No Action Required?

[Fred_Nickols3]

As I understand PCT, behavior serves to control perception; more
specifically, behavior controls perceptions of what are termed "controlled
variables" (Cv). I'm no master of scientific or PCT notation so bear with
me here: Let's say the reference signal for a Cv (i.e., its desired or
preferred state) is represented by Cv' (read that as Cv prime). And, let's
say that the actual state of that Cv is represented by (what else?) Cv. If
Cv equals or is within whatever tolerances might exist regarding Cv', no
action is required. That is the "no error" condition and, absent error,
there is no action. Correct?

Assuming I'm correct, then action - of any kind at any time - has to be the
result of a comparator producing an error signal; in other words, Cv does
not equal Cv'. It further seems to me that any gap or discrepancy between
Cv and Cv' can come about in two ways:

First, Cv was equal to Cv' but a "disturbance" created a gap, hence an error
signal and what I would ordinarily call "corrective action." In this case,
the action is intended to maintain the Cv'.

Second, Cv is not and never was equal to Cv' and so a continuing gap results
in continuing action. In this case, the action is intended to achieve the
Cv' for the first time. Maintenance is not yet an option.

So, in thinking about the behaviors I observe in various places at various
times, it strikes me that I have no way of determining if someone's behavior
is serving to "achieve" a Cv or "maintain" it. The difference between those
two strikes me as fundamental if not profound.

Comments anyone?

Regards,

Fred Nickols, CPT
Senior Consultant
Distance Consulting
"Assistance at A Distance"
nickols@att.net
www.nickols.us

[Fred_Nickols3]

[From Fred Nickols (2004.12.10.1724 EST)] --

Two things: First, my apologies for not doing the date/time thing in my
previous posting.

Second, a P.S. It occurs to me there is a third possible source of error
and thus action; namely, a brand-new, never before set Cv'.

Regards,

Fred Nickols
nickols@att.net

···

-----Original Message-----
From: Fred Nickols [mailto:nickols@worldnet.att.net]
Sent: Friday, December 10, 2004 5:06 PM
To: 'Control Systems Group Network (CSGnet)'
Subject: No Action Required?

As I understand PCT, behavior serves to control perception; more
specifically, behavior controls perceptions of what are termed "controlled
variables" (Cv). I'm no master of scientific or PCT notation so bear with
me here: Let's say the reference signal for a Cv (i.e., its desired or
preferred state) is represented by Cv' (read that as Cv prime). And,
let's say that the actual state of that Cv is represented by (what else?)
Cv. If Cv equals or is within whatever tolerances might exist regarding
Cv', no action is required. That is the "no error" condition and, absent
error, there is no action. Correct?

Assuming I'm correct, then action - of any kind at any time - has to be
the result of a comparator producing an error signal; in other words, Cv
does not equal Cv'. It further seems to me that any gap or discrepancy
between Cv and Cv' can come about in two ways:

First, Cv was equal to Cv' but a "disturbance" created a gap, hence an
error signal and what I would ordinarily call "corrective action." In
this case, the action is intended to maintain the Cv'.

Second, Cv is not and never was equal to Cv' and so a continuing gap
results in continuing action. In this case, the action is intended to
achieve the Cv' for the first time. Maintenance is not yet an option.

So, in thinking about the behaviors I observe in various places at various
times, it strikes me that I have no way of determining if someone's
behavior is serving to "achieve" a Cv or "maintain" it. The difference
between those two strikes me as fundamental if not profound.

Comments anyone?

Regards,

Fred Nickols, CPT
Senior Consultant
Distance Consulting
"Assistance at A Distance"
nickols@att.net
www.nickols.us

[Martin_Taylor2]

[Martin Taylor 2004.12.10.17.30]

[Fred Nickols, apparently 10 Dec 2004 17:05:57]

As I understand PCT, behavior serves to control perception; more
specifically, behavior controls perceptions of what are termed "controlled
variables" (Cv). I'm no master of scientific or PCT notation so bear with
me here: Let's say the reference signal for a Cv (i.e., its desired or
preferred state) is represented by Cv' (read that as Cv prime). And, let's
say that the actual state of that Cv is represented by (what else?) Cv. If
Cv equals or is within whatever tolerances might exist regarding Cv', no
action is required. That is the "no error" condition and, absent error,
there is no action. Correct?

No, there's a significant mistake here, that voids the rest of your argument.

The "no error" condition in any individual control system, elementary
or complext, indicates that the current action is effectively
opposing the existing disturbance. So "no error" doesn't mean "no
action". It means "keep doing what you are doing".

In the canonical Elementary Control Unit (ECU), the output function
is a simple integrator. If an integrator keeps being fed zero at its
input, its output will retain whatever value it had when the "zero"
state began.

Hope this helps. I have very little time these days.

Martin

[Richard_Marken2]

[From Rick Marken (2004.12.10.1450)]

Martin Taylor (2004.12.10.17.30) --

Fred Nickols ( apparently 10 Dec 2004 17:05:57)--

If Cv equals or is within whatever tolerances might exist regarding Cv', no
action is required. That is the "no error" condition and, absent error,
there is no action. Correct?

No, there's a significant mistake here, that voids the rest of your argument.

The "no error" condition in any individual control system, elementary
or complext, indicates that the current action is effectively
opposing the existing disturbance. So "no error" doesn't mean "no
action". It means "keep doing what you are doing".

In the canonical Elementary Control Unit (ECU), the output function
is a simple integrator. If an integrator keeps being fed zero at its
input, its output will retain whatever value it had when the "zero"
state began.

