[From Bill Powers (991005.0653 MDT)]
Steve Adey (991004.1400 GMT)--
Yes, the only things that the organism is allowed to use to generate
a purpose are what can be packed into the genetic code (which is
powerful but restricted in capacity) and what emerges from its
interaction with the environment (e.g. any social culture). Do we
have any models that show this production of purpose bottom up
by a developing system. I understand that fully formed models exhibit
'purposeful behavior', but then the purpose sort of comes from the
modeller.
Before we start generalizing about purpose, I think we had better say what
we mean by purpose.
I say an organism is acting purposively when it can internally select a
state of the world that doesn't currently exist, and then produce the
actions which bring it into being despite unpredictable disturbances. This
is precisely what the philosophers and psychologists of the early 20th
century insisted was impossible because it would require the future to
affect the present or would call for infallible predictions. The only kind
of organization I know about that can do this is a negative feedback
control system.
What's your definition, and what sort of system do you think can act
purposively?
Best,
Bill P.
···
--------------------------------------------------------------------
[From Kenny Kitzke (991002.1200EDT)]This is a perfect example of your self-imposed limits to
understanding human nature. All your speculations must relate
backward to your PCT concept. It even causes you to invent new
concepts and names for things that others have thought about and
tried to define for centuries. It seems you have to invent your
very own concepts and phenomena and definitions which satisfy your
own engineering concept constraints about how humans work. Your
jargon is unique to you and enhances your difficulty in
communicating with others who have dealt with similar issues in
psychology but used different terms for similar observations.
--------------------------------------------------------------------This need to 'invent' names to say things more precisely seems
inevitable when the objective can be seen as trying to engineer
a model that exhibits some key aspects of the emergent properties
of the mind. It has the unfortunate side-effect of generating a PCT
sub-culture with jargon that seems impenetrable to outsiders. However
the various strands of psychology each seem to have their own
jargon and any stimulus-response terms they may share in common are
the antithesis of PCT. As a delimited theory, PCT (and its HPCT
extension) only attempts to encompass a certain level of behavior
before it inevitable becomes speculation.--------------------------------------------------------------------
[From Dick Robertson,991902.2017CDT]If one feels compelled by the apparent evidence to postulate a
hierarchy of control systems (to explain where the references get
set for the lowest levels--which seem clearly established to
most people who have been at all interested in B:CP--you then come
up against two questions about the hierarchy:1) how does it develop in the first place (since babies clearly can
not control a lot of variables that they obviously increasingly
do, a la Piaget, Spitz, (several others I can't recall now) and
last and best, Plooij.2) does it go on and up forever, or is there a top level and if so
does it have reference values?
---------------------------------------------------------------------Yes,
1) The building of the perceptual control hierarchy from the bottom up
(or at least from where genetically controlled morphological
development leaves the system for perception to take over) is one of
the fundamental problem for all self-organizing brain systems.2) There is obviously not infinite recursion or a 'little-man' inside
the brain, so the top level references in the embryo must derive from
those in the genetic code. It is not yet clear to me whether the
nature of the top level references (not the intrinsic references of
any reorganization system) changes during development of the adult
brain.---------------------------------------------------------------------
Dick:
Let's take (1) first. I can't remember if Bill mentioned the
following in B:CP, but I recall reading about some of the early
attempts to build robots that could learn from experience.
(There was one by ____I'll get it if you insist, but my memory has
gaps nowadays) which, if you kicked its shell when it came to a
corner of the room it would stop moving to that corner.
They found that they had to put a "random signal generator" into the
circuitry of the robot in order for it to do something different
from what it had done previously whenever its control of its
environment was disturbed. If you stop and think about it, this is
purely logical. How can you do something you've never done before,
if you don't know what to do? If you do know how then it's not a
complete innovation. Yes, you might think you are imitating someone
else's behavior. But, you can test that out for yourself.
If you see someone doing a tennis stroke you have never done
before--can you then imitate it on the first trial? Or, do you
have to have "trial and error" until you latch on to it.
You do recognize, don't you, that "trial and error" is a synonym for
"make random movements?" I hope this is enough to start you
thinking about how a biological mess of nerves would have to
develope circuits that weren't there before and then have them
"tested" by whether or not the fixed (genetically) life-support
controls of such things as blood sugar level, o2 level, co2 level
etc. show readings moving toward the "red zone" or the "green zone."
