from [ Marc Abrams (990722.0102) ]
I'm sitting here by my PC mentally flipping back and forth between
reflecting on Jeff's and Fred's posts and going through the archives. In my
Don Quixoteisk trek for "Truth, Justice, and the American way", plus Chap 15
of B:CP My mind kept on creeping back to a thread I found very
compelling. The thread, as so many do on CSGnet, started in one place and
ended up in another. Actually, that is probably not all that unusual on any
list on the net. The importance of the thread ( in my mind ) lies in how the
ideas are presented. Not the technical PCT ones, but the way we assimilate
information and communicate with others about what we perceive. The thread
involved a bit of Bill Powers with Dag Forssell and Rick Marken as the main
players at This point This is _NOT_ about Dag or Rick or Bill. It's about
how ideas on this net are and aren't communicated. I could have changed the
names, but all this is a matter of public record. Just let me repeat this
one more time _THE INTENT OF THIS POST IS NOT TO FOCUS ON THE
INDIVIDUALS INVOLVED_ . All of us have been there and done that.
Bill might cringe at this, but I think it might be worthwhile to think
about how we can be more effective both in communicating and advocating our
ideas while we open ourselves up for inquiry into our own logic and
reasoning. I think both Bill and Dag did a very good job in both explaining
why and how we might do it.
Any comments?
Marc
[From Bill Powers (930915.1015 MDT)] >>[Avery Andrews (9309151309)]
[ Mike Fehling ]
I heartily second Avery Andrews' comment (930915.1309):
This discussion seems to be going the way of many, wherein PCT people seem
to want to perceive themselves as clearly enunciating the principles of PCT,
rather than figuring out what actually is on the other person's mind, >and
seeing whether there is a difference between that and any claim of PCT.>>
.. and also Michael Fehling's observation:
I find these attempts to drive a wedge between PCT and other theories
regarding causality are misleading and contentious.>>>
There are no doubt many differences to clear up, but the way to do it is
not to guess what's on the other person's mind. Most people who have gone
past a certain point in grasping the structure of PCT carry on most of the
"contention" within themselves, as situations arise that bring out conflicts
between their own previously-held concepts and the PCT interpretation. The
theory becomes self-teaching after that point; all that's needed is time. It
does no good to try to push a person into accepting basic conclusions from
PCT if that person hasn't built up the infrastructure of understandings that
make the conclusion look reasonable.>
I think that the best approach is to try to narrow differences down to
specific points where there is a clear contradiction. This allows people to
bring out the premises that are behind the contradictions and lay them out
in full view. Just arguing about conclusions is attacking the conflict at
the wrong level. A conclusion is just output at a certain level. The
conflict can be resolved only by finding out WHY that output is being used,
not just insisting harder on the "right" output.>
I am lecturing at myself as much as anyone else.
Me:
Great points, and I think Dag has a way of dealing with this. I did not
include Rick's post because Dag addressed the essential points of Rick's
post concerning our ability to communicate.
[From Dag Forssell (930916 1210)]
Rick Marken (930915.2330)] >>>Michael Fehling (930915 4:44 PM PDT)
You seem to have an equally "interesting" interpretation of controller
dynamics. In fact, p _seldom_ equals r, _dynamically_. If it did, one
wouldn't need the control loop in the first place.>>>
The dynamic variations of p about r are typically orders of magnitude
smaller than the potential dynamic range of either p or r. The control
loop keeps p virtually equal to r; the fact that p varies around r at
all represents failure of control; in a good control system the size of
the variations of p around r be so small that they are undetectable
by the instruments used to measure p.
As usual, Rick, you immediately go to the idealized condition of (almost)
infinite loop gain, by drawing your conclusions from the simplified math
where all the quantities divided by the loop gain have been dropped. Many of
the heated arguments on this net have been based on this extreme and
oversimplified position.>
I think you make PCT much harder to comprehend than it needs to be by
making this argument. If you consider our classic rubber band demonstration
(Michael, see July 17 video script p. 9 & 12), the loop gain is far from
infinite (it can still be considered a "good" control system) and Michael's
point above is completely valid.>
Me:
Here is the wisdom of Solomon speaking.. Absolutely on target
Dag:
A more balanced approach would be to point to an agreement with Michael, as
shown so clearly with the rubber bands, followed by an every-day example of
what you try to say. Perhaps one could say that when writing longhand, the
position of the pen is A) very closely controlled within a very small
fraction (.01 inch) of the potential range of your hand's reach from the
writing position (60 inches). It is so closely controlled that we B) do not
notice any discrepancy between the intended and perceived writing.?>
Rick:
Your statement above suggests that you believe that the dynamic deviations
of p from the reference level, r, are what drive (cause?) the outputs of the
control system. That is, dynamic (temporal) variations in (r-p) are the
cause of the temporal variations in the outputs that affect p, keeping p
near r. Is this what you think is going on in a control loop?>>
Dag:
This certainly is what I think. The temporal deviations of p from r create
an error signal e, which _contains >information_ used by the output function
(again, see video script p. 11). Rick, as I read your further argument in
this post, the best I can figure is that you write about some idealized
conception of a control system which takes an error signal as an instruction
to output any which way (which after trial and error proves successful).