Excellent, Martin. Clear, concise and correct.

Best

Rick

···

--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400

--------------------

This email message is for the sole use of the intended recipient(s) and
may contain privileged information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.

[Fred_Nickols3]

[From Fred Nickols (2004.12.10.1825 EST)] --

[Martin Taylor 2004.12.10.17.30]

>[Fred Nickols, apparently 10 Dec 2004 17:05:57]

>As I understand PCT, behavior serves to control perception; more
>specifically, behavior controls perceptions of what are termed
"controlled
>variables" (Cv). I'm no master of scientific or PCT notation so bear
with
>me here: Let's say the reference signal for a Cv (i.e., its desired or
>preferred state) is represented by Cv' (read that as Cv prime). And,
let's
>say that the actual state of that Cv is represented by (what else?) Cv.
If
>Cv equals or is within whatever tolerances might exist regarding Cv', no
>action is required. That is the "no error" condition and, absent error,
>there is no action. Correct?

No, there's a significant mistake here, that voids the rest of your
argument.

The "no error" condition in any individual control system, elementary
or complext, indicates that the current action is effectively
opposing the existing disturbance. So "no error" doesn't mean "no
action". It means "keep doing what you are doing".

In the canonical Elementary Control Unit (ECU), the output function
is a simple integrator. If an integrator keeps being fed zero at its
input, its output will retain whatever value it had when the "zero"
state began.

Hope this helps. I have very little time these days.

Martin

Thanks for taking the time to reply, Martin. Much appreciated. However, to
be honest, it helps and it doesn't help. A couple of additional points or
questions or whatever.

First, what you say above seems to me to assume ever-present disturbances.
Is that indeed the case? Is every Cv we try to control always subject to
disturbance? Further, "keep doing what you're doing" implies ongoing
action. What about start up or initiation?

Second, I'm an old fire control technician; cut my teeth on the fire control
problem as solved by electro-mechanical analog computers. My recollection
of the way a simple integrator works is that, first of all, it has more than
one input. It was also a dynamic device. If its inputs go to zero, its
outputs do too.

Third, thanks for the bonus: I was wondering what ECU stood for and I was
pretty sure it wasn't European Common Union.

Regards,

Fred Nickols

[Bruce_Abbott1]

[From Bruce Abbott (2004.12.10.1840 EST)]

Fred Nickols (2004.12.10.1724 EST) --

Two things: First, my apologies for not doing the date/time thing in my
previous posting.

Second, a P.S. It occurs to me there is a third possible source of error
and thus action; namely, a brand-new, never before set Cv'.

It seems to me that a "never before set Cv' " would imply that the control
system itself did not exist before. Existing systems always have a "set"
reference, even if it is only an implied one. (Otherwise, how can it be a
control system?) However, merely changing the value of a reference can
induce an error, if the change produces a difference between Cv and Cv'
that is large enough to be outside the "usual tolerances." This would
constitute the third possible source of error.

Bruce A.

[Richard_Marken2]

[From Rick Marken (2004.12.10.1600)]

Fred Nickols (2004.12.10.1825 EST)] --

Martin Taylor (2004.12.10.17.30)--

In the canonical Elementary Control Unit (ECU), the output function
is a simple integrator...

Third, thanks for the bonus: I was wondering what ECU stood for and I was
pretty sure it wasn't European Common Union.

I'm still wondering what an ECU is. What is an ECU, Martin? Is an ECU
defined by the nature of the output function? Is a proportional controller,
where o = ke, not and ECU?

Regards

Rick

···

--
Richard S. Marken
MindReadings.com
Home: 310 474 0313
Cell: 310 729 1400

--------------------

This email message is for the sole use of the intended recipient(s) and
may contain privileged information. Any unauthorized review, use,
disclosure or distribution is prohibited. If you are not the intended
recipient, please contact the sender by reply email and destroy all copies
of the original message.

[Fred_Nickols3]

[From Fred Nickols (2004.12.11.0534)] --

In case you're wondering why I'm sending out emails at 5:30 in the morning,
we have two new puppies. On to business...

It seems to me that children offer instances of control systems that don't
have reference signals for certain Cvs. Is it not the case that a child
learning to walk is developing (through reorganization or learning or just
plain practice) the reference signals necessary to control the perception of
standing upright and moving about? And, come to think of it, doesn't that
suggest that there is interaction between perception and the formation of
reference signals? (Or do I mean the level or value of a reference signal?)

I take what you're saying, Bruce, as meaning that we already have reference
signals for everything. I can see how the mechanisms (i.e., the control
system) needs to be in place but I don't see how all the reference signals
have to be there too. If that were the case, we wouldn't need to learn
anything; all we'd have to do is "turn on" the relevant reference signal.
I'm obviously missing something here. Maybe I'm confusing reference signal
with the value of the reference signal.

Oh well, I'll go back and see if I can formulate a better question. In the
meantime, What you say at the very end is what I thought I was saying so
thanks for confirming that - I think.

Fred Nickols

···

From Bruce Abbott (2004.12.10.1840 EST)]

>Fred Nickols (2004.12.10.1724 EST) --
>
>Two things: First, my apologies for not doing the date/time thing in my
>previous posting.
>
>Second, a P.S. It occurs to me there is a third possible source of error
>and thus action; namely, a brand-new, never before set Cv'.

It seems to me that a "never before set Cv' " would imply that the control
system itself did not exist before. Existing systems always have a "set"
reference, even if it is only an implied one. (Otherwise, how can it be a
control system?) However, merely changing the value of a reference can
induce an error, if the change produces a difference between Cv and Cv'
that is large enough to be outside the "usual tolerances." This would
constitute the third possible source of error.

