B:CP Chapter 4

Chapter 4, Feedback and Behavior

Study Guide

How would you describe
the main argument of this chapter? How does it relate to the question of whether
any behavior is open - loop?

The input
into our senses changes the output from our muscles, which changes the
environment, which changes the input into our senses. And this happens in more
of a simultaneous manner than a sequential manner. When one is doing a task in a research study, there are task instructions which the person is asked to follow. This makes all research studies a control task. Is the person following the task instructions?

Consider the following
statement: When one observes a behavior, PCT suggests that one should ask what
aspect(s) of a person’s experience is being altered, influenced, or regulated
by the behavior. Agree or disagree? Explain your answer.

I agree
with this. The function of a behavior for a person is to control some specific experience
within the person. To better understand a person, one should attempt to
identify the experience the person wants to control.

Consider the following
statement: It is found out that a person is engaging in self-cutting. What do
you think a person might be controlling by this action?

There is
probably no one answer. Maybe the person is being overwhelmed by stress and
this serves to reduce the stress in some way. The person may think that
punishment is needed and this serves to punish the person. The person may get
excited by seeing the blood come out. The difference between a PCT approach and
other ones, is that some effort will be made to test out the hypothesis of what
experience is being controlled.

If a person has no
preferences, that is, the person keeps saying “I don’t careâ€?, what do you think
the person might be controlling?

Again,
this question probably has no one answer. The person could be depressed and has
stopped trying to get what he/she wants. The person might be experiencing
frustration and this is a way to reduce it—Whatever happens is OK.

Do you think that a
person could “over-do-itâ€? when it comes to controlling his/her experiences?
Explain your answer.

Once again, there is probably no one correct
answer. What occurs to me is that the person might be “obsessive/compulsiveâ€?
and might be trying to control everything in his/her life to a high degree.
There are no priorities.

Try the leading questions at the end of this chapter. I will
hold off on these for now.

[From Lloyd Klinedinst (2013.08.03.1048 CDT)]

Here’s my read and replies for the current chapter.

In Chapter 3 Bill established premises regarding the internal arc of the control loop. In Chapter 4 he presents the external arc, the element of feedback, its fact, its role in control (FEEDBACK AS CONTROL section). He also deals with how purpose and goal, bugaboo terms in behavioral science, can be translated in control loop terms. He even proposes that this analysis of behavior obtains “all of the time”.

  1.  How would you describe the main argument of this chapter?  How does it relate to the question of whether any behavior is open - loop?
    

Bill clearly states his main points of the chapter in the two numbered items of the first paragraph: to analyze feedback and to introduce what kind of a system it might function in, in order to explain how these systems interact with the environment.

I don’t see any allowance for open-loop behavior.

  1.  Consider the following statement: When one observes a behavior, PCT suggests that one should ask what aspect(s) of a person’s experience is being altered, influenced, or regulated by the behavior. Agree or disagree? Explain your answer.
    

I agree. When I, used to viewing in PCT terms, observe an ACTION (I can’t observe behavior as defined in BCP), I consider that some CV in the environment is related to some RS in the person. I informally use TCV to see what the person might be controlling for. But I also realize that any particular RS is nested in an incredibly nested, interrelated complexity of RSs which make up that person’s world/experience.

  1.  Consider the following statement: It is found out that a person is engaging in self-cutting. What do you think a person might be controlling by this action?
    

No idea or too many ideas. I would start asking, testing for what their RS might be.

  1.  If a person has no preferences, that is, the person keeps saying “I don’t care”, what do you think the person might be controlling?
    

For me ‘preferences’ are RSs; a person can’t have NO RSs. Now “I don’t care” is a verbal action that I would need to test for to see what the person might be controlling for.

  1. [sic, 5] Do you think that a person could “over-do-it” when it comes to controlling his/her experiences? Explain your answer.

What I would call ‘tight control’ or ‘rigid control’ in a ‘control freak’ is a setup for continued error and dis-ease.

“Hang loose” or “go with the flow” seems to promise the best chances for an organism to find comfortable accommodation in an ever-changing environment. By this I do not mean the abandonment of all “standards”, personal RSs which define for me who am am and what I am comfortable with in my environment. There seem so far a sufficient number of degrees of freedom in any of my life circumstances to allow for this to happen.

Replies to text questions, p 55.

1 I tell it is ‘clenched’ by pressures felt on my skin in in my muscles, also some sense in my head that this is happening. I haven’t questioned myself on who is doing and experiencing this. Yes, I would say “it’s all perception” of whatever is happening inside and outside of me. It’s the only way ‘feedback’ reaches me.

2 No. No. Other people may let me know what I’m doing, but they’re perceptions, as well.

3 Doing things certainly emphasizes the outwardness of behavior, i.e. control of perceptions …I usually prefer a both-and rather than an either/or perspective. It is such an intimate relationship and constant interaction that an organism has with its environment, whether ‘looking’ inward or outward.

  1. “Not any more” is my answer to all these questions.

  2. Again because of the continuous, never-ending (til death) nature of behavior, even seeing or hearing things coming from the actions of others are still the perceptions of my seeing and hearing behavior. “Keeping one’s eyes and ears open” is one of our constant living behaviors.

  3. I didn’t hesitate or look far; I just closed my eyes. The visual image disappeared.

  4. Yes

  5. Yes

  6. Some increased muscle tension.

  7. Most proximately to stay in my lane, on the road.

  8. Yes.

  9. Yes.

  10. Having breakfast. [Was this question the baker’s dozenth egg?]

  11. About a dozen or so…

Lloyd

Dr. Lloyd Klinedinst
10 Dover Lane
Villa Ridge, MO 63089-2001
HomeVoice: (636) 451-3232

Lloyd Mobile: (314)-609-5571
email: lloydk@klinedinst.com

website: http://www.klinedinst.com

Hi David

The CSGNet server is not working. So nothing posted to CSGNet (such as
this post) is going to be posted to CSGNet. I tried to sign on to
CSGNet as a list manager so that I could try to release the server but
got a message saying the server was offline. So they are either doing
maintenance or CSGNet the server is just kaput. I'm copying this to
Gary Cziko in case he still has connections over there at the U of I
where the list is located.

Best regards

Rick

···

On Thu, Aug 1, 2013 at 8:46 AM, D GOLDSTEIN <davidmg@verizon.net> wrote:

Chapter 4, Feedback and Behavior
Study Guide

1. How would you describe the main argument of this chapter? How does
it relate to the question of whether any behavior is open - loop?
The input into our senses changes the output from our muscles, which changes
the environment, which changes the input into our senses. And this happens
in more of a simultaneous manner than a sequential manner. When one is doing
a task in a research study, there are task instructions which the person is
asked to follow. This makes all research studies a control task. Is the
person following the task instructions?

2. Consider the following statement: When one observes a behavior, PCT
suggests that one should ask what aspect(s) of a person�s experience is
being altered, influenced, or regulated by the behavior. Agree or disagree?
Explain your answer.
I agree with this. The function of a behavior for a person is to control
some specific experience within the person. To better understand a person,
one should attempt to identify the experience the person wants to control.

3. Consider the following statement: It is found out that a person is
engaging in self-cutting. What do you think a person might be controlling by
this action?
There is probably no one answer. Maybe the person is being overwhelmed by
stress and this serves to reduce the stress in some way. The person may
think that punishment is needed and this serves to punish the person. The
person may get excited by seeing the blood come out. The difference between
a PCT approach and other ones, is that some effort will be made to test out
the hypothesis of what experience is being controlled.

4. If a person has no preferences, that is, the person keeps saying �I
don�t care�, what do you think the person might be controlling?
Again, this question probably has no one answer. The person could be
depressed and has stopped trying to get what he/she wants. The person might
be experiencing frustration and this is a way to reduce it�Whatever happens
is OK.