---------------------------------------------------------------------This sort of learning is not random reorganization, based on intrinsic
variables violating intrinsic references. It seems to be something
driven by the nature of the error signal, I perceive that my serve
does not achieve the perception that I had in mind for the trajectory
of the ball. However, there appears to be semi-concious changing of
some undefined parameters (or perhaps references) by trial-and-error,
hopefully not randomly but under some degree of control.---------------------------------------------------------------------
Dick:
Now for (2) Allowing tentatively that the development of a
hierarchy of control systems could account for a lot of things about
behavior that are otherwise lacking a convincing explanation how
does it top off? There are several possibilities:
A. the topmost level in the brain gets references from outside
the skull by some means:
a) spiritual source
b) mysterious signals from space
c) cosmic rays (sort of like b, but maybe not)
I think I clearly understand that your answer would be a.
Well, that's OK as a matter of belief. But if you are
committed ultimately to explanations which could be
demonstrated to everyone's satisfaction (like that if you mix
O2 and H and touch a flame to it you get a bang) then a lot of
people can't yet concede that you have evidence for a, above.
B. the top level of the hierarchy never does fully develope, it
is slowly and randomly migrating through various settings
throughout life, maybe with convergence, maybe not.
C. The topmost level is in continual, slow reorganization as if
trying to build a higher level, and maybe in some people it
does, and in others they die first, and in still others, they
develop a systems concept (or most abstract reference setting
whatever you want to call it ) that sets/delegates/"instructs"
the level of Principles to avoid experiencing randomness as
much as possible.
B. and C. might look quite similar, but I believe I have worked out
that they are somewhat different elsewhere - to my satisfaction,
maybe nobody else's.
---------------------------------------------------------------------This is a very interesting post from Dick that gives another viewpoint
on what Bill was trying to say in his reply to me on Friday (see
below).---------------------------------------------------------------------
From [ Marc Abrams (991002.1938) ]I think a trip back to science 101 here might be helpful I am using
the text_Conceptual Physical Science_ by Hewitt, Suchocki. Hewitt
Published by Addison Wesley. It's always dangerous when "defining"
terms but I think these guys did a good job.A "speculation" is a non- scientific hypothesis about the nature of
something. Non-scientific, meaning that the speculation cannot be
disproven. An example of a speculation of this kind might be
"intelligent life exists on other planets somewhere in the
universe". _At present_ , reasonable or not, this statement is
"speculation". Although it can be proved correct by the
verification of a single instance, there is no way to prove the
hypothesis wrong, if life is not found. Many such statements are
quite reasonable and useful, but they are outside the domain of
science.In everyday usage there is no difference between the use of word
theory or hypothesis. However in science a theory is a large body of
information that encompasses well-tested and verified hypothesis
about certain aspects of the natural world. Scientific theories are
not fixed. They do through stages of redefinition and refinement.
This refinement is a strength of science.Now although the scientific method is powerful, good science isn't
always done this way. Many scientific advances have involved trial
and error, experimentation without guessing ( making hypothesis ),
or just plain old accidental discovery.The last important definition is that of a "fact". In science a
"fact" is generally close agreement among competent observers
studying the same phenomenon. A fact is not something that is
unchanging and absolute. "Facts" change as our knowledge changes.So what's the point? Well from the above definitions there are parts
of PCT that might be considered Scientific theory while there are
aspects of HPCT that _currently_ can be considered no more then
speculation.My point is this. These things will not change until people test
some of this speculations so scientific hypothesis may be made.
---------------------------------------------------------------------I would appreciate it if anyone has a catalogue of which items in B:CP
can be considered 'essential' parts of the theory and which are
hypothesis or speculation. This could be be used to guide thoughts on
modelling, because by the above definitions modelling of speculation
is best avoided by newcomers to the field.---------------------------------------------------------------------
[From Bill Powers (990930.1121 MDT)]
Steve Adey (990929.1740)--
Rick:
Why not implement your view as a working model and test it in
an experiment with real, live control systems (people)?Me:
Has anyone succeeded in modelling these upper layers? Do you
think that someone could do it by modelling a simpler organism
with fewer levels (perhaps only one or two)? I know too little
of HPCT to say that I could sensibly model anything, but I know
that modelling reality is far harder than programming simpler
more abstract models! I am still researching the literature
on actual brain structures that could provide more recent
ideas on physical higher level boundaries.Bill:
No, we haven't modeled (simulated) anything above the relationship
level, and the so-called "event" level is missing below that level.
Even in models of configuration control (only the third level) we
have to postulate that there is a configuration-perceiver; I have no
idea how a person can perceive the "shape" shown in Demo 1 (and
especially its variations), so the control model just has to assume
there is a perceptual signal corresponding to the computer variable
that is varying to alter the shape of the visual image.