[Bill_Powers1]

[From Bill Powers (2004.12.11.0728 MST)]

Fred Nickols (2004.12.10.1825 EST) –

[To Martin] First, what you say
above seems to me to assume ever-present disturbances. Is that indeed the
case? Is every Cv we try to control always subject to
disturbance?

This depends somewhat on the level of variable you’re talking about. The
analog variables (from relationships on down) are highly susceptible to
disturbances. When you reach for your toothpaste, how repeatable is the
position of the toothpaste tube, and how repeatable is the position of
your body as you prepare to reach? All it would take would be a one-inch
difference in either position for you to miss grasping the toothpaste
tube altogether – if you did this by repeating exactly the same actions
each time.
You are always adjusting your actions to counteract the effects of small
disturbances – both small and large, actually. This happens quickly and
without the need for thought, so you think you’re “doing the same
thing” every time. But pay close attention to anything you think you
do the same way every time. You’ll see those little adjustments, and big
adjustments, if you look. If I ask you to touch your nose, the first
moves you make depend on where your hand is when I ask, and which way
your head is turned. If I ask you to “do it again,” you’ll see
that all the moves are now different, because now your hand and head are
in different positions and orientations. Even though it feels as if
you’re doing the same thing again, touching your nose, you’re not. You’re
only calling it the same thing. The only reason you can keep
producing the same consequence is that you’re a control system and base
you actions on the difference between what you want and what you perceive
RIGHT NOW.

Disturbances having to do with categories, sequences, logic, principles,
and system concepts are not as common because it takes a larger
disturbance to cause a perceptible perturbation. But don’t forget that
even higher-order variables are continuous functions of the current state
of the analog world, so are still susceptiable to perturbation.
Disturbances can result from errors in action as well as from external
forces – when you speak or write, for example, you make little mistakes
– I corrected five in getting this far even in this sentence, though you
don’t see the result. We’re always editing, correcting, fixing, changing,
repairing, rearranging, because of disturbances from outside and from
inside.

But all our lives we’ve been juggling our actions to do make those
corrections, and are hardly aware that we’re doing them. Disturbances?
What disturbances?

Best,

Bill P.

···

Further, “keep doing
what you’re doing” implies ongoing

action. What about start up or initiation?

Second, I’m an old fire control technician; cut my teeth on the fire
control

problem as solved by electro-mechanical analog computers. My
recollection

of the way a simple integrator works is that, first of all, it has more
than

one input. It was also a dynamic device. If its inputs go to
zero, its

outputs do too.

Third, thanks for the bonus: I was wondering what ECU stood for and I
was

pretty sure it wasn’t European Common Union.

Regards,

Fred Nickols

[Bill_Powers1]

[From Bill Powers(2004.12.11.0753 MST)]

Fred Nickols (2004.12.11.0534) --

Is it not the case that a child
learning to walk is developing (through reorganization or learning or just
plain practice) the reference signals necessary to control the perception of
standing upright and moving about?

A reference signal can't help a child learn to stand upright and move
about, can it? I think what the child learns are the lower-order
perceptions and the actions needed to maintain an upright posture in
relation to the environment, a perception that is acquired when you help
the child stand upright for a moment. The child (sooner or later) wants to
repeat this experience; the perception is remembered and becomes a
reference signal against which the current state of pereived uprightness is
compared. That comparison produces an error signal, and the child then
starts learning what it has to do with its lower control systems to cause
that perception to appear again, erasing the error signal.

There are other routes to acquiring a reference signal fort being upright,
but that's one scenario that seems reasonable.

Best,

Bill P.

[Fred_Nickols3]

[From Fred Nickols (2004.12.11.1145 EST)] –

This is proving very helpful. I
obviously don’t understand very much.

I thought “disturbances” were factors
that interfered with my control of a variable.

In the example below, I understand that
the position of the toothpaste might vary as might my position relative to it
and so the precise behavior used on one occasion would not succeed on another.
If someone had a string attached to the toothpaste tube and jerked it out of my
reach every time I reached for it, that’s something I would consider a “disturbance.”
But, the toothpaste tube lying there passive and inert and me wishing to grasp
it, reaching for it and grasping it? Where is the disturbance in
that? I’m getting confused between “error” a difference
between the perceived state of a controlled variable and a reference signal for
that variable and a “disturbance.” Is your point essentially
that the last time I used the toothpaste I (or someone else since) set it in
some other place and that difference constitutes the disturbance?
Obviously, I don’t get it.

From Bill Powers (2004.12.11.0728 MST)]

Fred Nickols (2004.12.10.1825 EST) –

[To Martin] First, what you say above seems to me to assume
ever-present disturbances. Is that indeed the case? Is every Cv we try to
control always subject to disturbance?

This depends somewhat on the level of variable you’re talking about. The analog
variables (from relationships on down) are highly susceptible to disturbances.
When you reach for your toothpaste, how repeatable is the position of the
toothpaste tube, and how repeatable is the position of your body as you prepare
to reach? All it would take would be a one-inch difference in either position
for you to miss grasping the toothpaste tube altogether – if you did this by
repeating exactly the same actions each time.
You are always adjusting your actions to counteract the effects of small
disturbances – both small and large, actually. This happens quickly and
without the need for thought, so you think you’re “doing the same
thing” every time. But pay close attention to anything you think you do
the same way every time. You’ll see those little adjustments, and big
adjustments, if you look. If I ask you to touch your nose, the first moves you
make depend on where your hand is when I ask, and which way your head is
turned. If I ask you to “do it again,” you’ll see that all the moves
are now different, because now your hand and head are in different positions
and orientations. Even though it feels as if you’re doing the same thing again,
touching your nose, you’re not. You’re only calling
it the same thing. The only reason you can keep producing the same consequence
is that you’re a control system and base you actions on the difference between
what you want and what you perceive RIGHT NOW.

