Excerpt... Control

[From Rick Marken (2000.02.26.1010)]

Paul Stokes (2000.02.26.0950 GMT) replying to Bill Powers
(2000.02.25.0853 MST)

Thanks for taking the time to give this one some thought. I
think we have all benefited from your remarks which go a long
way, I assure you.

I agree.

(Now if only we could persuade Rick to chip in . . . )

There's really not much I can add to what Bill already said. I
always think it's a good idea to try to implement one's ideas
as a working model so I would encourage you to do this. I also
like data so I would also encourage you to present data that
supports your basic conclusion which (I think) is that people
control for being seen as a group member. (I agree with this
conclusion but I think models and data would give people
a better idea of what this means -- that is, what perceptual
variable(s) people are controlling when they control for "group
membership").

Best

Rick

···

--

Richard S. Marken Phone or Fax: 310 474-0313
Life Learning Associates e-mail: rmarken@earthlink.net
http://home.earthlink.net/~rmarken/

[Paul Stokes (2000.02.27.0228 GMT)]

Bill,

I have replied to your email in the attached MS Word file.

I accept almost in its entirety the contents of your email response to me.

Your email certainly put me straight on a few key points where there was
some residual confusion and misunderstanding and has brought me into closer
alignment with your way of thinking.

However, it also served to highlight those last highly resistant nuggets to
which I have devoted some effort in the enclosed document. I don't think
its essential that we agree on everything at this point. Rick is correct in
pointing out that a lot depends on empirical verification and I can surely
accept that.

Best wishes for now.

Paul

Attachment converted: Betelgeuse:Response to Bill.doc (WDBN/MSWD) (0000D210)

[From Bill Powers (2000.00.28.0734 MST)]

Paul Stokes (2000.02.27.0228 GMT)--

... [your post] served to highlight those last highly resistant nuggets
to which I have devoted some effort in the enclosed document. I don't think
its essential that we agree on everything at this point. Rick is correct in
pointing out that a lot depends on empirical verification and I can surely
accept that.

How pleasant to have a normal discussion.

···

========================================================
Closed-loop, Open System

Paul says:

The best way to conceptualise the apparent openness of the closed control

loop >is to accept that all the nodes around the loop consist of variables
and that >the closure of the loop exists in a set of fixed relationships
between these >variables. In other words closure does not exist between the
content of the >variables (which can be anything at any time in any place)
but between their >functions.

I don't follow this. "Variables" are simply quantities that vary. In any
particular control loop, there is some controlled variable, which we
observe as something specific in the environment (or the body as opposed to
brain) of the control system, something that is held constant or in a
constant pattern against disturbances. I don't see the different variables
we control as all being controlled by the _same_ control system, but by
physical distinct control systems.

As "open systems" and "closed systems" were conceived in physics, they were
open or closed strictly with respect to energy exchanges with their
environments. A closed system was one which neither received nor emitted
any energy, respectively, from or to its environment. Obviously organisms
do not belong in that class: they continually receive energy in the form of
food and oxygen; they continually lose it by doing work on the environment
and emitting heat. The food they eat is partly converted to their biomass,
and partly to stored energy which is expended when the muscles exert
forces. So the system is in fact completely open.

When information theory came along and people started using the entropy
metaphor (which seems to have some relationship, at first, to energy flow),
the "openness" of a system came to include information flow. But even if we
accept this dubious metaphor, it is clear that organisms are open systems
in this regard, too. Potential information comes into them from the
environment through millions upon millions of sensory receptors, and for at
least some organisms information is emitted via vocalizations and gestures
(as well as simply by actions). Of course transmitted information is
defined by the receiver, but the channels are clearly there and are clearly
used.

What is closed in PCT is the primary loop of causation. In this loop,
perceptual inputs are converted into actions, while actions are
simultaneously being converted into perceptual inputs. This is not to say
that the loop is closed to external influences. There can be disturbances
which have just as much effect on the perceptual input (or would have, but
for feedback effects) as the system's own action has; and the internal
reference signal, which comes from higher systems, influences the error
signal just as much as the perceptual signal does. External disturbances,
however, are prevented from having their "normal" effects on the input
because the action of the the system opposes and largely cancels those
effects. Nevertheless, we can see _some_ effect of the disturbances,
because we can predict that the actions of the system will very nearly
balance out the effect of the disturbances. So if I want to see you pushing
against my hand, I can use my hand to push on you.