5. Do you think that a person could �over-do-it� when it comes to
controlling his/her experiences? Explain your answer.
Once again, there is probably no one correct answer. What occurs to me is
that the person might be �obsessive/compulsive� and might be trying to
control everything in his/her life to a high degree. There are no
priorities.
Try the leading questions at the end of this chapter. I will hold off on
these for now.

--
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

[Martin Taylor 2013.08.06.17.07]

I guess this is outside the flow of the seminar, but here goes, on the basis that any comments were welcome.

For me, Chapter 4 is a fine introduction to the reasons why we have to deal only with perceptual control, but just as earlier we have to note some caveats about linearity being valid only for small signal excursions and the approximation of neural firing patterns by "neural currents", so in Chapter 4 there are a few caveats to keep in mind. But unless I skipped them somehow, they are caveats Bill failed to mention. Mainly, they are about places where Bill makes absolutist claims such as "never" and "always" and "only", where once one understands PCT these have to be softened into "very seldom", "usually", and "with few exceptions".

For example, the first para on P47 starts: "When feedback is found in the relationship between an organism and its environment, it will either be so weak that it can be ignored, or be negative--it will never be strong and positive." Yet Bill and Rick did at least one study in which they deliberately switched the feedback to be strong and positive -- not for very long, because the subject quickly learned to switch his tracking mode, but strong positive feedback did exist.

More practically, isn't a temper tantrum the result of an uncorrected positive feedback process? And as I discussed with Bill at CSG93, the "Bomb in the Hierarchy" <http://www.mmtaylor.net/PCT/DFS93/DFS93_8.html>, induces reorganization by ensuring that episodes of positive feedback do occur on occasion. Using the analogy that is often used to refer to evolutionary processes, which is equally applicable to e-coli reorganization, the effect of the Bomb is like a sandpile avalanche, usually unnoticed, but occasionally dramatic.

In the same paragraph Bill says "one will always be able to discover what the subject is controlling, for if disturbances are applied that do not in fact disturb the controlled aspect of the environment, the subject's behaviour will not oppose the disturbance." "Always", here is a gross overstatement for two reasons. Firstly, if disturbances are applied that disturb a correlate of the controlled perception, the subject will oppose them, and the only way that one could discover exactly what the subject is controlling is to try all possible variants and see what the subject most precisely opposes. Secondly, if the subject is not controlling well, there is no way to distinguish among the various correlates because for none of them will the opposition to the disturbance be very precise. Thirdly, except in a constrained laboratory condition, what the subject is controlling may change at any moment, as may its reference value if the controlled perception remains controlled. So although "always" may be appropriate in the introductory chapters, Bill should have done as he did earlier, and mention the caveats.

In the middle of the following paragraph he says "all behaviour all the time", when frequently on CSGnet, this is amended to "all intentional behaviour". From outside, one cannot distinguish side-effects from intended effects, nervous tics from an actor's presentation, and so forth. This is a caveat that depends on the meaning of "behaviour". The definition is circular if it depends on the observed actions being effective in influencing a controlled perception, but the statement in the paragraph is overstated if it is not. It's a good teaching statement that should have the desired effect of bringing up short someone not accustomed to thinking this way, but it deserves the same kind of "keep it in mind for later, when we get into detail" caveat as the others.

The first paragraph of page 42 is paraphrased in the second paragraph of the summary, where there's another similar caveat: "...sensory inputs affect and are affected by behaviour. This circle of cause and effect cannot be correctly analyzed as an alternating sequence". But if the lag between action and influence on perception is long, it _must_ be analyzed as an alternating sequence. Think of a mundane example with which I am confronted every day. My shower has a long delay between setting the temperature control and a change in the water temperature, which overshoots before it settles. I turn the handle, then wait to see whether the stable temperature is too hot or too cold, and then turn the handle again in the appropriate direction. In the abstract analysis, the mathematics is difficult precisely because it must be done piecewise, factoring in the "echoes" of previous phases of action after every lag cycle. Again, this is a caveat "to be kept in mind for later", rather than a problem with the exposition.

I think that if I had been writing B:CP instead of Bill, I would have written it the way he did (had I the skill), but would have had a little box at the end of the chapter listing the different caveats, which he clearly knew about, judging from long interaction with him in person and over CSGnet.

Martin

[David Goldstein (2013.08.07.10:46)]
[Martin Taylor 2013.08.06.17.07]

Interesting points.

I agree with your summarizing statement: “I think that if I had been writing B:CP instead of Bill, I would have written it the way he did (had I the skill), but would have had a little
box at the end of the chapter listing the different caveats, which he clearly knew about, judging from long interaction with him in person and
over CSGnet.”

David

···

From: Martin Taylor mmt-csg@MMTAYLOR.NET
To: CSGNET@LISTSERV.ILLINOIS.EDU
Sent: Tuesday, August 6, 2013 6:21 PM
Subject: Re: B:CP Chapter 4

[Martin Taylor 2013.08.06.17.07]

I guess this is outside the flow of the seminar, but here goes, on the basis that any comments were welcome.

For me, Chapter 4 is a fine introduction to the reasons why we have to deal only with perceptual control, but just as earlier we have to note some caveats about linearity being valid only for small signal excursions and the approximation of neural firing patterns by “neural currents”, so in Chapter 4 there are a few caveats to keep in mind. But unless I skipped them somehow, they are caveats Bill failed to mention. Mainly, they are about places where Bill makes absolutist claims such as “never” and “always” and “only”, where once one understands PCT these have to be softened into “very seldom”, “usually”, and “with few exceptions”.

For example, the first para on P47 starts: “When feedback is found in the relationship between an organism and its environment, it will either be so weak that it can
be ignored, or be negative–it will never be strong and positive.” Yet Bill and Rick did at least one study in which they deliberately switched the feedback to be strong and positive – not for very long, because the subject quickly learned to switch his tracking mode, but strong positive feedback did exist.

More practically, isn’t a temper tantrum the result of an uncorrected positive feedback process? And as I discussed with Bill at CSG93, the “Bomb in the Hierarchy” <http://www.mmtaylor.net/PCT/DFS93/DFS93_8.html>, induces reorganization by ensuring that episodes of positive feedback do occur on occasion. Using the analogy that is often used to refer to evolutionary processes, which is equally applicable to e-coli reorganization, the effect of the Bomb is like a sandpile avalanche, usually unnoticed, but occasionally dramatic.

In the same paragraph Bill
says “one will always be able to discover what the subject is controlling, for if disturbances are applied that do not in fact disturb the controlled aspect of the environment, the subject’s behaviour will not oppose the disturbance.” “Always”, here is a gross overstatement for two reasons. Firstly, if disturbances are applied that disturb a correlate of the controlled perception, the subject will oppose them, and the only way that one could discover exactly what the subject is controlling is to try all possible variants and see what the subject most precisely opposes. Secondly, if the subject is not controlling well, there is no way to distinguish among the various correlates because for none of them will the opposition to the disturbance be very precise. Thirdly, except in a constrained laboratory condition, what the subject is controlling may change at any moment, as may its reference value if the controlled perception remains controlled. So although
“always” may be appropriate in the introductory chapters, Bill should have done as he did earlier, and mention the caveats.

In the middle of the following paragraph he says “all behaviour all the time”, when frequently on CSGnet, this is amended to “all intentional behaviour”. From outside, one cannot distinguish side-effects from intended effects, nervous tics from an actor’s presentation, and so forth. This is a caveat that depends on the meaning of “behaviour”. The definition is circular if it depends on the observed actions being effective in influencing a controlled perception, but the statement in the paragraph is overstated if it is not. It’s a good teaching statement that should have the desired effect of bringing up short someone not accustomed to thinking this way, but it deserves the same kind of “keep it in mind for later, when we get into detail” caveat as the others.

