Wonderful conceptual explanation bill. helps immensely with my understanding of pct and mol. Hope all is well and your summer is moving along nicely. respectfully gary padover
[From Rick Marken (2009.07.29.1040)]
Martin Taylor (2009.07.28.17.43)
When one is trying to introduce PCT to a newbie, it can help if the terms
can be mapped onto something the newbie already knows. If control engineers
use the common language, in which the "output" is the controlled variable,
then they can perhaps come to terms with PCT more easily if they are told
not that they are wrong, but that in PCT discussions, the control loop is
split into halves, in which the "input" and the "output" are defined not by
the bounds of the control loop but by the seperatrix of the halves of the
loop. Yes, it's a simple idea, but not simple to one used to thinking of
inputs and outputs of a loop as being from and to the environment external
to the loop.
The problem with this (aside from all the excellent points that Bill
[Bill Powers (2009.07.29.0851 MDT)] just made) is that the diagram and
derivation that I posted were in a book that was explaining how
control theory applied to human behavior. Moreover, these people were
not "newbies" to PCT; they had read Bill's (and my) work and the
result was this: a diagram and derivation showing that human behavior
is controlled output. This is the kind of thing I have had to deal
with since getting into PCT: people who enthusiastically "understand"
my (and Bill's) work and then go off and do things that prove that
they don't. This is most disappointing when the "understanders" are
research psychologists because it means that these folks won't be able
to help me out with PCT research (and I can use all the help I can
get). PCT research requires that the researcher understand that
control involves the control of perceptual variables. Without that,
all one has to study are statistical illusions.
Best
Rick
···
--
Richard S. Marken PhD
rsmarken@gmail.com
[Martin Taylor 2009.07.29.14.42]
[From Bill Powers (2009.07.29.0851 MDT)]
Martin Taylor 2009.07.28.17.43 --
The inputs to a physical control system enter its sensors or input functions, and the outputs come from its effectors or output functions. Why make it more complicated than that?
What is "a physical control system" to a naive user? Consider a thermostat. To an everyday user, the input to the thermostat is the dial setting (to a PCT scientist, the reference value), while the output is the desired temperature (the controlled variable). Very simple.
Yes, for uninformed users. Typically of most oversimplifications, it misinforms more than it informs.
I omit the rest of your message (apart from a couple of quotes later), because I agree with pretty well all of it. But I think you miss the pedagogic implications of what you say, or at least you didn't bring them out in what you wrote. Your message consists thematically of "I'm not interested in ..." the way other people come to PCT, and "they are wrong". To me, that message is not useful in an effort to enlist support to the cause of developing and propagating PCT.
In my limited experience, when a newbie approaches PCT and is given encouragement, one of the first traps about which the newbie is warned is that in PCT language "perception" does not mean the content of conscious awareness. That's important, since even from people contributing to CSGnet one often sees a conflation of conscious awareness with the perceptual signal.
Never have I seen the trap about "input" and "output" made equally clear to the newbie. It's a trap for someone unused to control theory, but it's an even trickier trap for a control engineer who thinks in terms of constructed control systems. A constructed control system is designed so that the "input" reference signal causes the "output" controlled variable to take on a desired value. The constructor clearly has that purpose in mind. The purpose of the control system is to produce just that output, in the face of environmental variation (which we would call "disturbance"). The constructed control system IS an input-output system built with a feedback loop. It's not an error to say so, or to talk about it being so.
Now a control engineer comes to PCT, and would like to contribute his understanding of how control works. If he is not warned explicitly that what he calls "output" is what we call "the controlled variable", that the sensor that measures the state of that variable creates a "perceptual variable" that is really what is controlled, and that "input" is a combination of the disturbance with the effect of the "output" passed through the "environmental feedback function" then he will be in a perpetual state of confusion. Once the terminological difference is made clear (not by telling him that he is wrong, and stupidly so), the control engineer should be halfway to being able to contribute usefully to the development of PCT.
You say: "I'm not interested in thinking of the controlled variable naively; I want to think about it correctly and consistently." That's fine. You have a half-century of seeing all the implications of this way of thinking. But if one is naively becoming aware of PCT, one is perhaps unaware that the naive view and language is different from the PCT view and language.
