Behavioral Illusions

[FredNickols]

[From Fred Nickols (2018.03.21.1214)]

I think what you propose, Rick, is better than telling them they are succumbing to an illusion. However, telling them they are paying attention to the wrong things isn’t much better. In the first instance, they’re being fooled; in the second, they are misguided. Neither is likely to sit well. So how about saying simply that there’s more there than meets the eye. I think they’ll be more responsive to the notion that they’re missing something than they will to being told they’re wrong.

Fred Nickols

[Rick Marken 2018-03-20_18:33:02]

[FWN] SNIP

RM: So what do you think? I propose that we stop telling behavioral scientists that they are succumbing to an illusion and just say they are paying attention the the wrong things: irrelevant (but compelling) side effects of control rather than the variables being controlled.

Best

Rick

···

–

Richard S. Marken

Regards,

Fred Nickols

Managing Partner

Distance Consulting LLC

“Assistance at a Distance”

[Marken_Richard_S1]

[From Rick Marken (930413.1100)]

Dan Miller (930412.1330) --

PCT is invaluable to me in the classroom and in my research.

I think that's great. As Tom Bourbon (930412.1158 CDT) noted,
my list of "non-uses" of PCT was a tad facetious. The basic
idea was that PCT is not useful to those who want to study
the illusory phenomena that are considered to be important in
most departments of behavioral science.

Having said that, I suppose now is as good a time as any to
start explaining what these illusions are and why I consider
them "illusions". These illusions were described in my widely
distributed and rarely read "Blind men and the elephant" paper.
There are three of them -- which I call the S-R, reinforcement and
cognitive illusions. These are "illusions" in the same way that
the well-known perceptual illusions are illusions. In each case, there
is a "real" phenomenon to be experienced; but what is "really" going
on is not the same as the experience. For, example, in the well
known Mueller-Lyer illusion, two parallel lines with "fins" look to
be of unequal length even though we know, from measurement, that
they are exactly the same length. The phenomenon (appearance of
unequal line lengths) is real -- there is no denying that one line looks
longer than the other -- but, based on other considerations (measure-
ment) we know that the line lengths are equal.

The same kind of mismatch between experience and knowledge
occurs in the "behavioral" illusions described in the "Blind men"
paper. In the SR illusion it looks like a stimulus input causes response
outputs. The illusion is very common -- we see pedestrians "dodge"
cars and CSG-Net posters respond to an insulting posts. It looks like cars
and the insulting posts are stimuli that cause responses (dodging, snappy
replies). This is the illusion we see in tracking tasks -- we are seeing
the relationship between d and o. When d is visible (pursuit tracking)
it seems like o (handle movement) is caused by d (target movement).
When d is invisible (compensatory tracking) it is assumed that the
difference between cursor and target (i) causes output. PCT shows that
the apparent causal relationship between d and o (or between i and o)
is not what it appears to be. It appears that d (or i) cause o via the
organism. PCT shows that the relationship between o and d depends on
the feedback function, not on the organism at all; and the relationship
between i and o (where i is an independent variable) does not exist at
all.

The SR illusion is easy to demonstrate -- just ask a person to track your
moving finger with theirs. It looks like your finger is causing the
person to make their movements, doesn't it? One demonstration that
what you are seeing does not correspond to what is actually happening
can be found in Mind Readings (p. 61 -66). This demonstration is
equivalent to putting a ruler next to the lines in the Muller-Lyer
illusion.

In the reinforcement illusion it looks like the consequence of action
controls, selects or strengthens the action that produced it. This illusion
is also very common -- we see kids quickly coming back to ask for
more candy after they have been given one or we see cats whining at
the food bowl after getting fed earlier. In each case, a desirable con-
sequence (candy, food) seems to have increased the strength of the action
(asking, whining) that produced it. PCT shows that this apparent
relationship between consequence (reinforcement) and action is not what
it seems; it looks like reinforcement selects action; in fact, the action
controls some aspect of the reinforcement (ie. its rate of occurrence,
the amount/time period, etc). The relationship between action and
reinforcement exists because the organism is controlling for the
consequence.