The first paragraph of page 42 is paraphrased in the
second paragraph of the summary, where there’s another similar caveat: “…sensory inputs affect and are affected by behaviour. This circle of cause and effect cannot be correctly analyzed as an alternating sequence”. But if the lag between action and influence on perception is long, it must be analyzed as an alternating sequence. Think of a mundane example with which I am confronted every day. My shower has a long delay between setting the temperature control and a change in the water temperature, which overshoots before it settles. I turn the handle, then wait to see whether the stable temperature is too hot or too cold, and then turn the handle again in the appropriate direction. In the abstract analysis, the mathematics is difficult precisely because it must be done piecewise, factoring in the “echoes” of previous phases of action after every lag cycle. Again, this is a caveat “to be kept in mind for later”, rather than a problem with the
exposition.

I think that if I had been writing B:CP instead of Bill, I would have written it the way he did (had I the skill), but would have had a little box at the end of the chapter listing the different caveats, which he clearly knew about, judging from long interaction with him in person and over CSGnet.

Martin

[From Rick Marken (2013.08.07.1215)]

Martin Taylor (2013.08.06.17.07)

MT: ... in Chapter 4 there are a few caveats to keep in mind. But unless I skipped
them somehow, they are caveats Bill failed to mention. Mainly, they are about
places where Bill makes absolutist claims such as "never" and "always" and
"only", where once one understands PCT these have to be softened into "very
seldom", "usually", and "with few exceptions".

For example, the first para on P47 starts: "When feedback is found in the
relationship between an organism and its environment, it will either be so
weak that it can be ignored, or be negative--it will never be strong and
positive." Yet Bill and Rick did at least one study in which they
deliberately switched the feedback to be strong and positive -- not for very
long, because the subject quickly learned to switch his tracking mode, but
strong positive feedback did exist.

RM: Yes, I think this is a fair criticism, although Bill does say the
feedback will never be "strong and positive" implying that it could
sometimes be weak and positive, which is the case even in our little
study where we reversed the polarity of the effect of output on input.
That made the feedback positive but, as you say, the feedback was not
positive for long, so, integrating over time, the gain of this
positive feedback situation is pretty low. But it might have been
better for Bill to have simply said that feedback will "almost never
be strong and positive".

Since your comments are part of the B:CP course, I think a good
question for the "students" would be "Can you think of any examples of
human (or "lower" animal) behavior that seem to involve positive
rather than negative feedback"? You did off a temper tantrum as an
example of behavior involving positive feedback. Can you think of any
others? And also could you include in you answer an explanation of why
you think the behavior seems to involve positive feedback, where
positive feedback exists when efforts to correct error actually
increase it.

MT: In the same paragraph Bill says "one will always be able to discover what
the subject is controlling, for if disturbances are applied that do not in
fact disturb the controlled aspect of the environment, the subject's
behaviour will not oppose the disturbance." "Always", here is a gross
overstatement for two reasons. Firstly, if disturbances are applied that
disturb a correlate of the controlled perception, the subject will oppose
them, and the only way that one could discover exactly what the subject is
controlling is to try all possible variants and see what the subject most
precisely opposes. Secondly, if the subject is not controlling well, there
is no way to distinguish among the various correlates because for none of
them will the opposition to the disturbance be very precise. Thirdly, except
in a constrained laboratory condition, what the subject is controlling may
change at any moment, as may its reference value if the controlled
perception remains controlled. So although "always" may be appropriate in
the introductory chapters, Bill should have done as he did earlier, and
mention the caveats.

RM: I think these are also good points but I think the place to put
these "caveats" is in a paper dealing with the methodological details
of the test for the controlled variable -- a paper that hasn't been
written yet but should be.

MT: In the middle of the following paragraph he says "all behaviour all the
time", when frequently on CSGnet, this is amended to "all intentional
behaviour". From outside, one cannot distinguish side-effects from intended
effects, nervous tics from an actor's presentation, and so forth.

RM: This is also a good point. But I think it gets at a nuance that is
better discussed once the control models has been introduced. In this
chapter (Ch 4) Powers is arguing for the _fact_ of control; that
behavior is a control process; that it is purposeful. He's is
presenting this argument to psychologists who basically reject purpose
(or see it as an illusion). The fact that some things that we call
behavior -- like the behavior of tripping on a crack in the cement --
not done on purpose is a point that should be made; but I think it is
better made once we have a model of purposeful behavior.

MT: The first paragraph of page 42 is paraphrased in the second paragraph of
the summary, where there's another similar caveat: "...sensory inputs affect
and are affected by behaviour. This circle of cause and effect cannot be
correctly analyzed as an alternating sequence". But if the lag between
action and influence on perception is long, it _must_ be analyzed as an
alternating sequence.

RM: Here I think you are wrong. While it's true that there is a lag
between perception and action, perception and action are both
occurring continuously and at the same time. So while the present
perception is influencing future action it is being influence by
present action. A sequential analysis detaches action from perception,
turning a loop involving variables that vary continuously over time
into a sequences of discrete temporal events. The continuous analysis
takes into account the fact that action is affecting perception
_while_ perception is affecting action; the sequential analysis hides
this simultaneity.

So while it's true that a proper analysis of a control loop must
include the delayed influence of perception on action (and also of
action on perception) -- this delay being a result of transport lag
(the time for the effect of the perception to be transmitted via the
nervous system to the action system) and temporal integration (the
time for the effect of the influence of the perception of "build up"
to it's full effective value) a proper analysis of the behavior of a
control loop requires that all variations in the loop be treated as
though they were occurring simultaneously, not in sequence.

MT: Think of a mundane example with which I am confronted
every day. My shower has a long delay between setting the temperature
control and a change in the water temperature, which overshoots before it
settles. I turn the handle, then wait to see whether the stable temperature
is too hot or too cold,

RM: But you are acting while you are testing (perceiving) the
temperature to see if it's where you want it; your action is
"waiting"; you are always acting while you are perceiving. The point
of Ch. 4 is that we are _locked_ in a feedback loop (which the
stability of behavior suggests is negative); what we sense is at all
times a result of what we are doing (where "doing" includes not
moving). This is simply a result of the fact that our senses are
housed on the body whose actions they influence; so what we do (or
don't do) as a result of what we sense (perceive) is continuously
influencing what we sense. There is no sequence because there are no
gaps in perception; we don't stop perceiving while the perception that
influences a future action has it's effect. We are continuously
perceiving (and acting) and those perception are always simultaneously
influencing and being influenced by actions. Thus, we are always
controlling, which is the argument of Ch. 4.

Best regards

Rick

···

--
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

[Martin Taylor 2013.08.07.17.47]

[From Rick Marken (2013.08.07.1215)]

Martin Taylor (2013.08.06.17.07)
MT: ... in Chapter 4 there are a few caveats to keep in mind....

For example, the first para on P47 starts: "When feedback is found in the
relationship between an organism and its environment, it will either be so
weak that it can be ignored, or be negative--it will never be strong and
positive." Yet Bill and Rick did at least one study in which they
deliberately switched the feedback to be strong and positive -- not for very
long, because the subject quickly learned to switch his tracking mode, but
strong positive feedback did exist.

RM: ...

Since your comments are part of the B:CP course,

They were intended more as comments on the chapter, since they were not responsive to your questions.

MT: In the same paragraph Bill says "one will always be able to discover what
the subject is controlling, for if disturbances are applied that do not in
fact disturb the controlled aspect of the environment, the subject's
behaviour will not oppose the disturbance." "Always", here is a gross
overstatement.... So although "always" may be appropriate in
the introductory chapters, Bill should have done as he did earlier, and
mention the caveats.

RM: I think these are also good points but I think the place to put
these "caveats" is in a paper dealing with the methodological details
of the test for the controlled variable -- a paper that hasn't been
written yet but should be.