deny the obvious existence of a real, demonstrable control system in the
here and now, arguing instead for an ivory tower, unspecified function f()
with unreal properties, including the full effect of reorganization over a
long time period.>
It has been recently suggested that it may be better to seek points of
agreement before you engage in battle. Covey in _The Seven Basic Habits of
Effective People_ suggests: Seek first to understand, then to be
understood.>
This agrees very closely with Ed Ford's questioning approach.
It also agrees with the PCT lesson that people resist disturbances.
Please consider applying your understanding of PCT and reconsider your
reply to Michael.
Best, Dag
Next comes Rick's reply to Dag's post
[From Rick Marken (930919.2100)] >>[ Dag Forssell (930918 0025) says:
That is not the way I read you. You also said:
Michael Fehling:
in a good control system the size of the variations of p around
r be so small that they are undetectable by the instruments used
to measure p.
Dag:
So Michael was talking about a bad control system. I certainly read
your statement as a rejection of Michael's observation.
Rick:
Sorry. Poor writing on my part. I should have said "high gain" instead of
"good".
I said (to Michael) --
Your statement above suggests that you believe that the dynamic deviations
of p from the reference level, r, are what drive (cause?) the outputs of the
control system. That is, dynamic (temporal) variations in (r-p) are the
cause of the temporal variations in the outputs that affect p, keeping p
near r. Is this what you think is going on In a control loop?>
Dag says:
A good interpretation of Michael and an innocent question. You proceeded
to argue against it, without any qualifying transition, which I interpreted
as you saying that Michael is a half-baked idiot for thinking so. This
created a large disturbance in me. Your clarification does nothing to change
my impression.>>
Rick:
Apparently it was a disturbance to many perceptions -- including Bill's --
so you are in good company. There's not much I can do about your
interpretation of my post (it's yours). All I can do is assure you (and
Michael and everyone else who might care) that I am not in an intellectual
sparring match; I'm not trying to prove that I am smart and everyone else is
an idiot. I am sure everyone on this net is a lot smarter than I am. There
are one or two things that I think I understand -- so they are very precious
to me. But if it turns out that I am wrong about them (or anything else)
then that will be fine and I will have learned something. The net, for me,
is about batting around ideas -- NOT winning the battle for smartest person
(I already concede defeat). I'm sensitive about this because I know how
unpleasant such substanceless, intellectual sparring matches -- aimed only
at improving one's own imtellectual image by verbally humiliating your
opponent (a la W. F Buckley) -- can be. For me, it's the ideas that are
important -- not showing that "I know X, you don't so you're stupid".>
I am prepared to be shown that I have said something wrong. I'm trying to
learn stuff too. If I'm shown that I'm >wrong about something then this is a
learning experience. I'm sure that much of what I say is wrong -- I'm not
speaking ex cathedra here . Nor is discussion on this net an ego trip (for
me, anyway) -- it's a learning experience. I enjoy the ideas and I think
they're important. If someone says something that conflicts with my
understanding (meager as it is) of PCT then I will call them on it --
knowing that I might be wrong. I have been wrong plenty, I accept it when
someone explains how I am wrong (as I did fairly recently when Bill P.
pointed out that one of my efforts at a mathematical proof of some concept
was a hash).>
I participate on this net because 1) I want to communicate what I know (or
think I know) about PCT with those who are interested 2) I want to learn
from those who know things I don't know and 3) I want to work towards a
better understanding of organisms as perceptual control systems. If someone
says something that seems wrong to me, I feel free to say so -- and try to
explain what seems wrong about it. I expect people to call me on it >when I
say something that seems wrong (and they do call me -- and when they do I
don't feel like they are saying that I'm stupid -- even if they are)>.
Remember, when I say something like "That's wrong" I am not saying "you're
stupid and I'm smart" or "I'm better than you" or "even a 2 year old could
understand this" or anything else that has to do with intellectual
competition. I'm just trying to get the PCT perspective on living systems
right.>
So with that apologia out of the way, let's see what (I think) is going on
with a control systems.
