem and an infinite
number
levels to be considered. You also have error signals being the sole
source
of reference conditions for lower levels so the reference condition for
any
one level at any one time is dependent solely on the error signal
from
above.
What’s wrong with those ideas (at least the ones you got right)? I
repeat, there is no “little man” problem, since the highest
level is not a whole person with all the levels, but only the level that
perceives and controls system concepts. I add that the number of levels
most certainly is not infinite; There are 11 levels that I’ve been able
to describe in huyman beings. Animals probably have fewer. Yes, reference
signals for lower systems come from higher systems in this model. Got an
idea that works better? Trot it out.
Bill, what if there were no
hierarchy. What if it were a network instead.
You would have no ‘highest’ level, eliminating the ‘little man’ and
you
could account for various states of consciousness and ‘goals’ based on
the
patterns of connections. This would also allow a number of possibilities
for
generating reference conditions including error signals. Just a
thought.
Fine, work it out and present it. I’ve wondered about a network (or
“heterarchy” as Heinz von Foerster calls it), but haven’t been
able to see how to make it work. I picked a hierarchy because there seem
to be goals within goals within goals in human behavior, so we set some
goals as a means of achieving higher or more general goals, in many
layers. The control-system hierarchy simply implements that
organization.
Best,
Bill P.