Perhaps so, but the point of all my caveats was simply to say that although it was probably a good rhetorical choice for Bill to be absolutist, nevertheless, after people have read B:CP and taken it all in, they might go back to the chapter and say "Why did he write 'always' here, when 'almost always' would have been correct?" A small box at the end of the chapter would have avoided that, and reminded readers, as he did in Chapter 3, that he is offering a sketch draft of a very complicated structure. When you make the first sketch for a painting, you don't enter all the details, and nor could or should someone writing an introduction to a conceptual structure that is expected to be new to the reader. But readers should be warned that there are many details that will contradict the absolute nature of the statements.

MT: In the middle of the following paragraph he says "all behaviour all the
time", when frequently on CSGnet, this is amended to "all intentional
behaviour". From outside, one cannot distinguish side-effects from intended
effects, nervous tics from an actor's presentation, and so forth.

RM: This is also a good point. But I think it gets at a nuance that is
better discussed once the control models has been introduced.

This is true of all my caveats. But it wouldn't hurt to mention it when you are talking about "purpose".

MT: The first paragraph of page 42 is paraphrased in the second paragraph of
the summary, where there's another similar caveat: "...sensory inputs affect
and are affected by behaviour. This circle of cause and effect cannot be
correctly analyzed as an alternating sequence". But if the lag between
action and influence on perception is long, it _must_ be analyzed as an
alternating sequence.

RM: Here I think you are wrong. While it's true that there is a lag
between perception and action, perception and action are both
occurring continuously and at the same time.

I said nothing about a lag between perception and action. I was talking about a lag between action and perception. Though loop transport lag is really the technical issue, in most cases you can act on a perception much more quickly than your action can influence your perception (not in the standard computer-based tracking studies, but certainly if you are thinking about the effects of your biennial vote on political decisions).

... A sequential analysis detaches action from perception,
turning a loop involving variables that vary continuously over time
into a sequences of discrete temporal events. The continuous analysis
takes into account the fact that action is affecting perception
_while_ perception is affecting action; the sequential analysis hides
this simultaneity.

Yes, and if you do the continuous analysis, you find that if you act too fast, you get into a positive feedback oscillation. It was an early exercise in my engineering education to look at how the intrinsic dynamics of a loop can be separated from the transient effects of external influences. In this case, we have a loop with a lag and a gain that is negative at zero frequency. It isn't negative at a frequency where the output gain phase shift coupled with the time-lag equivalent phase shift is 180 degrees. To avoid oscillation, the absolute magnitude of the loop gain must be less than 1.0 at that frequency If it isn't, you get runaway oscillation. If it's high but below 1.0, you get ringing, like a bell. Since your perception-action connection is part of the loop, you can introduce the necessary filtering by moving no more often than once per lag time, or else moving very slowly. The latter means rather ineffective control.

In any case, we were talking about the analysis, not the way the controller maintains stability in the face of long action-perception lags. The analysis must take into account the lag, and the fact that a movement at t0 has an effect that is added to the current perception after one transport lag, echoes again after two, and so forth. Whatever the analytical technique, it must at base treat the loop behaviour as though the control actions were performed one at a time. In effect, the analysis has to use a time-comb filter with the "teeth" of the comb at one lag time. For distributed lag times it gets a bit more complicated.

... a proper analysis of the behavior of a
control loop requires that all variations in the loop be treated as
though they were occurring simultaneously, not in sequence.

Yes. There's no contradiction here.

...There is no sequence because there are no gaps in perception; we don't stop perceiving while the perception that influences a future action has it's effect. We are continuously perceiving (and acting) and those perception are always simultaneously influencing and being influenced by actions. Thus, we are always controlling, which is the argument of Ch. 4. Best regards Rick

We are indeed always perceiving and always controlling. But we don't perceive everything that's going on in the loop (though the analyst may). What we don't perceive before time t0+lag is the result of our action at t0. What we do perceive at time t0 is the ongoing disturbance plus the effects of our actions prior to t0-lag. What will lead us into a positive feedback loop is an attempt to control disturbances so fast that the phase shifts get to 180 degrees while the absolute magnitude of the loop gain in still greater than 1.0. (That's another answer to your question about when positive feedback happens in the real world, though people usually find out quickly enough that they shouldn't change action magnitude until they find out how the last step affected the controlled perception).

All that is too technical for the course, certainly at this stage of the development of the concepts, though it may be appropriate later. I can make it more mathematical if you want, but I think a caveat about being wary about discrete control actions when you have long transport lags is sufficient for the course.

Martin

[From Rick Marken (2013.08.08.1020)]

Martin Taylor (2013.08.07.17.47) -

RM: Here I think you are wrong. While it's true that there is a lag
between perception and action, perception and action are both
occurring continuously and at the same time.

... A sequential analysis detaches action from perception,

MT: In any case, we were talking about the analysis...The analysis
must take into account the lag, and the fact that a movement at t0 has an
effect that is added to the current perception after one transport lag,
echoes again after two, and so forth...

RM: Yes, the analysis must include time lags but a sequential
analysis, which assumes that input is _followed_ by output which is
_followed_ by input in discrete, non-overlapping steps, in an
alternating sequence, is wrong. That is why Powers says, in Ch. 4,
that a control loop "...cannot be correctly analyzed as an alternating
sequence". This is a very important point; Powers makes this point
specifically because it was sequential analyses of closed loop
behavior (such as the TOTE unit found in Miller, Galanter and
Pribrum's influential "Plans and Structure of Behavior", Neisser's
circle of cognition described in"Cognition and Reality" and sequential
descriptions of feedback loops found in many books on cognitive
psychology) that has allowed psychologists to maintain the assumption
that behavior can be explained in terms of a causal (input-output or
open-loop) model. The statement that a control loop "...cannot be
correctly analyzed as an alternating sequence" was not tossed off
accidentally; it is one of the most important things to know about
closed loop behavior and, I believe, it is the very essence of the
point made in Ch. 4.

When you can visualize behavior as a process where actions are
affecting sensory input _while_ sensory input is affecting actions
(rather than as a sequential process where action affects sensory
input _and then_ sensory input affects action _and then_ action
affects sensory input) then you really do start seeing behavior in a
new way -- through control theory glasses. It's not easy to do,
especially when brief transient disturbances (like a hammer tap on the
patellar tendon) appear to elicit actions (knee jerk). But you have to
always keep in mind that the angle of the knee is continuously
affecting sensed tendon tension _while_ sensed tendon tension is
continuously affecting knee angle. There is no sequence.

Best regards

Rick

···

--
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

[Martin Taylor 2013.08.08.14.03]

[From Rick Marken (2013.08.08.1020)]

While one can never be sure, I believe I understand everything you say. However, _if_ I understand what you say, it does not address my point. Certainly, if your second paragraph, which I will not quote, was addressed to me, you _really and deeply_ did not understand where I was coming from.

I understand why Powers had to correct the TOTE mistake, but as with my other caveats, the running text seems to overstate the correction. It's the same issue as feedback _always_ being negative, and so forth.

RM:

The statement that a control loop "...cannot be
correctly analyzed as an alternating sequence" was not tossed off
accidentally; it is one of the most important things to know about
closed loop behavior and, I believe, it is the very essence of the
point made in Ch. 4.

MT: And my point is that this is a good teaching point to make. It corrects against the approaches you cite, but to state it as an absolute is overkill and deserves a caveat. There are times when one does something, waits to see the effect, and then does whatever it takes to correct the result. Moreover, one does this because it is efficient and effective, when to continuously modify one's action could lead to positive feedback and exponentially increasing oscillation. As I said, to analyze correctly the behaviour of a control loop that includes lag requires one to do the mathematical equivalent of adding an infinite number of overlapping wait-and-see cycles. You usually don't do that in practice, for two reasons: (1) you are usually analyzing an experiment in which lag is minimized and the other parameters such as slow disturbance and relatively low gain are such as to avoid oscillation, and (2) you usually ignore the dynamics and analyze only the (never reached) steady state solution.