Dag says:
Would you please clarify what you have typed above. Is perceptual input
equivalent to the error signal e = r-p?>>
Rick:
I have been somewhat ambiguous, I'm afraid. In manual control the input to
the control system is error (r-p) where r and p (and e) are considered to be
variables in the environment. In PCT we are very clear about the fact that r
and p are inside the person and r is only manipulable by the person. But in
my development of the behavioral illusion the location and value of r and p
are irrelevant. From now on let's assume that the input to the system is p,
that r is zero and that the comparator is just an inverter -- so e=-p (which
is equivalent to your equation with r=0). The reason is that the behavioral
illusion does not depend on the location or value of r; it depends on the
fact that organisms are in a negative feedback relationship with their
environment. As Bill P. said in the Psych Review (1978) paper, the
behavioral illusion occurs because organisms are N - systems (systems in a
negative feedback relationship with their environment) not Z-systems (open
loop systems -- zero feedback relationship with their environment) or
P-systems (positive feedback relationship).>
Dag:
The temporal deviations of p from create an error signal e, which
_contains information_ used bythe output function (again, see video script
p. 11).>>
Rick:
This part is basically true
Dag:
So the only thing Michael said is true. But in a loop it loses its
validity? I don't think so.
Rick:
I was responding to the fact that what Michael said could be understood (by
me anyway) in wo ways: 1) as a description of a sequential input-output view
of how control works (the "Plans and Structure of Behavior" approach to
describing control) or 2) as a description of the relationships in a control
loop. The first meaning (which might not have been Michael's -- that's why I
asked) causes problems for me becuase it is at the heart >of psychology's
failure to understand the nature of control. It implies a sequential state
analysis of the behavior >of a control system -- which leads to the wrong
results.>
Dag:
I am aware of the behavioral illusion. I have always thought that the
illusion is to think that there is a direct causal relationship between
input p and output o (ignoring r).>>
Rick:
No. The illusion is that observed relationships between stimulus variables
(d) and output variables (o) in N- systems reflect characteristics of the
organism; in fact, they reflect characteristics of the feedback function
(not a part of the organism) that relates o to p). Experiment 4 on pp
156 -159 of LCS I is a good demonstration of the behavioral illusion; note
especially fig. 6 on page 157. This is a plot of the relationship between d
and o in a >tracking task; this is an input -- output plot; it would be
reasonable to assume that the nature of this relationship (cubic) depends on
characteristics of the subject; in fact, it depends entirely on the nature
of the environmental function connecting output to perceptual input.>
This is a stunning demonstration because it really pulls the main support
away from the foundation of scientific psychology. I think that it is VERY
important (for psychologists, anyway) to understand what is going on there.
You can demonstrate it to yourself with rubber bands, if you like. Do the
rubber band demo twice, once with a very tight and once with a very loose
rubber band connecting the subject to the knot. In the first case, you will
see that the subject makes very small finger movements per unit movement of
your finger. In the second case, >the subject makes very large movements for
the same unit movements of your rubber band. It looks like the subject has
become more sensitive in the second case; reacting much more to the same
disturbance that produced almost no reaction the first time. But there is no
change in the subject; just in the environmental connection from subject to
controlled variable. A psychologist who didn't know that the subject was
controlling the knot (and doing whatever was necessary to keep it at the
reference) might conclude that the subject is more sensitive to certain
kinds of rubber bands than to others. THAT'S the behavioral illusion.>
I was discussing a version of the illusion where one measures the
relationship between -p and o instead of between d and o. The illusion still
occurs (the observed relationship between -p and o is not what is expected
based on knowledge of the real function relating p to o) but I suppose I am
wrong to call it the behavioral illusion because it could only be measured
neurophysiologically; so maybe I should call it the neurological illusion.>
Dag:
Have you perceived that this is what Michael and I mean when we write
about a direct causal relationship between the error signal r-p and output
o?>>
Rick:
No, because my point had nothing to do with r (that's why I included it in
the input, as r-p). I think you didn't understand what I meant by
"behavioral illusion" (but I still think that you're smarter than I am).>
Dag:
If by output you mean the momentary muscle contractions and therefore hand
acceleration, I think Michael's and my notion of a direct causal
relationship between r-p and o holds very well indeed.>>
Rick:
Agreed. But remember, in my question I asked if what was being caused were
"the outputs that affect p, keeping p near r". The derivative of o (in the
model) is not what keeps p near r; it is what is caused, but it is the
current value of the integral that is the effective output. The integral is
what preserves the effects of all variables around the loop -- and makes it
so that the derivative that is "caused" at any instant happens to be the
"right" one -- i.e.. the one that has the effect on p which, in the context
of the current environmental circumstance, moves p toward r.>
Dag:
I think most people would intuitively think that their output is their
muscle contraction -- the way they try to move the hand.>>
Rick:
I think the output that matters is the one that influences the variable
under control.
Dag:
Please run your hypercard again and confirm my expectation that r-p is
linearly related to the derivative of o, o'.>>
Rick:
You clever devil. I did it before you asked and your prediction was
confirmed; the derivative is linearly related to p (or r-p with r=0). But
the derivative is not what produces control of p. If you just add the
derivative of o (instead of o) to p each time you get no control.>
Dag:
I agree completely, o versus p is the basis for the behavioral
illusion. And the integration is not the issue you make it.