The reinforcement illusion is quite compelling -- that's why people have
always resorted to rewards and punishments to control behavior. And it
works as long as the organism maintains its reference for the reinforcement
and discovers no other way of controlling for the reinforcer. But the
important point is that the appearance of selection by consequences is
just a compelling illusion. One demonstration of this illusion is given
in Mind Readings (p. 79 -85). This demonstration shows that the
consequences of action can be controlled even though they are random. If
the consequences actually controlled the actions, the resulting actions
would be random; they are not.

In the cognitive illusion it looks like behavior is generated by internal
plans for action. Most everyday behavior looks like this -- possibly
accounting for the dominance of "cognitive" type theories in the
behavioral sciences. Examples of the cognitive illusion are almost too
easy to find -- just look around; people are generating behavioral outputs
all over the place. A particularly germane example (considering the
discussions of social phenomena) is dancing (an apt analogy used by
Tom Bourbon in his paper on interpersonal control). When you watch Fred
and Ginger, what you see appears to be rather precisely coordinated,
programmed outputs. The dancers are producing the same outputs in
synchrony. It looks like both dancers have learned the same program
(Ginger's program a mirror of Fred's) and, at the down beat, start
outputting the steps of these programs in perfect synchrony. The
illusion is that what you are seeing is the end result of commands
(cognitions) generated in the heads of each dancer. According to PCT,
what you are actually seeing is the observable side effects of an
exceptionally skillful process of controlling perceptual inputs.
The end result of the cognitive commands are not outputs; they are
perceptions.

The fact that perception, not output, is being controlled in the cognitive
illusion is demonstrated in the "mirror image reversal" tracking task
described on pp 121-129 of Mind Readings. At a signal, the subject
changes from mirror to non-mirror "pursuit" tracking or vice versa. It
takes about 400 msec for the change to occur; during that 400 msec it
seems like the cursor movement is "programmed" output. In fact, the
cursor movement during this period is a controlled input because the
course of the cursor movement is defended against disturbance. The cursor
movements are not a set of outputs generated by an unfinshed "DO" loop--
they are intended perceptual results of action that are made to
occur, despite disturbance.

The cognitive illusion results from the fact that disturbances and the
variations in outputs that compensate for them are often invisible.
When Fred and Ginger dance we can't see the slight variations in the
environment that require slightly different adjustments from each dancer
so that the same (perceptual) result is achieved.

In summary, the three behavioral illusions are real phenomena -- they are
what we see happening every day. They are the basis of all conventional
approaches to understanding behavior. They are illusions only in the sense
that what appear to be the reasons for what we see (SR causality, selection
by consequences, programmed output) are not. According to PCT, the reason
we see SR, reinforcement and cognitive behavior is because people control
their own perceptions. PCT shows that we can understand ALL behavior as
control of perception. The way to achieve this understanding is to test
for controlled (perceptual) variables Thus, whenever you see SR,
reinforcement, or cognitive behavior, instead of asking "what stimulus
causes the response?" or "what reinforcer strenthens behavior?" or
"what program is being used to generate the output?" ask "what perception(s)
is (are) being controlled?".

Next episode: The Test

Best

Rick

[Richard_Marken]

[From Rick Marken (950701.1200)]

Bill Powers (950630.?) --

I think that Bruce is talking about a somewhat more complex theory
than that simple version.

Yes. It's the version that can explain everything. That's why I have decided
to take over the role of reinforcement theorist and explain what their theory
actually says;-)

If we can demonstrate by using disturbances that the SAME
consequence is associated with the entire possible range of behaviors,
then this consequence would have to be equally reinforcing of all the
behaviors that bring it about, and therefore incapable of reinforcing the
particular behavior needed to counteract a particular disturbance.

I am working on such a demo now; I should have it ready soon. My prediction,
however, is that reinforcement theorists will have no problem accounting for
the results, even if they can't account for them. Reinforcement theorists,
like OJ's defense team, can't lose.

