Looking for Trouble

[From Bruce Gregory (990714.1305 EDT)]

One problem some people have with accepting PCT is the conviction that
we act in order to achieve goals. The PCT view is that we most often act
to counter disturbances, not to achieve goals. (If control were perfect,
this would always be true.) One cause of misunderstanding may be a
failure to distinguish between actions under two differing
circumstances. In the first case, the actions _are_ the perception I am
controlling. In the second case the actions counter a disturbance to a
different perception that I am controlling. For example, if you ask me
why I am going to my garage in the morning, my answer would that I need
to be at a meeting in Cambridge at 10:00 a.m. This certainly sounds like
a goal, and it seems if my actions are "generated" by my efforts to
reach this goal.

The story from a PCT perspective it that what I am doing is controlling
a sequence of perceptions, one of which is the perception that I am
walking from the foyer to the garage. My goal is to perceive myself
walking and that is what I perceive--I am successfully exercising
control--I have achieved this particular goal. This is what I mean by my
actions being the perception I am controlling. If I tripped, my actions
would be those necessary to stabilize the perception that I am walking.

I don't think I've seen a discussion that makes this distinction clear.
Perhaps because I haven't looked in the right place, or because my
thinking is off base. (In any case, the meeting at 10:00 is at the send
of a length sequence of desired perceptions, but it plays no causal role
in my trip to my garage.)

Have I got any of this right, Rick?

Bruce Gregory

i.kurtzer (990714.1525)

[From Bruce Gregory (990714.1305 EDT)]

I do not think it would be accurate to say that..

The PCT view is that we most often act
to counter disturbances, not to achieve goals. (If control were perfect,
this would always be true.)

if only this is a casting-nets type of question, if by "achieve goal" you
mean a shift in a reference. I don't think the perfection of control enters
into the question. How would it, and what would that mean?

One cause of misunderstanding may be a
failure to distinguish between actions under two differing
circumstances.

If by between a shift in reference signal and a output correlated with a
disturbance than people have distinguished between "emitted" and "elicted"
behavior. I don't think that is sufficient.

In the first case, the actions _are_ the perception I am
controlling. In the second case the actions counter a disturbance to a
different perception that I am controlling.

I don't think I've seen a discussion that makes this distinction clear.
Perhaps because I haven't looked in the right place, or because my
thinking is off base.

Well, the following statement is disagreeable.

(In any case, the meeting at 10:00 is at the send
of a length sequence of desired perceptions, but it plays no causal role
in my trip to my garage.)

The "meeting at 10:00" is at the beginning and throughout the trip as a
reference and plays the causal role as one signal entering a comparator,
along with a perceptual signal, that drives the output. Thats why you drive
faster when you leave your house at 9:45 than 8:00.

i.

···

Have I got any of this right, Rick?

Bruce Gregory

[From Bruce Gregory (990714.1700 EDT)]

i.kurtzer (990714.1525)

if only this is a casting-nets type of question, if by
"achieve goal" you
mean a shift in a reference. I don't think the perfection of
control enters
into the question. How would it, and what would that mean?

I use goal in the sense of reference level.

> One cause of misunderstanding may be a
> failure to distinguish between actions under two differing
> circumstances.

If by between a shift in reference signal and a output
correlated with a
disturbance than people have distinguished between "emitted"
and "elicited"
behavior. I don't think that is sufficient.

I'm not sure what you mean. No behavior in my view is either emitted or
elicited.

Well, the following statement is disagreeable.

> (In any case, the meeting at 10:00 is at the send
> of a length sequence of desired perceptions, but it plays
no causal role
> in my trip to my garage.)
>

The "meeting at 10:00" is at the beginning and throughout the
trip as a
reference and plays the causal role as one signal entering a
comparator,
along with a perceptual signal, that drives the output.
Thats why you drive
faster when you leave your house at 9:45 than 8:00.

What control system do you envisage as having "meeting at 10:00" as an
input? If I look at my watch and it is 8:45, what does this control
system do? (If I leave my house at 9:45 for a 10:00 meeting, I have to
drive _really_ fast! [About 300 mph]). The speed at which I drive is
determined by my desire to perceive myself driving at that speed. As I
see it, its part of the planning process.