I haven't done this, but it wouldn't be hard to do. Do a trivial compensatory or pursuit tracking study in which the effect of moving the mouse is delayed by a few tens and then hundreds of milliseconds, and see whether you get to a lag duration at which the subject starts to move the mouse stepwise. Use a disturbance waveform consisting of a single step change in location. You can't use yourself or someone who knows what you are looking for as a subject in this one.

Martin

···

[From Rick Marken (2013.08.08.1020)]

Martin Taylor (2013.08.07.17.47) -

RM: Here I think you are wrong. While it's true that there is a lag
between perception and action, perception and action are both
occurring continuously and at the same time.
... A sequential analysis detaches action from perception,

MT: In any case, we were talking about the analysis...The analysis
must take into account the lag, and the fact that a movement at t0 has an
effect that is added to the current perception after one transport lag,
echoes again after two, and so forth...

RM: Yes, the analysis must include time lags but a sequential
analysis, which assumes that input is _followed_ by output which is
_followed_ by input in discrete, non-overlapping steps, in an
alternating sequence, is wrong. That is why Powers says, in Ch. 4,
that a control loop "...cannot be correctly analyzed as an alternating
sequence". This is a very important point; Powers makes this point
specifically because it was sequential analyses of closed loop
behavior (such as the TOTE unit found in Miller, Galanter and
Pribrum's influential "Plans and Structure of Behavior", Neisser's
circle of cognition described in"Cognition and Reality" and sequential
descriptions of feedback loops found in many books on cognitive
psychology) that has allowed psychologists to maintain the assumption
that behavior can be explained in terms of a causal (input-output or
open-loop) model. The statement that a control loop "...cannot be
correctly analyzed as an alternating sequence" was not tossed off
accidentally; it is one of the most important things to know about
closed loop behavior and, I believe, it is the very essence of the
point made in Ch. 4.

When you can visualize behavior as a process where actions are
affecting sensory input _while_ sensory input is affecting actions
(rather than as a sequential process where action affects sensory
input _and then_ sensory input affects action _and then_ action
affects sensory input) then you really do start seeing behavior in a
new way -- through control theory glasses. It's not easy to do,
especially when brief transient disturbances (like a hammer tap on the
patellar tendon) appear to elicit actions (knee jerk). But you have to
always keep in mind that the angle of the knee is continuously
affecting sensed tendon tension _while_ sensed tendon tension is
continuously affecting knee angle. There is no sequence.

Best regards

Rick

[From Rick Marken (2013.08.08.1400)]

Martin Taylor (2013.08.08.14.03)--

RM: The statement that a control loop "...cannot be
correctly analyzed as an alternating sequence" was not tossed off
accidentally; it is one of the most important things to know about
closed loop behavior and, I believe, it is the very essence of the
point made in Ch. 4.

MT: And my point is that this is a good teaching point to make. It corrects
against the approaches you cite, but to state it as an absolute is overkill
and deserves a caveat.

RM: Perhaps, but it seems pretty absolute to me. I say this based not
only on the points Bill makes in Ch. 4 -- that what we are sensing
influences what we are doing and at the same time what we are doing
influences what we are sensing -- but also on my experience with
modeling; all my control models involve simulating on a sequential
state machine ( digital computer) the simultaneous equations that
define a control loop.

MT: There are times when one does something, waits to see
the effect, and then does whatever it takes to correct the result.

RM: I think "doing" is controlling. So when one "does something" like
turn on the faucet for a shower, they are controlling for water coming
out of the shower head. So there is a perception of what's coming out
of the shower head while the action that influences that perception
(turning the faucet) is occurring. Now the faucet was probably turned
as part a program (perception) that might be called "getting the water
ready for my shower". That program perception is occurring while the
actions that control the lower level perceptions -- turning on faucet,
vary faucet settings while testing water to bring water to right
temperature -- that keep that program perception under control.

There are elements of the "prepare for shower" program that can look
like sequential steps in a control loop (the mistake made when MG&P
proposed the TOTE loop). For example, the faucet is turned (action)
then after some time the water is tested (perception) and if the water
is still not right there is another faucet turn (action) and later
another perception. So it looks like water temperature is being
controlled in a sequence of steps. But I would argue that this
sequence is not a control process; at least not a process of water
temperature control. When you want to control water temperature you
hold your hand under water _while_ adjusting the faucet. So what's
going on with the sequence "adjust", "perceive", "test", "adjust",
"perceive", "test"? I would say that the person is controlling for the
perception that is simultaneous with the actions that control it --
the perception of the relationship between turning the faucet and the
subsequent increase or decrease in water temperature. That perception
is present while you are taking action (turning and then not turning
the faucet) to keep it under control.

My belief is that, while you can control a perception of a sequence of
actions, a sequence of actions cannot control a perception.

I know I'll never convince you of this, Martin, so we will have to
just agree to disagree, I guess.

But I will change my belief if you can show me a sequential system
that controls.

Best

Rick

Moreover,

···

one does this because it is efficient and effective, when to continuously
modify one's action could lead to positive feedback and exponentially
increasing oscillation. As I said, to analyze correctly the behaviour of a
control loop that includes lag requires one to do the mathematical
equivalent of adding an infinite number of overlapping wait-and-see cycles.
You usually don't do that in practice, for two reasons: (1) you are usually
analyzing an experiment in which lag is minimized and the other parameters
such as slow disturbance and relatively low gain are such as to avoid
oscillation, and (2) you usually ignore the dynamics and analyze only the
(never reached) steady state solution.

I haven't done this, but it wouldn't be hard to do. Do a trivial
compensatory or pursuit tracking study in which the effect of moving the
mouse is delayed by a few tens and then hundreds of milliseconds, and see
whether you get to a lag duration at which the subject starts to move the
mouse stepwise. Use a disturbance waveform consisting of a single step
change in location. You can't use yourself or someone who knows what you are
looking for as a subject in this one.

Martin

[From Rick Marken (2013.08.08.1020)]

Martin Taylor (2013.08.07.17.47) -

RM: Here I think you are wrong. While it's true that there is a lag
between perception and action, perception and action are both
occurring continuously and at the same time.
... A sequential analysis detaches action from perception,

MT: In any case, we were talking about the analysis...The analysis
must take into account the lag, and the fact that a movement at t0 has an
effect that is added to the current perception after one transport lag,
echoes again after two, and so forth...

RM: Yes, the analysis must include time lags but a sequential
analysis, which assumes that input is _followed_ by output which is
_followed_ by input in discrete, non-overlapping steps, in an
alternating sequence, is wrong. That is why Powers says, in Ch. 4,
that a control loop "...cannot be correctly analyzed as an alternating
sequence". This is a very important point; Powers makes this point
specifically because it was sequential analyses of closed loop
behavior (such as the TOTE unit found in Miller, Galanter and
Pribrum's influential "Plans and Structure of Behavior", Neisser's
circle of cognition described in"Cognition and Reality" and sequential
descriptions of feedback loops found in many books on cognitive
psychology) that has allowed psychologists to maintain the assumption
that behavior can be explained in terms of a causal (input-output or
open-loop) model. The statement that a control loop "...cannot be
correctly analyzed as an alternating sequence" was not tossed off
accidentally; it is one of the most important things to know about
closed loop behavior and, I believe, it is the very essence of the
point made in Ch. 4.