Disturbances having to do with categories, sequences, logic, principles, and
system concepts are not as common because it takes a larger disturbance to
cause a perceptible perturbation. But don’t forget that even higher-order
variables are continuous functions of the current state of the analog world, so
are still susceptiable to perturbation. Disturbances can result from errors in
action as well as from external forces – when you speak or write, for example,
you make little mistakes – I corrected five in getting this far even in this
sentence, though you don’t see the result. [To Bill: And even then you missed at least one –
assuming you meant to correct it. J ] We’re always
editing, correcting, fixing, changing, repairing, rearranging, because of
disturbances from outside and from inside.

But all our lives we’ve been juggling our actions to do make those corrections,
and are hardly aware that we’re doing them. Disturbances? What disturbances?

[Martin_Taylor2]

Re: No Action Required?
[Martin Taylor 2004.12.11.13.14]

[From
Fred Nickols (2004.12.11.1145 EST)] -

This
is proving very helpful. I obviously don’t understand very
much.

I
thought “disturbances” were factors that interfered with my
control of a variable.

This isn’t actually wrong, but
it’s misleading, and it is probably not a good idea to think of
disturbances like that.

Why is it misleading? Because
"interfering with my control of a variable’ seems to imply that
something has affected your ability to influence the variable. In
other words, thinking of the canonical diagram of an ECU,

it would be altering the bottom feedback path. A disturbance
doesn’t do that.

Why is it not actually wrong? What a disturbance does, if you do
nothing to oppose it, is alter the value of the input to the
perceptual function, and therefore the value of the perceptual signal.
So its immediate effect is to create error. The output function inputs
the error and outputs an effect on the disturbed variable, reducing
the error. So, if you think “interfering with my control” is
synonymous with “increasing the error in the controlled
variable”, then a disturbance does that – without interfering
with the ECU’s ability to control.

A disturbance isn’t the only thing that can create momentary
error. The same happens if the reference value changes. Disturbance
and Reference are the only two signals that enter the loop from
outside. So, even if there is nothing in the environment affecting the
controlled variable except the output of the control system, transient
errors may still occur. But you wouldn’t say that a reference signal
interferes with your control of a variable, would you? So I think it’s
not a good idea to think of a disturbance interfering with your
control, either.

So, if the value of the disturbance is zero, and the error is
zero, the output must be zero, and your “No action required”
suggestion would be correct. But if there is a steady external
influence on the environmental variable, then to maintain zero error
in the perceptual signal, the output must be equal, opposite, and
continuing.

[From Fred Nickols (2004.12.10.1825
EST)]
Further, “keep doing what you’re
doing” implies ongoing
action. What about start up or
initiation?

Does the foregoing answer that? Start-up implies a change of
reference signal, doesn’t it?

My recollection
of the way a simple integrator works is
that, first of all, it has more than
one input. It was also a dynamic
device. If its inputs go to zero, its
outputs do too.

I don’t know what you called an integrator in your
electro-mechanical analogue computer days, but what I meant was a
function that implemented the mathematical operation of integration. I
can’t easily use math notation here, but in words:

x(t1) = x(t0) + (integral from time t0 to t1 of y(t) dy)

Here, y(t) is the input waveform to the integrator, x(t) its
output.

if y(t) has been zero between t0 and t1, x(t1) = x(t0), not
zero.

Martin

[Fred_Nickols3]

Re: No Action Required?

[From Fred Nickols (2004.12.11.1435 EST)]
–

Many thanks for the reply, Martin, and
many more for the animated drawing. I sat and watched it for several
minutes. (How did you do that?)

Before I burden you with any more
questions, let’s see if I have the symbols correct (I know I don’t
have one of them right):

R = reference signal

P = perceptual signal

E = error the difference between R & P

The little blue circle where R and P are
inputs is the comparator.

What do you call the red triangle? I
would call it an effector.

O = output

D = disturbance

Output and disturbance are inputs to a
little green circle. That little green circle is what I would think of as
an integrator.

S = ??? Does it refer to signal of
some kind? My current understanding of such matters is that I would think
of it as the net effect of my actions (output) plus whatever effect the disturbance
is having on the controlled variable. However, that seems to be incorrect
in light of what you’re saying.

So, what is S and what is the yellow
rectangle?

One other question: You show arrows going
off at angles from O and P. What do those represent?

More questions embedded below…

[Martin Taylor 2004.12.11.13.14]

[From Fred Nickols (2004.12.11.1145 EST)]

I thought “disturbances” were
factors that interfered with my control of a variable.

This isn’t actually wrong, but it’s
misleading, and it is probably not a good idea to think of disturbances like
that.

Why is it misleading? Because
"interfering with my control of a variable’ seems to imply that something
has affected your ability to influence the variable. In other words,

thinking of the canonical diagram of an
ECU,

it would be altering the bottom feedback path. A disturbance doesn’t do
that.

[From Fred: What are you designating
the “bottom feedback path”? From O to little green circle or
on through to the yellow rectangle? But a disturbance would alter the
perceptual signal, right?]