Bruce Gregory

[From Rick Marken (990714.1550)]

Bruce Gregory (990714.1305 EDT)--

The PCT view is that we most often act to counter disturbances,
not to achieve goals.

I think the PCT view is that we act to counter disturbances
to controlled variables _and_ to bring those variables into
a match with what are often time varying reference signal values.
Actions often are doing _both_ of these things simultaneously.
For example, when you move your finger to point at different
locations on the screen, one controlled variable is the visual
position of your finger; this variable is controlled relative
to a time varying reference signal which specifies the different
visual locations for the finger. Varying limb positions are the
actions that affect the state of the controlled variable (visual
finger position). These actions move the finger to the different
positions specified by the reference signal _while_ compensating
for disturbances (such as changing forces on the finger) that
would keep the finger from reaching each reference state.

One cause of misunderstanding may be a failure to distinguish
between actions under two differing circumstances. In the first
case, the actions _are_ the perception I am controlling. In the
second case the actions counter a disturbance to a different
perception that I am controlling...

I don't think I've seen a discussion that makes this distinction
clear.

What you are talking about is _hierarchical control_; an "action"
used by a higher level system to keep its perception under control
(matching a _possibly varying_ reference signal) can contribute
to the _reference_ for the perception controlled by a lower
level system _as the means_ of controlling the higher level
perception. For example, the action used by the system controlling
visual finger location is actually a _reference_ for the limb
positions _to be perceived_ by the lower level, limb position
perception control system. The lower level action (in this case,
limp positioning) is not _itself_ a controlled perception; it's a
_reference_ for the state of a controlled perception. At the
lower level (limb position control) limb position in controlled
(by setting references for still lower level, muscle force
perceptions) relative to the _varying_ reference sent by the
higher level, visual finger position control system. The control
systems at the lowest level of this hierarchy are the actions
(rather than reference settings) that actually have physical effects
on the real world (in this example, these actions are probably
variations in muscle tension that produce the forces that move the
limbs). These lowest level actions are not perceptive (and, thus,
not controlled).

It does take a while to understand this. It might help if you
try running my Excel "hierarchical control" spreadsheet simulation.
But that's pretty abstract. When I get the memory experiment
running (maybe a few weeks) I think it will show this hierarchical
relationship between control systems. Oh, and Bill's "Little Man"
or "Arm pointing" demo shows these hierarchical relationships
beautifully.

Best

Rick

···

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

i.kurtzer (990714.2030)

[From Bruce Gregory (990714.1700 EDT)]

i.kurtzer (990714.1525)
>
> if only this is a casting-nets type of question, if by
> "achieve goal" you
> mean a shift in a reference. I don't think the perfection of
> control enters
> into the question. How would it, and what would that mean?

I use goal in the sense of reference level.

I always give you the benefit of doubt.

> > One cause of misunderstanding may be a
> > failure to distinguish between actions under two differing
> > circumstances.
>
> If by between a shift in reference signal and a output
> correlated with a
> disturbance than people have distinguished between "emitted"
> and "elicited"
> behavior. I don't think that is sufficient.

I'm not sure what you mean. No behavior in my view is either emitted or
elicited.

I agree that that distinction is not a very good one. I'm just saying it
was attempt to divide the causes of action into enviromental events and
those initiated by the organism.

>
>
> Well, the following statement is disagreeable.
>
> > (In any case, the meeting at 10:00 is at the send
> > of a length sequence of desired perceptions, but it plays
> no causal role
> > in my trip to my garage.)
> >
>
> The "meeting at 10:00" is at the beginning and throughout the
> trip as a
> reference and plays the causal role as one signal entering a
> comparator,
> along with a perceptual signal, that drives the output.
> Thats why you drive
> faster when you leave your house at 9:45 than 8:00.

What control system do you envisage as having "meeting at 10:00" as an
input? If I look at my watch and it is 8:45, what does this control
system do? (If I leave my house at 9:45 for a 10:00 meeting, I have to
drive _really_ fast! [About 300 mph]). The speed at which I drive is
determined by my desire to perceive myself driving at that speed. As I
see it, its part of the planning process.