I entered this thread because I could see no evidence one way or the other in what Rick presented to suggest that the authors had no idea about how control systems worked in living systems, and I could also see that the contributors to the thread had all assumed that Rick's implication was true. Without the evidence that might be obtained from reading more of the book, I thought it quite reasonable to assume that the authors were using the everyday, usual, naive nomenclature, that the output of a control system is the controlled variable (as the diagram and their correct analysis showed).
MMT earlier:
All I'm saying is that nomenclature can be a barrier to understanding, if the same words are used to mean something one does not expect them to mean. And most people think of the output of control as being the controlled variable.
BP response:
Which is a barrier to understanding. They should change their understanding because it is wrong and impossible to use consistently.
If I call a faulty car a "lemon" I am also wrong, because a lemon is a yellow citrus fruit. If the whole English speaking world other than a hundred or so CSGnet readers use the word "output" for the controlled variable of a control system, it's arguable as to who is "wrong". Also, I can't imagine why it is impossible to use that naming consistently. One can use "controlled variable" consistently for it, or one could call the controlled variable the "Xylogism" equally consistently. Why is it impossible to use "output" for that variable? Is it simply because you personally want to use "output" to mean what emerges from a skin bag to influence the world outside the bag, and to use it for both would call confusion?
Look only at the control loop, and not at its physical instantiation, and you see two, and only two, outputs. One is the perceptual signal (the controlled variable), and the other is the side-effects. Look only at the physical instantiation and not at the abstract function, and you see one output, the muscular effects on the world (or in the case of the thermostat, a voltage that is sent to a furnace or airconditioner. As I pointed out (and you agreed, though in a castigatory mode) the difference is the point of view. The physical output of the effectors is within the control loop; it is not an output of the control loop. The PCT preference in nomenclature is to mix the points of view, looking at the abstract loop while defining the output from the physical instantiation viewpoint. You emphasise this mixture of viewpoint (between "effector output" and "control system output" quite effectively when you say: "controlled variables are not outputs from the control system, and I can prove it by measuring the actual effector outputs and showing that they do not always, or even very often, covary with the state of the controlled variable." But the controlled variable (the perceptual signal value) is indeed an output of the control system in the usual PCT diagram -- but the arrow points upward, not downward.
Think sympathetically of the person coming to PCT with linguistic habits that are built from a non-PCT lifetime. What they know about control may be right or it may be wrong, but it is guaranteed that they will use words such as "perception", "input", and "output" in a way with which they are familiar. It is guaranteed that their way of using these words is not the way they are used in PCT discussions, and until they realize the difference, it is guaranteed that they will not be able to understand PCT.
Martin
[Martin Taylor 2009.07.29.15.29]
[From Rick Marken (2009.07.29.1040)]Martin Taylor (2009.07.28.17.43)When one is trying to introduce PCT to a newbie, it can help if the terms can be mapped onto something the newbie already knows. If control engineers use the common language, in which the "output" is the controlled variable, then they can perhaps come to terms with PCT more easily if they are told not that they are wrong, but that in PCT discussions, the control loop is split into halves, in which the "input" and the "output" are defined not by the bounds of the control loop but by the seperatrix of the halves of the loop. Yes, it's a simple idea, but not simple to one used to thinking of inputs and outputs of a loop as being from and to the environment external to the loop.The problem with this (aside from all the excellent points that Bill [Bill Powers (2009.07.29.0851 MDT)] just made) is that the diagram and derivation that I posted were in a book that was explaining how control theory applied to human behavior.
Yes, you said that. But you have never shown a shred of evidence that
they don’t understand how control theory applies to human behaviour.
Since you have read the book, I’m quite willing to accept that you are
correct. Why don’t you provide the evidence that must exist in the rest
of the book?
Moreover, these people were not "newbies" to PCT; they had read Bill's (and my) work and the result was this: a diagram and derivation showing that human behavior is controlled output.
Read my messages about nomenclature. I’m not going to repeat yet again
why it’s perfectly reasonable to expect someone writing for the general
public or for students unacquainted with PCT to use the term “output”
to refer to the controlled variable. It’s pure hubris to require the
world at large to change their language because a hundred or so
scientists use the words differently. If they indeed knew Bill’s work
and your own, it seems more likely that the diagram was drawn so as to
be more familiar to people using everyday language. Only a reading of
the rest of the book can indicate whether they follow through correctly
with the pedagogy.