Why is it not actually wrong? What a disturbance does, if you do
nothing to oppose it, is alter the value of the input to the perceptual
function, and therefore the value of the perceptual signal. So its immediate
effect is to create error. The output function inputs the error and outputs an
effect on the disturbed variable, reducing the error. So, if you think
“interfering with my control” is synonymous with “increasing the
error in the controlled variable”, then a disturbance does that – without
interfering with the ECU’s ability to control.

[From Fred: Would I be correct in
assuming that “the perceptual function” is the yellow rectangle and
that S is the input to it? If so, and if the little green circle
represents the controlled variable “out there”, then output and
disturbance are both affecting it and through it, the input (S) to the
perceptual function. If I’ve got this wrong feel free to whack me
alongside the head.]

A disturbance isn’t the only thing that can create momentary error. The
same happens if the reference value changes. Disturbance and Reference are the
only two signals that enter the loop from outside. So, even if there is nothing
in the environment affecting the controlled variable except the output of the
control system, transient errors may still occur. But you wouldn’t say that a
reference signal interferes with your control of a variable, would you? So I
think it’s not a good idea to think of a disturbance interfering with your
control, either.

[From Fred: Well, actually, I would
say that changing a reference value does (or can) interfere with my ability to
control a variable. If the new value represents a degree of control that
greatly exceeds anything I’ve accomplished before it seems to me that it
might be beyond my current ability to control.]

So, if the value of the disturbance is zero, and the error is zero, the
output must be zero, and your “No action required” suggestion would
be correct.

[From Fred: What you say directly
above is what I thought I was saying first time around. I obviously
didn’t say it very well.]

But if there is a steady external influence on the environmental
variable, then to maintain zero error in the perceptual signal, the output must
be equal, opposite, and continuing.

[From Fred: I take the sentence
immediately above to mean that if disturbances are present that I must act to
keep the controlled variable in whatever state I want it to be.]

[From Fred Nickols (2004.12.10.1825 EST)]

Further, “keep doing what you’re doing” implies ongoing

action. What about start up or initiation?

Does the foregoing answer that? Start-up implies a change of reference
signal, doesn’t it?

My recollection

of the way a simple integrator works is that, first of all, it has more
than

one input. It was also a dynamic device. If its inputs go
to zero, its

outputs do too.

I don’t know what you called an integrator in your electro-mechanical
analogue computer days, but what I meant was a function that implemented the
mathematical operation of integration. I can’t easily use math notation here,
but in words:

x(t1) = x(t0) + (integral from time t0 to t1 of y(t) dy)

Here, y(t) is the input waveform to the integrator, x(t) its output.

if y(t) has been zero between t0 and t1, x(t1) = x(t0), not zero.

Martin

[Bill_Powers1]

I thought “disturbances” were
factors that interfered with my control of a
variable.
[From Bill Powers (2004.12.11.1335 MST)]
Fred Nickols
(2004.12.11.1145 EST) –
Disturbances are simply factors other than your own actions that
independently affect a variable. A wind disturbs the leaves on a tree. An
earthquake disturbs the land and things on it. Sunlight disturbs the
temperature of the air, the sidewalk, your skin. This happens whether
you’re controlling those things or not.
Everything you try to control is affected by events and forces in the
environment that occur whether you act or not. Since many of the things
you try to control involve relationships between you and something else,
anything that alters either what you do or what the environment does is a
disturbance that can perturb the controlled relationship. Your own
previous actions, because they can change the state, position ,
orientation, or velocity of your body, can be disturbances of
relationships between your body and other objects (or people).
The way you can tell that a disturbance is present or has acted is that
in order to achieve a result you previously achieved, you have to act in
a way that is somewhat different from the way you previously acted. If
there were no disturbances, you would find that to repeat a given result
you would have to act exactly as you did the last time you
produced it. That is very rare, and the more carefully you examine the
details of control actions, the rarer it looks.

If
someone had a string attached to the toothpaste tube and jerked it out of
my reach every time I reached for it, that’s something I would consider a
“disturbance.” But, the toothpaste tube lying there passive and
inert and me wishing to grasp it, reaching for it and grasping it?
Where is the disturbance in that? I’m getting confused between
“error” a difference between the perceived state of a controlled variable
and a reference signal for that variable and a “disturbance.” Is
your point essentially that the last time I used the toothpaste I (or
someone else since) set it in some other place and that difference
constitutes the disturbance? Obviously, I don’t get
it.

In the case I described, the disturbance was not caused by any movement
of the toothpaste tube, but by the fact that your own body starts in a
different position and orientation each time you reach for the
toothpaste. Your own behavior prior to reaching is the disturbance. The
muscles that operate your arm and hand must NOT repeat the same tensions
they produced the last time you reached for the toothpaste. If they did,
your hand would come to the same position it did before, not relative to
the toothpaste but relative to your body. That would leave your hand to
one side of the toothpaste, or above or below it, or short of it or
beyond it.

What happens is that your brain adjusts the muscle tensions to make the
visual image of the hand (or kinesthetic image, in the dark) approach the
position of the toothpaste image, continually correcting the muscle
tensions as the distance is reduced, and ending up with the hand touching
the toothpaste tube regardless of where the hand started or what other
changes have happened before or during the movement.

This will happen even if someone else has moved the toothpaste, or ties a
string to it and pulls, as you suggest. It will happen if your muscles
grow fatigued so they respond less to the neural signals driving them.
Your control systems will alter their neural outputs as required to make
the controlled input match the reference level.