The input to the control system would not be "meeting at 10:00". The
reference for that control loop would be "be at the meeting come 10:00".
According to legend that could be classed with those controlled variables
called principles. Being at meetings on time is one thing that is part
about being a "good worker" which is a system concept. To that one would
vary programs such as..1st eat at McD's, 2nd drop off kids by 9:00, get gas
by 9:15, finish cigarette by 9:50, wash hands and check breath by 9:55,
step into office at 10:00. All those items within a program are
perceptions that are ticked off so that the the perception of "being in the
office at 10:00" matchs the reference of "be at the office at 10:00".
Driving at a certain speed is doubtfully planned at program level. If it
is then your plan better encounter a very stable world where there are no
unexpected traffic jams. That does not mean that we do not have some rough
and ready memories of how fast one can get to work, which lanes are the
quickest, and so one, just that 55mph is doubtfully a programmed variable.
Of course this should all be researched and this is just the gist.

i.

[From Bruce Gregory (90729.1315 EDT)]

i.kurtzer (990714.2030)

The input to the control system would not be "meeting at 10:00". The
reference for that control loop would be "be at the meeting
come 10:00".
According to legend that could be classed with those
controlled variables
called principles. Being at meetings on time is one thing
that is part
about being a "good worker" which is a system concept.

This raises an issue that I have questions about. In the hierarchical
model, you think of the "principle level" as establishing references for
systems below it in the hierarchy. In this sense, the principle level is
"in charge". This doesn't seem to fit my experience. Most of the time,
principles do _not_ seem to be in charge. Perhaps this is simply because
my attention is elsewhere. It may be that attention is not drawn to
systems in which control is being exercised successfully. Perhaps
attention "shows up" only when control is problematical. We know that a
mechanism in the visual system moves the focus of attention to motion in
the visual field outside the fovea region. If we think of the reference
level as being "no motion", motion generates an error signal that is
eliminated by moving the central visual field so that it encompasses
whatever is moving. Perhaps this is a general mechanism that moves
attention to where it is likely to be most useful.

Bruce Gregory

i.kurtzer (990729.2030)

[From Bruce Gregory (90729.1315 EDT)]

i.kurtzer (990714.2030)

This raises an issue that I have questions about. In the hierarchical
model, you think of the "principle level" as establishing references for
systems below it in the hierarchy. In this sense, the principle level is
"in charge". This doesn't seem to fit my experience. Most of the time,
principles do _not_ seem to be in charge. Perhaps this is simply because
my attention is elsewhere.

I would think that both the issue of delineating levels and the focus of
attention are researchable topics.

i.

[From Bruce Nevin (990730.1006 EDT)]

Bruce Gregory (90729.1315 EDT)--

In the hierarchical
model, you think of the "principle level" as establishing references for
systems below it in the hierarchy. In this sense, the principle level is
"in charge". This doesn't seem to fit my experience. Most of the time,
principles do _not_ seem to be in charge. Perhaps this is simply because
my attention is elsewhere.

You may be confusing principles with precepts. At the risk of being fatuous
--please don't take offense -- principles are what you actually want; you
can identify them by observing what you do (what disturbances you resist).
Precepts are easier to identify as statements of what you should want; you
can identify them by interrogation. Conflict between them is called hypocrisy.

  Bruce Nevin

[From Bruce Gregory (990730.1457 EDT)]

Bruce Nevin (990730.1006 EDT)

You may be confusing principles with precepts.

I may be, but I don't think so.

At the risk of
being fatuous
--please don't take offense -- principles are what you
actually want; you
can identify them by observing what you do (what disturbances
you resist).

By observing the fact that I drive back and forth between Pomfret,
Connecticut and Cambridge, Massachusetts five days a week, you can infer
that some higher level perception is establish the reference level
associated with the control of the perceptions associated with this
behavior. I was simply noting that I am normally not aware of what that
perception is unless you ask me about it. The fact that I am
successfully controlling the perception seems to "signal" awareness that
its presence is not needed.