PCT research requires that the researcher understand that control involves the control of perceptual variables. Without that, all one has to study are statistical illusions.
Yes, but attitudes like yours are counterproductive. Rather than
complaining that they use “output” to refer to the controlled variable,
which is, after all, one of the outputs of a control system in PCT, why
not simply point out that their output IS the controlled variable, and
complain (if the complaint is valid) that they don’t correctly analyze
how the controlled variable relates to behaviour.
Martin
[From Rick Marken (2009.07.29.1430)]
Martin Taylor (2009.07.29.14.42) --
I entered this thread because I could see no evidence one way or the other
in what Rick presented to suggest that the authors had no idea about how
control systems worked in living systems
I didn't say they had no idea; I said they had the wrong idea, at
least as evidenced by the diagram and derivation.
Without the evidence that might be obtained from reading more of the book, I
thought it quite reasonable to assume that the authors were using the
everyday, usual, naive nomenclature, that the output of a control system is
the controlled variable (as the diagram and their correct analysis showed).
But a controlled variable is defined by the function that converts
physical variables into the controlled perceptual variable. In a
thermostat that function is computed by the transducer that converts
physical variables (like temperature, pressure and humidity) into the
electronic (perceptual) signal that is actually controlled. If the
authors of the diagram (and derivation) were using the word "output"
to refer to a "controlled variable" then we would expect to find in
that diagram some indication of such a transducer. But there is none.
Clearly, what the authors meant by "output" is the variable that the
observer perceives (via the observer's own transducer -- ie., some
temperature measuring device or their own senses) as the output of the
system, which is not necessarily the same as what the system is
perceiving and controlling (due to the nature of the
transducer/perceptual function the thermostat may actually be
perceiving and controlling some weighted combination of temperature
and humidity, say). This latter fact is the reason why we have to test
to determine what a living control system is controlling in order to
understand its behavior.
The whole point of PCT is that you can't just look at behavior (the
perceived output of the system, either it's actions or consequences of
its actions) and see what it is controlling. The authors of the
diagram I posted clearly don't understand this point at all. They
assume that the output under control can simply be "seen"; that it is
obvious to an observer. In fact, the builder of an artificial control
system does know what variable the system _should_ be controlling (the
variable, like room temperature, that the system is designed to
control) and so the observer of the system would measure this "output"
variable to see how well the system is controlling what it's designed
to control. But the situation is completely different when you are
dealing with living control systems. The problem with living control
systems is that we don't know what they are controlling; all we have
are hypotheses about what variables they might be controlling. We have
to test these hypotheses if we want to actually know what the system
is doing.
So, no, the problem with the diagram/derivation is not just on of
terminology: you say "output" and I say "input". The problem is
missing the most important point that control theory has to make about
living control systems: behavior is the control of perception! And you
can't perceive what another system is perceiving. All you can do is
_guess_, which is the first step in the test for the controlled
variable.
Best
Rick
···
--
Richard S. Marken PhD
rsmarken@gmail.com
[From Rick Marken (2009.07.29.1450)]
Martin Taylor (2009.07.29.15.29) --
But you have never shown a shred of evidence that they
don't understand how control theory applies to human behaviour.
The evidence is the diagram (and derivation) that show the output as
the controlled variable. The other piece of evidence is that nowhere
in the book do they discuss the procedure for testing to determine
what variable is under control. If they were really thinking of the
output as a controlled variable then they should have had a section in
the book about how to test for the controlled output. But of course
they didn't have such a section because they would see no need for it;
the output is just obvious, as it is with the thermostat, or so they
think (they're not even necessarily right about what the thermostat
output is, as I noted in the previous past)
Read my messages about nomenclature. I'm not going to repeat yet again why
it's perfectly reasonable to expect someone writing for the general public
or for students unacquainted with PCT to use the term "output" to refer to
the controlled variable.
I explained in the previous post why this is not just an issue of
nomenclature. There is nothing in the diagram, derivation or book that
suggests that the output variable is a function of physical variables
or that the nature of the function that produces this "output"
(actually, controlled) variable is unknown to an observer of the
system.