Note that the concept of operant conditioning assumes that repeating the
same behavior will always cause the same result to occur. If that were
not true, repeating a given behavior that previously led to a
reinforcement would not lead to another reinforcement, or a different
behavior would be needed to produce the same reinforcement. If the
behavior changed so as to continue producing the reinforcement, the
idea of reinforcement would have to be rejected. So what does happen when
you change the conditions that make the reinforcer depend on behavior?

Best,

Bill P.

[Martin_Taylor2]

Re: No Action Required?
[Martin Taylor 2004.12.11.16.38]

[From
Fred Nickols (2004.12.11.1435 EST)] -

Many
thanks for the reply, Martin, and many more for the animated drawing.
I sat and watched it for several minutes. (How did you do
that?)

Before
I burden you with any more questions, let’s see if I have the
symbols correct (I know I don’t have one of them
right):

R =
reference signal
P =
perceptual signal
E =
error the difference between R & P

The
little blue circle where R and P are inputs is the
comparator.

What
do you call the red triangle? I would call it an
effector.

Most people call it “the output function” in
PCT-land.

O =
output
D =
disturbance

Output
and disturbance are inputs to a little green circle. That little
green circle is what I would think of as an
integrator.

That’s where the confusion lies. It’s just an adder or combiner.
There’s no special PCT-ish name for it that I know.

S =
??? Does it refer to signal of some kind?

Sensory input.

So,
what is S and what is the yellow rectangle?

Yellow rectangle = perceptual input function.

One
other question: You show arrows going off at angles from O and P.
What do those represent?

The one from O represents effects on aspects of the world that do
not influence the ECU’s perceptual signal.

The signal from P is the perceptual signal that forms the input
to higher-level ECUs in the complex control system that is the HPCT
hierarchy (or whatever other structure you care to imagine).

[Martin
Taylor 2004.12.11.13.14]

[From
Fred Nickols (2004.12.11.1145 EST)] -

I
thought “disturbances” were factors that interfered with my
control of a variable.

This
isn’t actually wrong, but it’s misleading, and it is probably not a
good idea to think of disturbances like that.

Why is
it misleading? Because "interfering with my control of a
variable’ seems to imply that something has affected your ability to
influence the variable. In other words,

thinking of the canonical diagram of an
ECU,
it would be
altering the bottom feedback path. A disturbance doesn’t do
that.

[From
Fred: What are you designating the “bottom feedback path”?

From O to little green circle or on through to the yellow rectangle?
But a disturbance would alter the perceptual signal,
right?]

To the green circle. Beyond that, the disturbance has been
combined. Everything up through to signal P is still technically part
of the feedback path, but I only meant up to the green combiner
circle.

Why is it
not actually wrong? What a disturbance does, if you do nothing to
oppose it, is alter the value of the input to the perceptual function,
and therefore the value of the perceptual signal. So its immediate
effect is to create error. The output function inputs the error and
outputs an effect on the disturbed variable, reducing the error. So,
if you think “interfering with my control” is synonymous
with “increasing the error in the controlled variable”, then
a disturbance does that – without interfering with the ECU’s ability
to control.

[From
Fred: Would I be correct in assuming that “the perceptual
function” is the yellow rectangle and that S is the input to it?
If so, and if the little green circle represents the controlled
variable “out there”, then output and disturbance are both
affecting it and through it, the input (S) to the perceptual
function. If I’ve got this wrong feel free to whack me
alongside the head.]

So far, so good. The “out there” variable has been
called the “CEV” for “Controlled Environmental
Variable”.

A
disturbance isn’t the only thing that can create momentary error. The
same happens if the reference value changes. Disturbance and Reference
are the only two signals that enter the loop from outside. So, even if
there is nothing in the environment affecting the controlled variable
except the output of the control system, transient errors may still
occur. But you wouldn’t say that a reference signal interferes with
your control of a variable, would you? So I think it’s not a good idea
to think of a disturbance interfering with your control,
either.

[From
Fred: Well, actually, I would say that changing a reference
value does (or can) interfere with my ability to control a variable.
If the new value represents a degree of control that greatly exceeds
anything I’ve accomplished before it seems to me that it might be
beyond my current ability to control.]

Yes, but that’s a complication of nonlinearity that obscures the
issue you were initially dealing with. Non-linear systems change their
characteristics with changing signal values, and it’s certainly true,
for instance, that if you can reach only 8 ft high, a reference to
stand on the floor and touch a 10 ft ceiling would have a different
effect on your control than would a reference to touch a 7 ft ceiling.
In this discussion, I’ve been assuming linearity, so that the control
equations are the simple ones often bandied around in tutorial
discussions.

So, if the
value of the disturbance is zero, and the error is zero, the output
must be zero, and your “No action required” suggestion would
be correct.

[From
Fred: What you say directly above is what I thought I was saying
first time around. I obviously didn’t say it very
well.]

No, you made the special case into the general case. That’s the
point I was trying to make here. That it’s ONLY true if the
disturbance is zero and the initial condition is that the error is
zero. That’s very rare. It’s kind of like expecting to come into a
room and seeing one of your pencils happening to be standing on its
point. Possible, but one wouldn’t often expect to see it.

But if there
is a steady external influence on the environmental variable, then to
maintain zero error in the perceptual signal, the output must be
equal, opposite, and continuing.

[From
Fred: I take the sentence immediately above to mean that if
disturbances are present that I must act to keep the controlled
variable in whatever state I want it to be.]

Change that to “I must have been acting appropriately and
should continue the same action…”

Martin

[Fred_Nickols3]

[From Fred Nickols (2004.12.11.1716 EST)] –

Thanks for taking time to respond,
Bill. Comments are embedded below.