When you can visualize behavior as a process where actions are
affecting sensory input _while_ sensory input is affecting actions
(rather than as a sequential process where action affects sensory
input _and then_ sensory input affects action _and then_ action
affects sensory input) then you really do start seeing behavior in a
new way -- through control theory glasses. It's not easy to do,
especially when brief transient disturbances (like a hammer tap on the
patellar tendon) appear to elicit actions (knee jerk). But you have to
always keep in mind that the angle of the knee is continuously
affecting sensed tendon tension _while_ sensed tendon tension is
continuously affecting knee angle. There is no sequence.

Best regards

Rick

--
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

[Martin Taylor 2013.08.08.23.02]

[From Rick Marken (2013.08.08.1400)]

Martin Taylor (2013.08.08.14.03)--

RM: The statement that a control loop "...cannot be
correctly analyzed as an alternating sequence" was not tossed off
accidentally; it is one of the most important things to know about
closed loop behavior and, I believe, it is the very essence of the
point made in Ch. 4.

MT: And my point is that this is a good teaching point to make. It corrects
against the approaches you cite, but to state it as an absolute is overkill
and deserves a caveat.

RM: Perhaps, but it seems pretty absolute to me. I say this based not
only on the points Bill makes in Ch. 4 -- that what we are sensing
influences what we are doing and at the same time what we are doing
influences what we are sensing -- but also on my experience with
modeling; all my control models involve simulating on a sequential
state machine ( digital computer) the simultaneous equations that
define a control loop.

As they should. But it is evident that either you haven't done much simulation of experiments involving loops with long lags compared to the rate at which the action component of the loop can act, or you haven't appreciated the implications of your simulations.

...

I know I'll never convince you of this, Martin, so we will have to
just agree to disagree, I guess.

But I will change my belief if you can show me a sequential system
that controls.

It's very strange that while I pretty much agree with everything you say about control, you finish by saying _we_ have to agree to disagree. I do not agree to disagree; it's _you_ who agrees with yourself to disagree. It's your choice to ignore what I say, and substitute your own statements, with which you are quite free to disagree. Play your own solitaire game all you want. I agree with what you say about control, none of which contradicts what I said in my two previous messages, the ones you choose to disagree with.

Martin

[From Rick Marken (2013.08.09.0940)]

Martin Taylor (2013.08.08.23.02)--

MT: It's very strange that while I pretty much agree with everything you say
about control, you finish by saying _we_ have to agree to disagree. I do not
agree to disagree; it's _you_ who agrees with yourself to disagree.

RM: Perhaps. What I _think_ I disagree with you about is the idea that
control can be represented as a _sequential_ process. I know that
there can be long lags between input and output (neural transport lag)
and, especially, between output and input created by physical
processes that intervene between what we do and what we perceive (I've
done a shuttle docking simulation where the lag between my action --
accelerating or decelerating the shuttle--and result--gently docking
with the target object--were on the order of a minute). But these lags
are occurring in a non-sequential control process where inputs are
occurring _while_ outputs are occurring; the input seen at time t(0)
may be largely the result of outputs that occurred at in the past
(time t(-tau)), but the input at t(0) is still occurring at the same
time as the output at t(0); it's in that sense that the control
process is non-sequential, from my point of view. But these points are
relevant to the basic control model discussed in Ch. 5. So maybe we
can discuss this in that context, particularly with respect to control
system dynamics.

Best regards

Rick

···

--
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

[David Goldstein (2013.08.09.17:45)]

Re.: [From Rick Marken (2013.08.09.0940)]

Re.: Martin Taylor (2013.08.08.23.02)–

I think that a lot of the difference in your views may lie in the level of hierarchy

which is being discussed. The higher levels take longer to

control their perceptual signals. But doesn’t a control

system at each level control in the manner of the Live Block

diagram? The difference is in the nature of the perceptual signal.

Take the example of a person controlling
the temperature of the water.

Martin describes it as a sequence. When I get into the shower, I set the

control knob all the way to hot, but don’t let the water into the shower head

right away. I put my foot under the water from the faucet and test the water

temperature. It is usually too hot, so I adjust the control
knob. I keep on doing this

until the water temperature is just right. Then I allow the water to come through the

shower head. When the water lands on my body, I may make a final adjustment, if

necessary. Doesn’t this sound more like a program level? If there were a change in the water

temperature by some unknown
forces, I would make an adjustment. This is a control

system at a program level. The program is part of a “Taking a shower routine.”

Something which seems more like a sequence is normal walking–left, right; left right.

If I see someone go left, left; right right; left, left, right right, I notice this as an abnormal
walking pattern.

The perception and action of the normal walking works in the same way as the Live Block diagram, doesn’t it?

David

···

From: Richard Marken
rsmarken@GMAIL.COM
To: CSGNET@LISTSERV.ILLINOIS.EDU
Sent: Friday, August 9, 2013 12:38 PM
Subject: Re: B:CP Chapter 4

[From Rick Marken (2013.08.09.0940)]

Martin Taylor (2013.08.08.23.02)–

MT: It’s very strange that while I pretty much agree with everything you say
about control, you finish by saying we have to agree to disagree. I do not
agree to disagree; it’s you who agrees with yourself to disagree.

RM: Perhaps. What I think I disagree with you about is the idea that
control can be represented as a sequential process. I know that
there can be long lags between input and output (neural transport lag)
and, especially, between output and input created by physical
processes that intervene between what we do and what we perceive (I’ve
done a shuttle docking simulation where the lag between my action –
accelerating or decelerating the shuttle–and result–gently docking
with the target object–were on the order of a minute). But these lags
are occurring in a non-sequential control
process where inputs are
occurring while outputs are occurring; the input seen at time t(0)
may be largely the result of outputs that occurred at in the past
(time t(-tau)), but the input at t(0) is still occurring at the same
time as the output at t(0); it’s in that sense that the control
process is non-sequential, from my point of view. But these points are
relevant to the basic control model discussed in Ch. 5. So maybe we
can discuss this in that context, particularly with respect to control
system dynamics.

Best regards

Rick


Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

[Martin Taylor 2013.08.09.23.15]

(To David Goldstein: I am talking about a standard pursuit or complementary tracking task in a situation in which there is a substantial time delay between acting on the environment and perceiving the effect of that action on the perception the action is intended to influence.)

[From Rick Marken (2013.08.09.0940)]

Martin Taylor (2013.08.08.23.02)--
MT: It's very strange that while I pretty much agree with everything you say
about control, you finish by saying _we_ have to agree to disagree. I do not
agree to disagree; it's _you_ who agrees with yourself to disagree.

RM: Perhaps. What I _think_ I disagree with you about is the idea that
control can be represented as a _sequential_ process.

I haven't suggested that it might be, though since control can be represented in a myriad of different ways, I suppose one might be able to find a way it could be done.

What Bill was contrasting with smooth control was an analysis in which "a stimulus causes a response via the organism, then the response causes a new stimulus or modifies the next stimulus, and the cycle begins again....The real organism behaves in a smoothly continuous fashion." It is this last statement that requires the caveat, not the idea that perception, action, and all the signals in the loop are continuously and simultaneously varying. Let's take a trivial counter-example. A person throws a dart at a dartboard. Later he throws another one, having observed the fall of the first. Does this "real organism" continuously throw darts, correcting the throw of each while the continuous stream of darts is on its way to the board? It does not. It may well observe the _individual_ dart in flight on its way to the board, and it may well be modifying its aim during that time, but it does not throw the next dart until after the first one has landed. At whatever control level the dart throwing is analyzed, this "real organism" does not "behave in a smoothly continuous fashion" at that level.

  I know that
there can be long lags between input and output (neural transport lag)
and, especially, between output and input created by physical
processes that intervene between what we do and what we perceive (I've
done a shuttle docking simulation where the lag between my action --
accelerating or decelerating the shuttle--and result--gently docking
with the target object--were on the order of a minute). But these lags
are occurring in a non-sequential control process where inputs are
occurring _while_ outputs are occurring; the input seen at time t(0)
may be largely the result of outputs that occurred at in the past
(time t(-tau)), but the input at t(0) is still occurring at the same
time as the output at t(0); it's in that sense that the control
process is non-sequential, from my point of view. But these points are
relevant to the basic control model discussed in Ch. 5. So maybe we
can discuss this in that context, particularly with respect to control
system dynamics.