Correlation of r-p with o is not an issue that is central to the
behavioral illusion. More than that: It has nothing to do with it.
Rick:
In the rubber band version of the behavioral illusion that I suggested
there is no change in r (the reference position of the knot) and yet the
illusion occurs; it is an especially impressive illusion if all you can see
is d and o -- not the controlled variable OR the feedback function (the
rubber bands that connect subject to knot). If we could dip into the brain
and measure variations in p (with r = 0) and o (finger position) in the same
experiment (with changed feedback function rubber bands) we would get the
same apparent change in the relationship between p and o.>
Of course, if you knew that the relationship between p and o was likely to
be misleading, you might look for ways to transform o so that you linearized
the relationship; however, taking the derivative only linearizes if the
output function is a pure integrator.>
Dag:
Please consider applying your understanding of PCT and reconsider your
reply to Michael.
Rick:
This is not a religion group. I don't want to feel like I have to tip toe
around, worrying about whether I am treading on some sacred beliefs. I will
TRY to be more polite -- I'm REALLY not trying to be mean -- but I think it
MUST be OK to suggest that someone MIGHT have an incorrect notion about PCT
(heck, people say I don't know what I'm talking about and nobody jumps in to
say "take back what you said to Rick"). If I say some idea might be wrong
or if someone says something about PCT is wrong) then that should be fine,
ESPECIALLY if people take the trouble to try to show --mathematically,
experimentally or computationally-- why it's wrong. I have no problem with
people saying PCT is wrong; in fact, I wish people would say it and then
show how it's wrong -- not just in words ("I don't like it because...") but
by test ("if you measure this variable while manipulating that one you find
this instead of what PCT says you would find") . Now THAT would be
refreshing. But even verbal disputes are helpful because they help us find
ways to clarify what we are trying to say. If people are personally offended
when someone says "that's wrong" then I think they are looking at this
debate thing from the wrong level of their own perceptual control
hierarchy -- in my humble opinion.>
Dag:
Are you less surprised now?
Rick:
No.
Dag:
I think it is highly worth while to clarify the behavioral illusion. But
let us be very clear about what it is.>>
Rick:
It sounds to me like you are saying that I should be sure that I am right
before I post anything. I can't make that guarantee. This is not a refereed
medium so a lot of mistakes will get posted as "facts". I try to be careful
but basically we're thinking on our feet -- like at a conference. I've made
mistakes, Bill's made mistakes; everyone's made mistakes. Only religions
expect (and require) infallibility; scientific ideas are expected to be
wrong. The hope is that some new ideas and realizations can come from the
uninhibited discussions on this net. When we start requiring respect for
ideas (rather than people) then I'm outta here. Question ideas; love
people.>
Dag Forssell (930918 1125)--
You are thoroughly trained in IV-DV research and use it below to show
something that happens to be true, but has no significance. You draw false
conclusions -- unless I shall prove mistaken,>>
Rick:
I think my mistake was saying that this relationship was an accurate
reflection of the feedback function. This is true for the relationship
between d and o but not necessarily for that between r-p (or -p) and o --
which is what I plotted. The relationship I plotted is a good example of an
"illusory" representation of the real relationship between (r-p) and o --
but it's probably not the inverse of the feedback function.>>
Dag:
Shall we call this the CORRELATION ILLUSION?
Rick:
No. Correlation is just a measure of the degree of (usually) linear
relationship between variables. I was not making inferences about
populations based on the correlation and then applying the results of that
inference to individuals -- THAT's one of the BIG problems with using
statistics in psychology (the other is accepting a low but significant
measure of correlation as reasonable data on which to base models of
individuals).>
Dag:
Your mistaken application of correlation analysis above, in lieu of
consideration of functional relationships, is interpreted by this observer
as an excellent illustration of Bill's thesis.>>
Rick:
I plotted one variable against another. Where was the "mistaken application
of correlation analysis"?>
Dag:
It seems one is much better off without studying statistics.
Rick:
I didn't report ANY statistics; a statistic is a measure of sample
characteristics -- like mean, standard deviation, r, t or F -- that can be
used to make inferences about the corresponding measure of the population
from whence the sample data were drawn). Statistics, per se, are just fine
(or do you object to RMS as a measure of voltage)?>
Dag:
Standing by for your counter-blast in this exchange of disturbances.
Rick:
Boooom. Or Beeeep. Depending on how you hear it.
Talk to you tomorrow, probably.
Love Rick
···
Date: Wed Sep 15, 1993 11:15 am PST
Subject: More misc
Date: Thu Sep 16, 1993 1:21 PM PST
Subject: Rick on control
Date: Sun Sep 19, 1993 8:45 P.M. PST
Subject: Illusions, Confusions, Contusions