From Bill Powers (2004.12.11.1335 MST)]

Fred Nickols (2004.12.11.1145 EST) –

I thought “disturbances” were
factors that interfered with my control of a variable.

Disturbances are simply factors other than your own actions that independently
affect a variable. A wind disturbs the leaves on a tree. An earthquake disturbs
the land and things on it. Sunlight disturbs the temperature of the air, the
sidewalk, your skin. This happens whether you’re controlling those things or
not.
Everything you try to control is affected by events and forces in the
environment that occur whether you act or not. Since many of the things you try
to control involve relationships between you and something else, anything that
alters either what you do or what the environment does is a disturbance that
can perturb the controlled relationship. Your own previous actions, because
they can change the state, position , orientation, or velocity of your body,
can be disturbances of relationships between your body and other objects (or
people).
The way you can tell that a disturbance is present or has acted is that in
order to achieve a result you previously achieved, you have to act in a way
that is somewhat different from the way you previously acted. If there were no
disturbances, you would find that to repeat a given result you would have to
act exactly as you did the last
time you produced it. That is very rare, and the more carefully you examine the
details of control actions, the rarer it looks.

[From Fred: Okay; I can see how what
I said initially is less than clear. I’ll try again in light of
what you’d said above. I view “disturbances” as “other
actors and factors” that affect some variable I’m attempting to
control. These include most of the examples you list above. I had
not thought of my own previous actions as disturbances but, klutz that I am, I
can see how that could be the case in some cases. I’m having a
little difficulty squaring that with your first statement above which says “factors
other than your own actions.” For now, I’ll assume it
ties to the distinction you seem to be making between current and previous
behavior.]

In the case I described, the disturbance was not caused by any movement of the
toothpaste tube, but by the fact that your own body starts in a different
position and orientation each time you reach for the toothpaste. Your own
behavior prior to reaching is the disturbance. The muscles that operate your
arm and hand must NOT repeat the same tensions they produced the last time you
reached for the toothpaste. If they did, your hand would come to the same
position it did before, not relative to the toothpaste but relative to your
body. That would leave your hand to one side of the toothpaste, or above or
below it, or short of it or beyond it.

What happens is that your brain adjusts the muscle tensions to make the visual
image of the hand (or kinesthetic image, in the dark) approach the position of
the toothpaste image, continually correcting the muscle tensions as the
distance is reduced, and ending up with the hand touching the toothpaste tube
regardless of where the hand started or what other changes have happened before
or during the movement.

This will happen even if someone else has moved the toothpaste, or ties a
string to it and pulls, as you suggest. It will happen if your muscles grow
fatigued so they respond less to the neural signals driving them. Your control
systems will alter their neural outputs as required to make the controlled
input match the reference level.

Note that the concept of operant conditioning assumes that repeating the same
behavior will always cause the same result to occur. If that were not true,
repeating a given behavior that previously led to a reinforcement would not
lead to another reinforcement, or a different behavior would be needed to
produce the same reinforcement. If the behavior changed so as to continue
producing the reinforcement, the idea of reinforcement would have to be rejected.
So what does happen when you change the conditions that make the reinforcer
depend on behavior?

[From Fred: Hmm. I’m no more
of an expert in operant conditioning than I am in PCT but I think the
behaviorists’ definition of an “operant” in behavioral terms
accounts for the difference. The folks I know would probably say that it
is a general behavior pattern that is being reinforced – a.k.a. an “operant”
– not a particular configuration or motion of that configuration.
In other words, the experimenters are reinforcing “pressing the lever”
not a particular body movement that also happens to result in pressing the
lever (and, no, I’m not about to go into “successive approximations”.)
But, there are far more qualified people on this list than I to respond to that
one.

Thanks, Bill. I appreciate your
time, interest and effort.]

[Fred_Nickols3]

Re: No Action Required?

[From Fred Nickols (2004.12.11.1705 EST)]

···

–

Thanks again, Martin. I think my understanding
grows closer to yours.

More comments embedded below. I’ve
trimmed all but the parts to which I’m replying.

But, first, how did you make that animated
diagram?

Martin Taylor 2004.12.11.16.38]

[From Fred Nickols (2004.12.11.1435 EST)]

What do you call the red triangle? I
would call it an effector.

Most people call it “the output function” in PCT-land. [From Fred: OK; that makes
sense to me]

O = output

D = disturbance

Output and disturbance are inputs to a
little green circle. That little green circle is what I would think of as
an integrator.

That’s where the confusion lies. It’s just an adder or combiner.
There’s no special PCT-ish name for it that I know. [From Fred: Yep, that’s where the confusion (at
least part of it) lies. An “adder” or “combiner”
is what I think of as an integrator. It was shown in the old computer
schematics as a circle with an X inside. It could have two or more inputs
and produce a single output (or two outputs). Or, it could have one input
and several outputs (in which case it was called a “differentiator”
instead of an “integrator”). “Adder” or “combiner”
works for me.

S = ??? Does it refer to signal of
some kind?

Sensory input.

So, what is S and what is the yellow
rectangle?

Yellow rectangle = perceptual input function.

[From Fred: Thanks for clearing
those up. I don’t feel nearly as stupid now.]

One other question: You show arrows going
off at angles from O and P. What do those represent?

The one from O represents effects on aspects of the world that do not
influence the ECU’s perceptual signal.
[From Fred: Thanks]

The signal from P is the perceptual signal that forms the input to
higher-level ECUs in the complex control system that is the HPCT hierarchy (or
whatever other structure you care to imagine).