The functions of reference level, perception, comparator and behaviour are
determinably linked in a closed loop and derive their meaning one from

another >in a highly specific manner. The loop is not for varying or
opening but the >content can and does - all the time.

"Does vary", I take it, and I agree with that. What you call "content" is
what I would call "magnitude." A given perceptual signal always refers to
just one perceived entity, with its magnitude indicating how much of that
entity is present. A different entity would be represented by a physically
distinct perceptual signal in a different control system. Remember that the
PCT model is a highly parallel system with many control loops in action at
once. VERY many.

Paul, out of sequence:

Clearly the events that are being described here are taking place at

different >levels of logical typing and are hence incommensurable. What is
closed at one >level can be open at another.

The problem here is in using "closed" with different meanings in the same
argument. A closed loop is simply a chain of causes that is connected into
a loop (unbroken, and therefore closed in the same sense that an electrical
circuit becomes closed when you connect the last two wires to each other).
This sense of "closed" has nothing to do with isolation or impermeability.

This same loop, because it receives perceptual information from the
environment and because its actions expend energy on the environment
(energy that was originally received from the environment) is
simultaneously a _thermodynamically open_ system. Half of the control loop
is inside the behaving system; the other half is in its environment. While
only negligible energy input is involved in perception, there is still a
continuous exchange of significant amounts of energy between the system and
its environment through its actions.

This system is also subject to perturbations from outside the loop: the
reference signal and disturbances. Those two entities are part of the model
simply to give it maximum generality. So even a loop that is causally
closed is open to perturbations from outside it. The words "closed" and
"open" are simply too vague to cover all these meanings.

Paul:

The variables themselves have changeable content. Furthermore, the

changeable >content can come from anywhere. Just as the environment can
disturb our >perceptions so can our reference objects and their reference
levels as well as >the variety of our behavioural response be derived
exogenously. Hence: closed >loop - open system.

I'm not sure what you mean by "changeable content" of a variable. In PCT, a
variable has a fixed identity (after learning is complete). The variable
"green" comes to be represented as a perceptual signal which from then on
never means anything but "green." The magnitude of the signal indicates
_how much_ green is present. A control loop involving this perception would
control for a specific amount of green, the amount specified by the
reference signal it receives. These signals will never stand for "red" or
"smooth." A given control loop consists of neurons dedicated to the control
of one particular variable, and never controls any other variable. If a
different controlled variable is controlled, at least a different
perceptual function, comparator, and output function are involved. The kind
of model used in PCT thus requires a great many control systems, but at the
same time these control systems are extremely simple, so each one can be
implemented with a small number of neurons.

I'm also not sure what you mean by a "reference object." If you mean "an
external object that is referred to by a perceptual signal," the meaning of
the word "refer" in that context is completely different from its meaning
in the term "reference signal."

A reference signal is a neural signal of the same physical form as a
perceptual signal; it is received from higher systems by a control system's
comparator. The perceptual signal is compared with the reference signal by
subtraction (which, neurally, entails quantitative inhibition). The
magnitude of the perceptual signal can vary from zero to some maximum
amount. The magnitude of the reference signal also can vary between zero
and a maximum amount. When the reference signal's magnitude is held steady
at some level within this range, it becomes a fixed target to which the
control action matches the magnitude of the perceptual signal. If the
perceptual signal corresponds reliably to the state of some external
variable (reasonably true most of the time), we will observe that external
variable being brought to some particular state and being maintained there.
Of course if the magnitude of the reference signal changes, the control
action will cause the perceptual signal to change in the same way, its
magnitude remaining close to the magnitude of the reference signal.

That may be exactly what you meant, of course, but it always pays to be sure.

Paul:

I think that you tend to argue like a rational choice theorist and that

this >is the nub of our problem. (Like an economist you take a minimalist
view of >human needs.)

I may use simple examples to make my point, but I don't rule out _a priori_
any proposal concerning a human need. People vary so much in what they
"need" at higher levels of organization, however, that it's simpler to
speak of things like food, which they all need. But if you want to say that
all human beings need fun, power, belonging, and such things (as Glasser
proposes), I'll hold still while you try to prove it to me. However, if
you're claiming to define a need that is common to all human beings, I
reserve the right to demand that all human beings, without significant
exception, be shown to have it. Not four out five, or 51 per cent, but ALL.
Otherwise you're probably talking about an acquired need, not a universal one.