Me:

The causal relationship between o and d is (approximately) the inverse
of the feedback function, h, relating o to q; it does not reflect a direct
connection between o and d and it does not reveal anything about the
nature of the organism that is controlling a perception of q (see Powers,
Science, 1973).

Martin Taylor (950630 17:30) --

That is very close to asserting that you could not, by observation
of behaviour, distinguish a sheep from a shepherd, since all you could
see would be the differences between their two environments.

You missed my point completely, Martin. You seem to think I was saying that one
cannot learn anything about the internal organization of a control system by
observing its behavior. In fact, you can learn plenty about the internal
organization of a control system by observing its behavior -- but you have to
know what behavior to observe.

Conventional psychological research is based on the idea that you can learn
about the internal organization of a living organisms by observing the nature
of the relationship between o (responses; dependent variable) and d (stimuli;
independent variable). If, however, organisms are perceptual control systems,
then the relationship between o and d reveals only the external organization of
the relationship between the organism and its environmentl; it reveals nothing
about the internal organization of the organism.

I think it is this fact about the nature of living control systems that causes
the most problems for conventional psychologists. Psychological research is
based on a causal model of organisms; it assumes that o = f(d). PCT shows that
the causal relationship between o and d reveals nothing at all about f. The
implication is clear, even if unspoken: all research results based on the
assumption that o = f(d) tell us nothing about the nature of organisms. Since
all psychological research is based on this assumption, it is not surprising
that PCT has made no inroads in conventional psychology.

Once you realize that organisms are perceptual control systems you simply stop
trying to understand their internal organization using conventional research
methods. But you certainly don't stop trying to understand their internal
organization; you just start doing it using different methods. The most
important method is, of course, The Test; the goal of The Test is to reveal the
nature of the environmental quantities that an organism controls. Once you
discover a controlled quantity, you are severly constrained in the nature of
the internal organization that can perceive and control that quantity.

The problem is in the juxtaposition of "(approximately)" and "does not
reveal anything." If the approximation is inexact, then the revelation is
not totally denied. And no living control system controls perfectly.

True but irrelevant. The relationship between o and d reveals nothing about
the internal organization of a control system even when that system controls
imperferctly. Imperfect control just means that the relationship between
o and d gives an imperfect picture of h, the feedback function; it doesn't
mean that it gives a better picture of f, the internal organization of the
organism.

I think that, ultimately, rejection of PCT (or failure to see it as anything
other than an "alternative theory of behavior") by conventional psychologists
is based on failure to grasp (or accept) the fact that, for a control system,
o = -1/h(d), NOT o = f(d).

It looks like o = f(d); just take a swing at a person's face, give them $100
every time the turn on a light, say they are stupid, etc. Each of these stimuli
(d) will be associated with an observable response (o). It is reasonable to
assume (as psychologists have assumed since the time of Fechner) that the
relationship between o and d reveals something about the organism. PCT shows
that this is not the case; the relationship between o and d only tells us about
the environmental link between the organism and the environmetal quantities it
controls. This is called the Behavioral Illusion; it is probably the most
important (and disturbing, to conventional psycholoigists) of the revelation
of PCT.

In his talk in San Francisco, Gary Cziko presented an excellent demonstration
(using a variant of the rubber band demo) of the "behavioral illusion"; the
fact that the relationship between o and d reveals nothing about the organism
but plenty about the environment in which it does its controlling. Gary's
innocent little demo is the litmas test for "getting" PCT; a conventional
psychologist who see Gary's demo and still thinks that it is possible to
learn about the nature of living organisms by studying functional
relationships between independent (d) and dependent (o) variables does not
"get it". They (understandably) prefer to labor in the grip of the Behavioral
Illusion rather than desert their peers and start over on the road to
understanding living control systems.

Best

Rick

[_Martin_Taylor]

[Martin Taylor 950704 17:10]

Rick Marken (950701.1200)

You missed my point completely, Martin.