Precepts are easier to identify as statements of what you
should want; you
can identify them by interrogation. Conflict between them is
called hypocrisy.

I view hypocrisy somewhat differently. The person who has been labeled
as a hypocrite may not be experiencing any conflict. When we label
someone a hypocrite we seem to be saying, "If I said that and I acted
that way, _I_ would experience conflict." Calling another a hypocrite
says something about oneself, not necessarily about the other.

Bruce Gregory

[From Erling Jorgensen (990730.1530 CDT)]

Bruce Gregory (90729.1315 EDT)

This raises an issue that I have questions about. In the hierarchical
model, you think of the "principle level" as establishing references for
systems below it in the hierarchy. In this sense, the principle level is
"in charge". This doesn't seem to fit my experience. Most of the time,
principles do _not_ seem to be in charge.

Here is my take on it...

For control systems to work in a hierarchy, connections must be established.
At a minimum, that means perceptions being passed on from lower levels (and
being reconstructed according to the perceptual input function of the control
system in question, here one at the "Principles" level), as well as relevant
output connections to lower levels that can impact the desired perception.

In my experience, principles are generated more "locally," that is with regard
to a specific context, which then may get generalized to other contexts. I
have also postulated in the past on the net (though I have not figured out
a way to test it), that the nature of a principle-perception may be that of
a probability distribution, which is judged to be true or not true _on average_.
For instance, if I _usually_ act "honestly", then I may let a specific slip go
by (perhaps one evaluated by a parallel principle as being "minor") and not
take action to correct it.

These points suggest various ways that a given principle may not be experienced
as "in charge":

a) lower level perceptions may not be perceived as relevant -- i.e., the
ascending connection may not be there! -- in which case we generally say "the
principle does not apply."

b) a given principle control system may not know how to affect lower level
perceptions -- i.e., the output connections may not yet be established -- in
which case we speak of "acting inconsistently with one's principles."

c) if principles are enacted and evaluated on the basis of multiple occurrences
-- i.e., a distribution must be set up first -- then 1) there may not be enough
cases to bring the principle into play ("that part of my value system is still
in flux"), or 2) a given occurrence may be an acceptable outlier ("it's an
exception") -- although in this instance, we usually think of that principle
still being operative.

d) You suggest a fourth possibility, namely --

Perhaps this is simply because my attention is elsewhere.

At the present time, I am very reluctant to think that control systems operate
on the basis of attention. To me, attention seems more like a monitoring
system. I agree, however, that it seems related to the presence of sufficient
error -- as when you say,

Perhaps attention "shows up" only when control is problematical.

We know that a
mechanism in the visual system moves the focus of attention to motion in
the visual field outside the fovea region. ... Perhaps this is a general
mechanism that moves attention to where it is likely to be most useful.

I like this as a metaphor for how to think about attention. Attention follows
error, just as our visual system follows motion. Where I disagree is in
whether attention does anything "useful," at least in terms of initiating
corrective action. When I lose my balance or start to twist an ankle, my
attention is immediately drawn to it, but it seems I am already in the process
of doing something about it, like shifting weight to the other leg and trying
to brace an upcoming fall. I don't think those systems waited for attention
to give any commands.

I guess that's how I currently view principles, and attention, being integrated
into the hierarchical PCT proposal.

All the best,

        Erling

[From Bruce Gregory (990730.1723 EDT)]

Erling Jorgensen (990730.1530 CDT)

At the present time, I am very reluctant to think that
control systems operate
on the basis of attention. To me, attention seems more like
a monitoring
system.

Yes, I agree. What I was trying to say was that I might not be _aware_
that the system controlling for the perception of a principle was "in
charge".

I like this as a metaphor for how to think about attention.
Attention follows
error, just as our visual system follows motion. Where I
disagree is in
whether attention does anything "useful," at least in terms
of initiating
corrective action. When I lose my balance or start to twist
an ankle, my
attention is immediately drawn to it, but it seems I am
already in the process
of doing something about it, like shifting weight to the
other leg and trying
to brace an upcoming fall. I don't think those systems
waited for attention
to give any commands.

Again, I agree completely.

Bruce Gregory