RM: PCT research requires that the researcher understand that
control involves the control of perceptual variables. Without that,
all one has to study are statistical illusions.MT: Yes, but attitudes like yours are counterproductive. Rather than complaining
that they use "output" to refer to the controlled variable, which is, after
all, one of the outputs of a control system in PCT, why not simply point out
that their output IS the controlled variable, and complain (if the complaint
is valid) that they don't correctly analyze how the controlled variable
relates to behaviour.
OK, they don't correctly analyze how the controlled variable relates
to behavior. I described what is incorrect about their analysis in my
previous post: they don't understand that the controlled variable
(their "output") is a perceptual function of physical variables and
that an observer cannot know what this "output" actually is without
doing the test for the controlled variable.
Another problem, by the way (which involves only nomenclature) is that
if you use "output" to refer to the controlled variable then what do
you call the unlabeled variable in the diagram that comes out of the
system and joins with the disturbance to produce the "output"?
Amplified error?
Best
Rick
···
--
Richard S. Marken PhD
rsmarken@gmail.com
[From Bill Powers (2009.07.29.1458 MDT)]
Martin Taylor 2009.07.29.14.42 –
… I think you miss the
pedagogic implications of what you say, or at least you didn’t bring them
out in what you wrote. Your message consists thematically of “I’m
not interested in …” the way other people come to PCT, and
“they are wrong”. To me, that message is not useful in an
effort to enlist support to the cause of developing and propagating
PCT.
Of course that’s not what I say to them. So far I’ve had only one failure
to communicate the “control of input” idea to control
engineers. In one case the engineer said he was too used to the other way
of thinking about it and didn’t intend to change, and signed off. He
understood it immediately; he just didn’t like it, and his reasons for
that, as he expressed them, had nothing to do science or
engineering.
Another factor you’re overlooking is that when I teach PCT, my main
message is that the person I’m talking to is the control system and he is
looking at the world from the standpoint of that system, not of the
external observer. It’s not hard to get people to see that whatever
they’re controlling, they’re perceiving it and controlling the
perception, and I’ve never (before) run into an objection to calling a
perception an input. Similarly, people can easily see what actions
they’re using to control perceptions, and it’s really not hard to
convince them that their actions are (perceived) outputs. It’s only
people who view control systems entirely from outside who have any
problems with this, and even they normally have no difficulties with it
when the diagram is carefully explained. After all, most engineers are
used to block diagrams and use them often to communicate.
In my limited experience, when a
newbie approaches PCT and is given encouragement, one of the first traps
about which the newbie is warned is that in PCT language
“perception” does not mean the content of conscious awareness.
That’s important, since even from people contributing to CSGnet one often
sees a conflation of conscious awareness with the perceptual signal.Never have I seen the trap about “input” and “output”
made equally clear to the newbie. It’s a trap for someone unused to
control theory, but it’s an even trickier trap for a control engineer who
thinks in terms of constructed control systems.
Perhaps, but it’s not hard for people other than engineers to understand.
It’s not even hard for most control engineers I have met to understand.
After all, it’s a pretty simple idea, and engineers are not dumb. When
most engineers see what I am calling the output of the controller, they
agree that it is the output of the controller. And when I point out that
in an organism, it is the perception that is actually being controlled,
they usually understand that immediately, too, or after a moment’s
thought.
The constructed control
system IS an input-output system built with a feedback loop. It’s not an
error to say so, or to talk about it being so.
But an organism is not an input-output system, and it is an error, if it
has been explained that the input we refer to is a sensory input.
“Oh, that input!” the engineer says. Even control-system
engineers consider sensors to be input devices. It makes even more sense
after it is explained that the reference input to the comparator in a
living control system is not accessible for manipulation as an
independent variable from outside the organism: the only inputs
that can be affected from outside are the sensory inputs (ingestion and
injection aside). It’s only in systems manufactured for a customer’s use
that a way is provided for an external agent to manipulate the reference
signal directly. Organisms are not manufactured for a customer’s use. I
haven’t had any great difficulties in communicating that idea.
Now a control engineer comes to
PCT, and would like to contribute his understanding of how control works.