[From Fred: So you’re saying
there’s some kind of splitting function; the same signal goes up a level
and stays at the level where it is. It’s processed at one level and
one or more levels up?]

[From Fred: Thanks for clearing
things up for me, Martin. I think my grasp of matters is improved.

[Acree_Michael]

[From Mike Acree (2004.12.13.1515 PST)]

Bill Powers(2004.12.11.0753 MST)--

A reference signal can't help a child learn to stand upright and move
about, can it? I think what the child learns are the lower-order
perceptions and the actions needed to maintain an upright posture in
relation to the environment, a perception that is acquired when you

help

the child stand upright for a moment. The child (sooner or later) wants

to

repeat this experience; the perception is remembered and becomes a
reference signal against which the current state of pereived

uprightness is

compared. That comparison produces an error signal, and the child then
starts learning what it has to do with its lower control systems to

cause

that perception to appear again, erasing the error signal.

A professor of mine, Al Trieschman, observed that the pattern of a
child's motor learning was that first we sit our babies up, then they
learn to sit up by themselves; we stand them up, then they stand by
themselves; we walk them, then they walk by themselves. He thought this
was why many children expected to be able to fly, because the one thing
we do for them that they can't do by themselves is to pick them up.

Mike

[B_Thalhammer]

[From Bryan Thalhammer (2004.12.13.2000 CST)]

Mike,

The other day I was at a party, and there was a young family who had brought
their two little twin kids. They were like 3 months old (tell me if I am off
here) and just getting comfortable on their Sitz bones, sitting up straight,
but had not yet figured out how to crawl.

One was faster at learning or whatever than the other, and was just getting
up on all fours. The other one, more vocal, watched his brother make it up
on hands and knees, and twisted over to his belly. He strained and strained
pushing his arms down, but couldnt figure out how to get those knees under
him. Cute kids.

So he started sniffling and yelling, and so, I helped out, getting his knees
bent under him, with some encouraging words. All of a sudden, he stopped
screeching, finding himself in a desirable position. He looked at me, and at
his mother, and for a sec, he seemed to be smiling at what he had done.

Then, not knowing what to do, he started to get a little crabby, and his
mother picked him up and he had a good cry.

What was HE controlling for?

I agree that suggesting new reference settings (sit, crawl, stand, and fly)
all seem to be positions WE put them in, either as a way of positioning our
new little dolly, or in an "unconscious-innate-learned before speech???"
way, stand them up proper. Perhaps, kids finding their stomachs in a less
regurgitative state, their eyes higher above the ground, their inner ear in
a new orientation, and their noses above the stinky rug or crib or backside,
they want to preserve that position, to control for it, and their random
behavior gradually keeps them in positions more and more favorable, but its
only in their (our) dreams that they fly.

So, while I do not propose that one can give reference signals, or even
really suggest them to kiddies, BEING put in that position does bring some
benefits, and they (we) start to control for those positions, exploring
parts fo the environment these new positions allow them to reach with their
eyes, mouths and hands.

Good idea Mike!

--Bryan

···

[From Mike Acree (2004.12.13.1515 PST)]

Bill Powers(2004.12.11.0753 MST)--

>A reference signal can't help a child learn to stand upright and move
>about, can it? I think what the child learns are the lower-order
>perceptions and the actions needed to maintain an upright posture in
>relation to the environment, a perception that is acquired when you
help
>the child stand upright for a moment. The child (sooner or later) wants
to
>repeat this experience; the perception is remembered and becomes a
>reference signal against which the current state of pereived
uprightness is
>compared. That comparison produces an error signal, and the child then
>starts learning what it has to do with its lower control systems to
cause
>that perception to appear again, erasing the error signal.

A professor of mine, Al Trieschman, observed that the pattern of a
child's motor learning was that first we sit our babies up, then they
learn to sit up by themselves; we stand them up, then they stand by
themselves; we walk them, then they walk by themselves. He thought this
was why many children expected to be able to fly, because the one thing
we do for them that they can't do by themselves is to pick them up.

Mike

[Richard_Thurman]

[From Richard Thurman (2004.12.20.2100)]

>Bill Powers(2004.12.11.0753 MST)

Fred Nickols (2004.12.11.0534) --

>>Is it not the case that a child

learning to walk is developing (through reorganization or learning or just
plain practice) the reference signals necessary to control the
perception of
standing upright and moving about?

A reference signal can't help a child learn to stand upright and move
about, can it? I think what the child learns are the lower-order
perceptions and the actions needed to maintain an upright posture in
relation to the environment, a perception that is acquired when you help
the child stand upright for a moment. The child (sooner or later) wants to
repeat this experience; the perception is remembered and becomes a
reference signal against which the current state of pereived uprightness is
compared. That comparison produces an error signal, and the child then
starts learning what it has to do with its lower control systems to cause
that perception to appear again, erasing the error signal.

>There are other routes to acquiring a reference signal fort being upright,
>but that's one scenario that seems reasonable.
>
>Best,
>
>Bill P.

Bill, could you elaborate a little more on how a control system
acquires reference signals and how they are used? I'm currently
struggling with the idea of memory as a reference signal. It wasn't
until yesterday that the thought of memory as a reference signal
really took hold. The standard PCT diagram does not show memory's
function in references -- and so I don't usually think about memory
when modeling a control system. However, as I was looking at the
diagrams in BCP (page 218 & 221). The idea of memory as reference
began to make sense. But I'm have a bit of a struggle with it. Any
ideas about how an individual can use memory in the control process
would be greatly appreciated. (And any ideas on how to build such a
process in code would be incredibly appreciated.)

Best wishes,
Rich Thurman