I believe that rational choice theorists claim that we determine our
actions by rationally considering our needs and the means available for
meeting them. I would agree that some people do that with respect to some
needs (a stock market analyst, for example), but I do not by any means
agree that this is a basic or a universal way of handling the meeting of
needs. The really basic physiological needs, and probably many others, I
think are controlled through reorganization, which (as technically defined
in PCT) is anything but a rational process.

Paul:

The examples given are all typically to do with our own individual welfare

and >well being. This is why I use the phrase 'radical individualism' to

characterise this aspect of PCT and it has nothing to do with your point

about >perception (which I accept completely and will amend accordingly).

I'm glad you said "the examples given." I could clearly give other
examples, and probably should have done so to avoid giving a misleading
impression. It is perfectly possible for a person to have a reference level
that says "The welfare of other people is far more important than my own."
The controlled perception is then the welfare of others relative to my own,
and the control actions will see to it that the perception maintains the
desired proportionality between others' and my own welfare. This in no way
changes the fact that the "welfare of others" is a perception in the brain
of an individual, and that the goal also resides in the brain of the
individual.

My "radical individualism" should not be confused with self-centeredness or
selfishness or egocentrism (as I think you may be doing). I am merely
saying that all concepts of anything are perceptions in the mind of some
individual, and that all goals, whether selfish, altruistic, or coldly
objective, exist in the mind of the individual. There is no other place for
them to exist, is there?

Paul:

The implication is that there are our human needs - which are basic and

rather >limited. A further implication is that our sociality as a species
is based on >the fact that we may have to use other people a lot of the
time in order to >fulfil these needs. A final implication might well be:
better not piss them >off - better to act in ways likely to improve our
chances of need fulfilment.

We certainly have to use other people to satisfy a personal need to make
them happy. Of course by "use" you mean "take advantage of" or some other
pejorative sense, but that's not at all what I propose. I'm merely pointing
out that whatever goals you have regarding other people, they are _your_
goals, and the perceptions you're trying to control are _your_ perceptions,
the only ones you can know about for sure.

Paul:

I am not denying that such calculations are part and parcel of social life

and >social interactions. They are. But what is altogether missing from
your >argument and that of other PCTers is an acknowledgement that there
are such >things as intrinsically social needs and goals that have nothing
whatsoever to >do with our immediate physiological, and biological welfare
and hedonistic >desires. (Although, as Wilkinson and James point out,
failure to meet these >intrinsic social needs can have lethal physiological
consequences).

If there are intrinsically social needs and goals -- such as a desire to be
with others of one's species -- they can only exist inside the individual.
I certainly do not deny the possibility that such needs may exist in some
people, or most people, or in some cases, all people. How they got there, I
can't say, and neither can you. But nothing in PCT denies their possibility.

Neither do I deny the existence of explicitly social perceptions. Anything
having to do with interactions among people is a social perception that
specifically refers to other people. Robinson Crusoe couldn't even aspire
to being a leader until Friday showed up. I think the legitimate subject of
study for sociology is the way people control social perceptions, and what
happens when, as autonomous control systems, they necessarily interact with
each other. What I will not admit is that the locus of control exists
anywhere but in the individual. I can interact only with _my_ perception of
society; I can have desires only for _my_ perception of what another person
is. If my perceptions and goals are incompatible with reality, hard
experience will eventually set me straight, provided I survive.

My only in-principle objections to social theories have arisen when they
neglect the individual and therefore are not grounded in first principles.
Social laws can't exist in the air between people: they must emerge from
the properties of interacting individuals. There is simply no other way,
physically, for them to work.

It may be that some social goals are inherited, but as you know, it isn't
simple to separate what is inherited from what is simply passed down
through communications (and actions!) between generations. I'm happy to
leave that question to the future. I can't answer it.

When you speak of exhaustively citing evidence for social needs, I think
you are called upon to consider that while there may be overwhelming
evidence that _some_ people out of any group will have some specific social
need, the evidence is distinctly underwhelming that _every_ person in a
group has it. My view is that any time a generalization proves to be true
of only part of a group, the chances are pretty high that we're talking
about something that is either transmitted culturally, or simply a logical
consequence of circumstances which some, but not all, have figured out.