If he is not warned explicitly that what he calls “output” is
what we call “the controlled variable”, that the sensor that
measures the state of that variable creates a “perceptual
variable” that is really what is controlled, and that
“input” is a combination of the disturbance with the effect of
the “output” passed through the “environmental feedback
function” then he will be in a perpetual state of
confusion.
Sorry, Martin, but this whole post is just a philosophical argument for
the sake of argument. You can think about the variables and their
connections any way you please, and attach any labels to them you please,
and the system will go on working in exactly the same way. Each variable
has a value determined by the variables that affect it; each variable
affects one or more other variables. The equations describing the loop
are not affected by the words you use. And their solutions remain the
same.
I can explain the PCT diagram to any engineer without using the words
input or output at all.
The variables qi and qo are names of physical variables in the
environment of the controller. Qo is a function of the error signal only;
the perceptual signal p is a function of qi only. Qi represents the value
of a joint function of qo and d, the disturbance. It is qi that is held
at a specific value when d fluctuates; qo fluctuates in such a way as to
nearly cancel the effects of d on qi. By the same token, variations in qo
hardly affect the value of qi. The action of the closed loop is such as
to maintain the perceptual signal in a close match to the reference
signal. If maintaining that close match is the definition of control,
then this closed-loop system controls the perceptual signal. The behavior
of the controller, its action on the environment, is seen in the changing
values of qo.
No customary usages of terms like input and output matter a bit when you
explain the various functions in the control loop mathematically. An
engineer looking at the PCT diagram, in which more aspects of the loop
are shown explicitly than in the customary engineering diagram, would see
nothing unfamilar – there would simply be more detail than
usual.
That is all that has to be said to any engineer. All the rest is just
playing with words.
Best,
Bill P.
[Martin Taylor 2009.07.30.09.59]
[From Rick Marken (2009.07.29.1450)]Martin Taylor (2009.07.29.15.29) --But you have never shown a shred of evidence that they don't understand how control theory applies to human behaviour.The evidence is the diagram (and derivation) that show the output as the controlled variable. The other piece of evidence is that nowhere in the book ...
Thank you. Finally, this is what I wanted to hear from you. As you
know, I did not think the diagram and derivation provided any evidence
at all for your claim. If you can say “Nowhere in the book…” listing
all the things that ought to be in the book and are not, that is
evidence that does support your claim.
Thanks again for finally presenting the evidence. I think we can put
this portion of the thread to rest.
I am continuing the other portion, about nomenclature, under a new
subject heading “Control units, systems and loops”.
Martin
[Martin Taylor 2009.07.30.10.13]
[From Bill Powers (2009.07.29.1458 MDT)]
Martin Taylor 2009.07.29.14.42 –
In my limited
experience, when a
newbie approaches PCT and is given encouragement, one of the first
traps
about which the newbie is warned is that in PCT language
“perception” does not mean the content of conscious awareness.
That’s important, since even from people contributing to CSGnet one
often
sees a conflation of conscious awareness with the perceptual signal.Never have I seen the trap about “input” and “output”
made equally clear to the newbie. It’s a trap for someone unused to
control theory, but it’s an even trickier trap for a control engineer
who
thinks in terms of constructed control systems.Perhaps, but it’s not hard for people other than engineers to
understand.
It’s not even hard for most control engineers I have met to understand.
Of course it’s not hard for them to understand IF the difference in
nomenclature is made clear. That’s my point! I almost think you are
agreeing with me, but making an argument because you like to make
arguments.
The constructed control
system IS an input-output system built with a feedback loop. It’s not
an
error to say so, or to talk about it being so.But an organism is not an input-output system, and it is an error, if
it
has been explained that the input we refer to is a sensory input.
Quite. “if it
has been explained that the input we refer to is a sensory input”.
Sorry, Martin, but this whole post is just a philosophical argument for
the sake of argument.
If you think that, you miss the point. It’s a practical argument, whose
point has been made by the problems sometimes encoutered in introducing
PCT, and has been re-emphasised by each of your messages in response to
my initial contribution to the thread, and especially by the message to
which I am responding, in which you remake my point by saying things
like “once it has been explained” and
I can explain the PCT diagram to any engineer without using the words
input or output at all.