I did, indeed. I see a litle better what you were getting at, now. But
I'm afraid I still don't see your point very clearly.

But you certainly don't stop trying to understand their internal
organization; you just start doing it using different methods. The most
important method is, of course, The Test; the goal of The Test is to reveal the
nature of the environmental quantities that an organism controls. Once you
discover a controlled quantity, you are severly constrained in the nature of
the internal organization that can perceive and control that quantity.

There is a constraint on the perceptual functions involved, for it is those
functions that define the stable complex in the environment that the
experimenter can more or less see using the Test. But I don't get the
connection with "and control that quantity" in the absence of control
imprecision. "What" is not "How."

"How" is the internal organization, whereas the Test tells you "What." If
control is very good, I don't see how you can access the internal organization,
other than the relationships among controlled perceptions, no matter what
you discover through the Test.

Or am I once again being obtuse?

Martin

[Bill_Leach1]

<[Bill Leach 950704.20:51 U.S. Eastern Time Zone]

[Martin Taylor 950704 17:10]

There is a constraint on the perceptual functions involved, for it is
those functions that define the stable complex in the environment that
the experimenter can more or less see using the Test. But I don't get
the connection with "and control that quantity" in the absence of
control imprecision. "What" is not "How."

"How" is the internal organization, whereas the Test tells you "What."
If control is very good, I don't see how you can access the internal
organization, other than the relationships among controlled perceptions,
no matter what you discover through the Test.

Or am I once again being obtuse?

Well, not to me I believe.

It is, to a large measure, system dynamics and biology that will be
needed for much of the "internal organization" information. The
"reaction time" demonstrations are an excellent example of course
determination of "levels" of control.

I would say that PCT is currently at the point where it is "trying" to
determine controlled perceptions, show that they really are controlled
against disturbance and then create a closed loop negative feedback model
that correctly emulates the living organism behaviour with respect to that
controlled perception. This, of course, necessitates quantitative
measurement of behaviour. The difficulty that PCT faces is similar to
that of EABers. In some instances worse and in others maybe not as bad.

A PCTer is however, faced with the necessity to acknowledge that anytime
the researcher overwhelms the control system(s) that the specific
characteristics of the subject may change completely and for purposes of
the experiment -- permanently.

The HPCT hypothesis asserts that it is control from top to bottom but
only the first few levels have strong scientific support.

I think that Rick is acknowledging that we have a great deal more
experimental work to do before we will be able to "generate" much more
in the way of structure theory.

Bill P. is, doubtless to me, the best there is at analyzing for possible
controlled perceptions. I know that many of the things that I thought
were (and no doubt others that I still think are) controlled perceptions
just can not be such. I think that anyone that becomes convinced that
PCT is the correct theory for behaviour will spend the rest of their
lives becoming increasingly astonished at just how little they understand
of their own behaviour much less anyone or anything else's.

We have all spent a lifetime creating belief systems that while generally
"workable" are probably mostly false in their detail.

The enormous difficulties associated with attempting to identify even a
significant number of the controlled perceptions for a simple "task" is
a daunting proposition.

Rick's comment that "you are severly constrained in the nature of the
internal organization that can perceive and control that quantity." is I
think, a gross oversimplification.

In the first place, identified "perceptions" are themselves almost
infinitely complex in the nature of the components signals that
eventually are THE perception. Then there is the matter that the
perception of the perception is not THE perception even in the subject
much less the researcher...

Secondly, we can rest in reasonable comfort in the knowledge that the
organization of the working PCT models is at best a mild approximation
of the complexity of the real thing, yet the performance is impressive.

We already have some experience with the concept that control loops that
disturb each other can and do operate quite well in the physical control
realm. I don't believe that there is a similar experience base for
control loops that might be involved in concepts similar to what we
typically call "decision making" and "cognitive processes".

Just the difficulty in attempting to analyze and discuss Han's model over
the last few months demonstrates amply that we are an awful long way from
applying much in the way of "constraints" to the organization.

-bill