I can agree with all of that. Interestingly, in space there isn't much opportunity for random disturbances to influence the perception. Space physical dynamics are pretty predictable. That kind of predictability is used in some submarine and ship control, where the helmsman is shown where the vessel will go if the control surfaces are held where they are. That matters, because it is often a long time before the effect of changing the control surfaces is visible to the helmsman's eye, and there are disturbances whose effect can result in predictable course changes down the road. Control in a complicated situation such as entering Rotterdam harbour (one case I saw demonstrated) is greatly improved by the real-time prediction, because the helmsman can see how the new settings will affect the course with very little lag.

The experimental question is under what conditions of lag does a compensatory or pursuit tracker act like the dart thrower, and under what conditions does the tracker act smoothly, as we see in the usual tracking experiments. In both cases, perception and action are occurring simultaneously. However, when lag becomes a substantial phase shift at frequencies that occur in the disturbance, the analysis has to take into account that the effects of any action echo discretely around the loop as a kind of cycle similar to the kind Bill was talking about. That does NOT make it a sequential analysis of the kind Bill was objecting to.

Martin

[John Kirkland 20130813]

Possible candidates for positive feedback where control goes haywire:

Infections, e.g. Tetanus

Muscle cramp

A Tazer strike

Comments welcomed

JohnK

···

On Sat, Aug 10, 2013 at 3:48 PM, Martin Taylor mmt-csg@mmtaylor.net wrote:

[Martin Taylor 2013.08.09.23.15]

(To David Goldstein: I am talking about a standard pursuit or complementary tracking task in a situation in which there is a substantial time delay between acting on the environment and perceiving the effect of that action on the perception the action is intended to influence.)

[From Rick Marken (2013.08.09.0940)]

Martin Taylor (2013.08.08.23.02)–

MT: It’s very strange that while I pretty much agree with everything you say

about control, you finish by saying we have to agree to disagree. I do not

agree to disagree; it’s you who agrees with yourself to disagree.
RM: Perhaps. What I think I disagree with you about is the idea that

control can be represented as a sequential process.

I haven’t suggested that it might be, though since control can be represented in a myriad of different ways, I suppose one might be able to find a way it could be done.

What Bill was contrasting with smooth control was an analysis in which “a stimulus causes a response via the organism, then the response causes a new stimulus or modifies the next stimulus, and the cycle begins again…The real organism behaves in a smoothly continuous fashion.” It is this last statement that requires the caveat, not the idea that perception, action, and all the signals in the loop are continuously and simultaneously varying. Let’s take a trivial counter-example. A person throws a dart at a dartboard. Later he throws another one, having observed the fall of the first. Does this “real organism” continuously throw darts, correcting the throw of each while the continuous stream of darts is on its way to the board? It does not. It may well observe the individual dart in flight on its way to the board, and it may well be modifying its aim during that time, but it does not throw the next dart until after the first one has landed. At whatever control level the dart throwing is analyzed, this “real organism” does not “behave in a smoothly continuous fashion” at that level.

I know that

there can be long lags between input and output (neural transport lag)

and, especially, between output and input created by physical

processes that intervene between what we do and what we perceive (I’ve

done a shuttle docking simulation where the lag between my action –

accelerating or decelerating the shuttle–and result–gently docking

with the target object–were on the order of a minute). But these lags

are occurring in a non-sequential control process where inputs are

occurring while outputs are occurring; the input seen at time t(0)

may be largely the result of outputs that occurred at in the past

(time t(-tau)), but the input at t(0) is still occurring at the same

time as the output at t(0); it’s in that sense that the control

process is non-sequential, from my point of view. But these points are

relevant to the basic control model discussed in Ch. 5. So maybe we

can discuss this in that context, particularly with respect to control

system dynamics.

I can agree with all of that. Interestingly, in space there isn’t much opportunity for random disturbances to influence the perception. Space physical dynamics are pretty predictable. That kind of predictability is used in some submarine and ship control, where the helmsman is shown where the vessel will go if the control surfaces are held where they are. That matters, because it is often a long time before the effect of changing the control surfaces is visible to the helmsman’s eye, and there are disturbances whose effect can result in predictable course changes down the road. Control in a complicated situation such as entering Rotterdam harbour (one case I saw demonstrated) is greatly improved by the real-time prediction, because the helmsman can see how the new settings will affect the course with very little lag.

The experimental question is under what conditions of lag does a compensatory or pursuit tracker act like the dart thrower, and under what conditions does the tracker act smoothly, as we see in the usual tracking experiments. In both cases, perception and action are occurring simultaneously. However, when lag becomes a substantial phase shift at frequencies that occur in the disturbance, the analysis has to take into account that the effects of any action echo discretely around the loop as a kind of cycle similar to the kind Bill was talking about. That does NOT make it a sequential analysis of the kind Bill was objecting to.

Martin

[Martin Taylor 2013.08.12.17.48]

[John Kirkland 20130813]

Possible candidates for positive feedback where control goes haywire:
Infections, e.g. Tetanus
Muscle cramp
A Tazer strike

Comments welcomed

JohnK

Are the actions in such cases intentional -- in other words, do they occur because some controlled perception differs from its reference value?

Martin

[From Rick Marken (2013.08.12.1530)]

[John Kirkland 20130813]

Possible candidates for positive feedback where control goes haywire:
Infections, e.g. Tetanus
Muscle cramp
A Tazer strike

Comments welcomed

Evidence of positive feedback is when action that is aimed at reducing
error seems to increase it, as evidenced by an increase in the action.
So an infection does not seems to be a positive feedback situation
since the action (lymphocyte production) aimed at reducing the error
goes down as the error (infection) decreases. Muscle cramp may involve
positive feedback; I don't know. And a Tazer strike is just a huge
disturbance that temporarily puts some control systems out of action;
no positive feedback there.

I think a good example of positive feedback is scratching a bug bite;
the more you scratch the more you feel like scratching.

Another good example is sex (I'm doing this from memory so I may be
wrong;-); the more you kiss and hug the more you want to kiss and hug,
until things, well, limit out.

That's all the excitement I can take for today;-)

Best

Rick

···

--
Richard S. Marken PhD
rsmarken@gmail.com
www.mindreadings.com

to
[John Kirkland 20130813]

Nice question Martin, which brings me to another of my little conundrums: the extent to which any system control, and perception too for that matter, is intentional. Awareness may be an artifact, it just happens in certain conditions but is not foundational.

Oliver Sacks once remarked we don’t have a disease, the disease has us. I quite like this reversal of conventional thinking: it helps put us in our biological place.

Similarly, perhaps, we don’t control anything (that’s another illusion); rather, control systems control us (at least so far as our perceptions may become associated with intentions, aka intelligence) which in simulations may be interpreted as self-evident control. Hence the idea of perception is a convenience, a way-station within a process. Do cells ‘perceive’? No, and yes.

What brought me to this point is the different ways the word ‘control’ appears to be deployed. In some instances it references a system (it is a control system; affect and be affected) and in others it represents an inclusion of personal agency (the person is apparently doing some controlling, eureka there’s volition).

I’m gradually coming to understand these double-speak aspects of PCT.
The diagram in Fig 5.1 helped my understanding since it characterises an isomorphism between objective/physical actual real world instances (objectivity, to the left of centre) and presumed psychological workings (to the right) (also line 5, pg 60). Though for the life of me I cannot ‘see’ the ‘higher loop’ referred to on line 2. pg 58.