Paul:

Instead of looking to the connections between social context and

consequent >behaviour we must look to how people's perceptions of
themselves both as >individuals and of themselves as societies changed over
the period from the >12th Century in Europe onwards. Once we have this in
place we can say that the >behaviour of the mass of the population, first
of one group then another, also >changed in the direction of more refined
behaviour in order to stabilise these >new perceptions.

What a fabulous thesis! This sends shivers up my spine: it sounds like
something that could actually be worked out to some real factual
conclusions. It is totally consistent with PCT, yet it will take us into
areas of knowledge that PCT has not yet even touched -- not to mention
sociology.

We're getting somewhere, I think.

Best,

Bill P.

[From Paul Stokes (2000.02.27.0228 GMT)]

[Bill Powers (2000.00.28.0734 MST)]

How pleasant to have a normal discussion.

Indeed and I hope to continue it.

How embarassing it is to witness your razor-sharp scalpel of a mind dissect
my rather loosely phrased arguments with such clinical precision!

I think there is some terminological confusion between us - largely my
fault, I think - but there is almost nothing of substance in this that I
would disagree with. I do not want to parry further over details or over
matters of legitimate debate which are peripheral to the main thrust of the
argument. I am delighted that you have answered my points so clearly and -
this surprised and cheered me immensely - so positively.

Your wrote:

What a fabulous thesis! This sends shivers up my spine: it sounds like
something that could actually be worked out to some real factual
conclusions. It is totally consistent with PCT, yet it will take us into
areas of knowledge that PCT has not yet even touched -- not to mention
sociology.

I think PCT is destined to make a big splash in this area for all the
reasons cited.

We're getting somewhere, I think.

I agree.

Best wishes.

Paul

[From Rick Marken (2000.02.28.1320)]

Today is my birthday. Happy birthday to me.

Paul Stokes (2000.02.27.0228 GMT)--

Instead of looking to the connections between social context and
consequent behaviour we must look to how people's perceptions of
themselves both as individuals and of themselves as societies
changed over the period from the 12th Century in Europe onwards.
Once we have this in place we can say that the behaviour of the
mass of the population, first of one group then another, also
changed in the direction of more refined behaviour in order to
stabilise these new perceptions.

Bill Powers (2000.00.28.0734 MST) --

What a fabulous thesis! This sends shivers up my spine: it sounds
like something that could actually be worked out to some real
factual conclusions. It is totally consistent with PCT, yet it
will take us into areas of knowledge that PCT has not yet even
touched -- not to mention sociology.

I agree. I'd really like to see proposals for testing this
notion. I think Paul is proposing that, at the aggregate level,
people change their controlling (the perceptions they control;
the reference states of these perceptions) so that the aggregate
is always keeping some "virtual" controlled perceptions in virtual
reference states (that are changing over time).

If I have this right, then I think this is a fascinating proposal
(similar to one proposed by Kent McClelland, I believe). I think
Paul's historical perspective is particularly interesting because
there are "natural experiments" that have been conducted over the
centuries (religious wars, Nazism, economic oppression, etc) that
can be seen as possible disturbances to the virtual perceptions
that are, hypothetically, under control by the aggregate.

Best

Rick

···

--
Richard S. Marken Phone or Fax: 310 474-0313
Life Learning Associates mailto: rmarken@earthlink.net
http://home.earthlink.net/~rmarken

[Norman Hovda (2000.00.28.1450 MST)]

I agree. I'd really like to see proposals for testing this
notion. I think Paul is proposing that, at the aggregate level,
people change their controlling (the perceptions they control;
the reference states of these perceptions) so that the aggregate
is always keeping some "virtual" controlled perceptions in virtual
reference states (that are changing over time).

If I have this right, then I think this is a fascinating proposal
(similar to one proposed by Kent McClelland, I believe). I think
Paul's historical perspective is particularly interesting because
there are "natural experiments" that have been conducted over the
centuries (religious wars, Nazism, economic oppression, etc) that
can be seen as possible disturbances to the virtual perceptions
that are, hypothetically, under control by the aggregate.

Best

Rick

This all reads (Kent and Paul) to me like what I'm attempting to do with
regard to applying PCT to markets (specifically DJIA futures.) I don't
have anything of substance to report (I'm still reading "Mind Readings"
and just ordered "Living Systems 1 & 2). From a most basic and
rudimentary spreadsheet fooling around POV I am encouraged.

The best description I now have is I'm trying to quantify the cumulative
amount of error present in the market moving through time.

Best regards,
nth