(thus avoiding the problem) andNo customary usages of terms like input and output matter a bit when
you
explain the various functions in the control loop mathematically. An
engineer looking at the PCT diagram, in which more aspects of the loop
are shown explicitly than in the customary engineering diagram, would
see
nothing unfamilar – there would simply be more detail than
usual.That is all that has to be said to any engineer. All the rest is just
playing with words.
Yes, yes, yes!!! THAT IS MY POINT! “Playing” with words misleads those
who don’t realize that you use words differently than they use them.
Mathematics doesn’t, for someone who understands the math.
You think it’s an unimportant detail. I think it’s a practical problem
that goes a little beyond the newbie tutorial issue, and I’m starting a
separate thread under the title “Control units, systems, and loops” to
consider it.
Martin
[From Rick Marken (2009.07.30.0910)]
Martin Taylor (2009.07.30.09.59)--
MT: But you have never shown a shred of evidence that they
> don't understand how control theory applies to human behaviour.
RM: The evidence is the diagram (and derivation) that show the output as
the controlled variable. The other piece of evidence is that nowhere
in the book ...
MT: Thank you. Finally, this is what I wanted to hear from you.
I'm sure I mentioned it before. But no matter; it's nice to have you
agree that the people who made the diagram/derivation that I posted
clearly do not understand how control theory applies to human
behavior. That is, they don't understand PCT. Again, the only reason I
posted this is to show that it is possible to be able to understand
control theory as an engineering and/or mathematical discipline and
have no idea how to apply it correctly to understanding living
systems.
Best
Rick
···
--
Richard S. Marken PhD
rsmarken@gmail.com
[From Bill Powers (2009.07.30.1100 MDT)]
Martin Taylor 2009.07.30.10.13] --
BP earlier: That is all that has to be said to any engineer. All the rest is just playing with words.
MT: Yes, yes, yes!!! THAT IS MY POINT! "Playing" with words misleads those who don't realize that you use words differently than they use them. Mathematics doesn't, for someone who understands the math.
BP: But what I don't understand is why you assume that I don't say these things to engineers or newbies, as appropriate. I always do and always have. All my demonstrations are designed to make these points clear, and I explain them in words, too. Maybe not every single time, but almost always, and eventually if not right away. What I say depends on what a listener seems not to understand, so with some people (like the present company) I may skip over that part.
MT: You think it's an unimportant detail.
BP: Not at all. It's essential, if there is reason to think it's not already understood. But it's also a little insulting to explain something to a person who is very likely to know it already (like a certain person at a CSG meeting, telling an audience consisting of probably 80% PhDs, what the "ring of fire" is, in tones indicating he really thought this was news).
Best,
Bill P.
[From Rick Marken (2009.07.30.1045)]
Bill Powers (2009.07.30.1100 MDT)--
Martin Taylor 2009.07.30.10.13] --
MT: You think it's an unimportant detail.
BP: Not at all. It's essential, if there is reason to think it's not already
understood. But it's also a little insulting to explain something to a
person who is very likely to know it already (like a certain person at a
CSG meeting, telling an audience consisting of probably 80% PhDs,
what the "ring of fire" is, in tones indicating he really thought this was
news).
I don't get the reference. I have a PhD and I don't know what a "ring
of fire" is. What CSG meeting was that? I've been to every one (except
for some international ones, I suppose) and I don't recall anyone
telling me about a "ring of fire". Or was it, possibly, me breaking
into my Johnny Cash imitation (although I'm a natural tenor I can get
down into Johnny Cash territory when I sing).
This whole discussion with Martin has been (as usual) pretty confusing
to me. Martin's argument with me seemed to be that the people who made
the diagram I posted actually understood PCT (the proper mapping of
control theory to living systems), they just used different words to
describe it (I think he's now conceded that these people did not
understand PCT correctly at all). His argument with you seems to be
that we have to be willing to talk about PCT in a language that
control theorists (like those who posted the diagram) would
understand, implying that the people who posted the diagram actually
didn't understand PCT but could be tutored in it if the right words
were used. I guess I'll just keep watching to see what happens.
But please tell me about the "ring of fire" thing.
Best
Rick
···
--
Richard S. Marken PhD
rsmarken@gmail.com