Further, line 1 after sub-heading on pg 59, raises a question: is there any other type of relationship than those that are self-evident? Where ‘self’ may not necessarily mean being aware, at least not in the conventional sense. Of course the system is aware, or it would not be functioning but this may not also mean it can articulate what this means on a different level.

I am enjoying being surprised: I enter many margin jottings (and report some of these in my notes to this net) and then in a next chapter find these queries tend to be addressed and even answered. Hence I would not be surprised to find more answers in subsequent chapters, resolving what I’m puzzling over at the moment.

Bill remains a wily teacher.

With kind regards

JohnK

···

On Tue, Aug 13, 2013 at 9:50 AM, Martin Taylor mmt-csg@mmtaylor.net wrote:

[Martin Taylor 2013.08.12.17.48]

[John Kirkland 20130813]

Possible candidates for positive feedback where control goes haywire:

Infections, e.g. Tetanus

Muscle cramp

A Tazer strike

Comments welcomed

JohnK

Are the actions in such cases intentional – in other words, do they occur because some controlled perception differs from its reference value?

Martin

Thanks Rick; as usual, helpful remarks to go on with.

···

On Tue, Aug 13, 2013 at 10:26 AM, Richard Marken rsmarken@gmail.com wrote:

[From Rick Marken (2013.08.12.1530)]

[John Kirkland 20130813]

Possible candidates for positive feedback where control goes haywire:

Infections, e.g. Tetanus

Muscle cramp

A Tazer strike

Comments welcomed

Evidence of positive feedback is when action that is aimed at reducing

error seems to increase it, as evidenced by an increase in the action.

So an infection does not seems to be a positive feedback situation

since the action (lymphocyte production) aimed at reducing the error

goes down as the error (infection) decreases. Muscle cramp may involve

positive feedback; I don’t know. And a Tazer strike is just a huge

disturbance that temporarily puts some control systems out of action;

no positive feedback there.

I think a good example of positive feedback is scratching a bug bite;

the more you scratch the more you feel like scratching.

Another good example is sex (I’m doing this from memory so I may be

wrong;-); the more you kiss and hug the more you want to kiss and hug,

until things, well, limit out.

That’s all the excitement I can take for today;-)

Best

Rick

Richard S. Marken PhD

rsmarken@gmail.com

www.mindreadings.com

[Martin Taylor 2013.08.12.19.28]

[John Kirkland 20130813]

I had asked [Martin Taylor 2013.08.12.17.48] : "Are the actions in such cases intentional -- in other words, do they occur because some controlled perception differs from its reference value?"

Nice question Martin, which brings me to another of my little conundrums: the extent to which any system control, and perception too for that matter, is intentional.

I don't think perception can be intentional at all. The deployment of sensor (turning the head, looking round a corner, cupping the ear) is likely to be controlled and therefore intentional, but what comes in through those sensors is whatever is available. "Intentional" is defined in the second half of the quoted question: [Intentional actions] occur because some controlled perception differs from its reference value.

I'm not sure what you mean by "system control", but some of your post reminds me of a time many years ago on CSGnet, when we were being very careful about keeping straight the viewpoint from which we were talking -- analyst's, control unit's, control system's, and I forget how many others. Mixing viewpoints in a message without being clear that you are doing so can lead to confusion, and possibly what you call "double-speak". Maybe some of that is happening here, or maybe the issue is something else. Since I really didn't understant what you meant in several places, it's hard for me to know.

Awareness may be an artifact, it just happens in certain conditions but is not foundational.

Yes. Awareness or conscious perception is a topic that has arisen on CSGnet from time to time when other amusements seem to pall. There are lots of correlative ideas as to the circumstances under which perceptions become conscious, but no suggestions as to mechanism. Maybe we are aware when control is difficult or failing, or when we are letting go of controlling one perception in order to be able to control another. Maybe it happens when we get to the "logical" levels of Bill's hierarchy. Maybe ...., maybe .... Different people have suggested different possibilities, but they are all largely just talking points, and all are peripheral to PCT, in which the existence of a controlled variable has no necessary connection to consciousness.

Oliver Sacks once remarked we don't have a disease, the disease has us. I quite like this reversal of conventional thinking: it helps put us in our biological place.

Neither has either. If you think of "disease" in terms of the causative agent rather than the process that is their result, we and they are each part of the environment of the other. The fact that we have a place in all sorts of ecological webs is just that -- a fact that relies only on the assumption that what we perceive has some relation to a "real world" that we presume to exist. When you talk about bacteria, we couldn't live without the trillions of them that live on or in us. I find it easier to think of "myself" as including all those bacteria (an order of magnitude more numerous than my genetic cells) and their interactions with my other components. Only occasionally do those interactions go wrong and let us experience what we experience as "disease".

Similarly, perhaps, we don't control anything (that's another illusion); rather, control systems control us (at least so far as our perceptions may become associated with intentions, aka intelligence)

There are about three things in that partial sentence that I don't understand. Is my problem with the introduction of the concept "we", presumably meaning some kind of conscious self-awareness that is outside the scope of science (so far)? Each control unit does control something -- the value of its perceptual variable, and that has to be true whether you are a believer in strict HPCT or only in the loosest concept of PCT. How can control be an illusion, if systems do control? Or are you denying the possible existence of a "real world" out there, and saying that everything you perceive is self-generated? Where does "intelligence" come into this? How do our components control "us"? A very mysterious partial sentence -- but I like mysteries. They lead to religions and fascinating rituals.

which in simulations may be interpreted as self-evident control. Hence the idea of perception is a convenience, a way-station within a process. Do cells 'perceive'? No, and yes.

How did "simulations" get in there?

Think of the definition of a "perception" in HPCT. Oversimplified, it is a variable that is at least in part a function of present and/or past values of other variables outside the boundary of the system under consideration (that boundary need not be the skin of an organism). A controlled perception is any such variable that exists within a feedback loop of which another element acts on the environment outside the boundary. Does a cell perceive? Inside just about any cell are many perceptions, a lot of which are controlled. The answer to your question depends on whether you, yourself, consider that fact to mean that the cell perceives. Or are you limiting "perception" to be something of which a conscious organism is knowingly aware. If the latter, you are going into the realm of talk-shop-philosophy rather than science.

What brought me to this point is the different ways the word 'control' appears to be deployed. In some instances it references a system (it is a control system; affect and be affected) and in others it represents an inclusion of personal agency (the person is apparently doing some controlling, eureka there's volition).

I'm not sure what you mean, control being a very specific mechanical construct; but I hazard a guess that you may be talking about levels of perception in the PCT model. A control unit controls one scalar variable -- its controlled perception. A control system is a connected set of one or more control units, maybe millions or trillions of them, which controls one scalar variable per control unit in the control system. The totality of control systems within an organism control what, tautologically, the organism controls. "Volition" at any level could be used to talk about the reference values coming in to the control units at that level, or it could be used to talk about the outputs from that level that provide reference values for the level below. "Personal agency" presumably refers to the entire set of control units, at the top of which (in HPCT) is a set of fixed reference levels, the values of which have been found either within the individual lifetime or over evolutionary time to keep the intrinsic (important for life and reproduction) variables within safe limits. Are those the kinds of meanings you intend those words to convey?

If you think about the mechanism in the theory, and don't overlay it with concepts that belong in a different domain, these sorts of issues (for me, anyway), go away. I can still puzzle about the mystery of consciousness and the sense of self-hood (just where am "I"? Behind my eyes? In my toes -- I don't think so? At my fingertips? It seems to have something to do with what sensors I'm most aware of at the moment. But it has nothing to do with PCT.

I'm gradually coming to understand these double-speak aspects of PCT.

I know of double-speak aspects in most psychological theories, but I haven't found any in PCT, nor have you explained sufficiently for me to understand wherein you have detected them.

Martin