---------------------------------------------------------------------I am interested in the fact that you skip level 5 in your model
because it represents an interesting level at which the organism
can perceive events over which it may have no control. When I first
found out about PCT, I found the statement "behavior is the control of
perception" as troubling because my interest in autonomous control
was concerned with simulated worlds where events happen that are
outside the agents control. The attention of the agent has to be drawn
to these events and their significance has to be recognised and then
the agent will act to counter them. Thus my description comes closer
to a stimulus-response than a control of perception. I recognise that
lower levels are indeed doing control of perception but at this event
level it is less obvious. What I wonder is whether a 5 level model
would be feasible (at the moment only as a thought experiment to
derive a specification/understanding), and whether it should
correspond to something a bit like a fly. That perceives some simple
features of the world and then when a rate exceeds some level it
interprets it as an event and responds to avoid trouble (e.g. my
hand). What is less clear is whether the fly could be classed as a
learning system with reorganization or whether everything is hard-
wired. Perhaps there is someone on CSGnet who can think of some
organism that has been investigated with classic stimulus-response
experiments that would count as a five level organism that learns.
On-the-other-hand perhaps S-R needs the sixth 'relational' level to
link stimulus to response, so a five level learning agent could be
verification that the classic S-R description is wrong. The attraction
of only modelling up to level 5, is that events are highest level that
Bill treats to a complete chapter in B:CP. I know that in B:CP, Bill
expresses some doubt about its existence as a separate level, but
maybe modelling will help us understand more about its likelihood.---------------------------------------------------------------------
Bill:
I expect that it will take fifty or a hundred years to sort out the
levels that actually exist, once the scientific community accepts
"levels" as something that needs research, and once it starts
devoting time and resources to this idea. The time and resources
now devoted to this kind of research are close to zero, and a good
portion of them are devoted to returning sniper fire.
---------------------------------------------------------------------As a serious investigator (at least conceptually, if not yet actually
coding), the sooner we get started on the more difficult cognitive
psychology problems, the sooner people will start to take notice.
Defending against Skinnerism seems to be a waste of time since
psychology has move on, even if it is currently stuck with S-R
terminology. I suggest that a key issue is to spend more time on
the higher levels that relate to the cognitive school, and less time
on the lower levels that relate more directly to behaviorism.---------------------------------------------------------------------
Bill:
I think Rick's suggestion that you model your proposal was somewhat
tongue-in-cheek.
---------------------------------------------------------------------But hopefully said in the right spirit. Currently I am waiting on
Rick to send me a zip file of his demos, which he hopes to deliver
by the end of the week. Only when I properly understand what has
been modelled so far will I see what next needs to be done (from
my point of view).---------------------------------------------------------------------
Me:
Bill, I thought that intrinsic references were the highest but you
have suggested some other interesting possibilities. Thanks for the
ideas, are they written down in one of your papers that I should be
reading?Bill:
Yes, in B:CP. Please re-read the section titled "The reorganizing
system" starting on page 182.
---------------------------------------------------------------------I have read it through but not yet extracted a complete understanding.
It is not yet clear to me whether a 5 level system would need to use
reorganization as a mechanism for building itself up in a fly or
whether it would all come pre-built (I presume the latter) which makes
modelling it easier but less useful.---------------------------------------------------------------------
Bill:
Looking at the figure on p. 188, I can see how the reorganizing
system might be mistaken for the highest-level system. But that was
not my intention. The output of the reorganizing system is shown as
affecting _all_ levels of the hierarchy, and its effects are
specifically labeled as "organization-altering effects," not as
changes in reference signals in the hierarchy (which is what a
highest-level control system would produce). Perhaps I should have
located the whole reorganizing system more to the left, and had
those organization-altering effects enter the hierarchy from
the left side.
---------------------------------------------------------------------The diagram was perfectly clear to me, however intrinsic reference
signal(s) violated as a cause of reorganization never seemed to be
mutually exclusive from some of them also being a source of references
for the top level of the HPCT hierarchy. How that is achieved from the
set of possibilities that Bill and Dick suggest we will have to leave
undefined at this stage.---------------------------------------------------------------------
Bill:
What is altered is the way the hierarchy interacts with and controls
its environment. The link through the environment from that
interaction back to intrinsic variables is very indirect. For
instance, one intrinsic variable might be the concentration of sugar
in the blood. But the behavior modified as a result of low blood
sugar could be anything from a pigeon's walking in a figure-eight
pattern to a woman's degree of subservience to her husband.
Such variables have nothing to do with blood sugar concentration
-- except for vagaries of the physical and social environment.
This is why _random_ reorganization is so important: it can lead to
behavioral control of blood sugar even though there is no natural or
rational reason for a link to exist between a specific control
behavior and the level of blood sugar.
---------------------------------------------------------------------That random reorganization is required to cope with such extremes of
weak 'causality' through the world does not invalidate my assumption
that such radical mechanisms will not normally be invoked to control
for minor deviations. However I am open to seeing the results from a
model to decide whether random reorganization should always be
invoked.regards
Steve Adey