[From Bill Powers (2010.02.06.1650 MST)]
Bruce Abbott (2010.02.06.1100 EST) –
BP earlier: I wouldn’t use those
common-language terms or say “because” as you say Skinner would
have done when it’s a non-sequitur. OK, the delivery has followed presses
in the past. What does that have to do with pressing the lever this time?
Could it be that the rat has learned what action to produce in order to
create a given perception?
BA: Do you really mean action?
BP: Let me see. Do I? In this case I think I do, although even though the
kind of action is pretty much determined, the amount needed (which could
include none or some in the opposite direction) is unpredictable. But
perhaps I should have said something more like “which perception to
vary as a means of causing food to be delivered”
But I was focused primarily on the “non-sequitur” part. An
event occurs somewhere and is experienced. What does that have to do with
pressing the lever this time? Trying to work out an answer to that
question is the basic problem for understanding behavior, isn’t it? A
large chunk of an animal’s internal organization lies between the
external event in the past and the present behavior, starting with the
sensory nerves and proceeding through some number of levels of
organization, with memory getting in there, as well as reorganization. To
say merely that the lever is pressed “because” pressing it
produced food in the past packs and hides too much in that one word,
“because.” And anyway, we know that in general the action must
be varied, not repeated, to get the same result as before. Sometimes
repeating the action works, especially in simplified circumstances as in
a laboratory, but most of the time it doesn’t. It’s the surprisingly
appropriate variations we really have to account for, not the repetitions
that work only under special circumstances. If we account for the first,
that accounts for the second, too. The reverse is not true.
BP earlier: That’s how we would
replace the “because” statements in PCT-compatible language.
There’s nothing about past events that can affect present behavior in the
slightest, unless there was some change in the physical system to alter
the relationship of actions to perceptions. Events don’t cause anything;
they just happen.
BA: PCT and reinforcement theory agree that past events affect present
behavior.
BP: Dammit, no they don’t. PCT says that is the wrong interpretation.
It’s true that after a past event, behavior may have changed, but that’s
not an indication that the event affected the behavior. In fact, the past
event may have disturbed something that was under control, and it was
that perturbation that resulted in opposing effects from behavior – but
any event that disturbed the same variable in the same way would have
resulted in the same opposing behavior, because it’s the change in the
controlled variable, not the disturbance, that is sensed and
opposed.
BA: They do so by affecting the
organisms present organization. In PCT, past behavior that has failed to
correct error in a controlled variable fails to slow reorganization. Past
behavior that has succeeded in correcting that error does slow
reorganization, more-or-less freezing in the current, relatively
successful organization.
BP: Wait a minute. Now I realize that when you say “changes
behavior” you don’t mean changing the amount or direction of a
behavior, but altering the manner in which the system acts when the
environment changes in some way – changing to a differnet form of
behavior. You’re talking about reorganization.
But now you seem to be saying that past events affect the organism’s
present organization. I don’t agree with that, either. Past events
stimulate the sensory endings, and direct physical effects like cuts and
bruises aside, that is all they do. They can’t reach inside the brain and
alter the organization, can they? It’s the internal reorganizing
processes that detect the state of some variable and institute
reorganizations as a result of detected error, and that, as you say,
“freeze” the organization that was present when, for any reason
whatsover, the detected error is corrected. If the same error occurs
again because of a change in the environment, reorganization might come
up with a completely different behavior, because all that matters is that
the error gets corrected, not how it gets corrected.
BA: In reinforcement theory,
behavior that produces certain types of events changes the internal
organization of the organism, so that under similar conditions such
behavior is repeated.
BP: Yes, and that is why that theory is wrong. It is not behavior that is
repeated, but certain consequences, regardless of what behavior is
required to produce them.
BA: Either way, our current
organization is a function of certain past events, including perceptible
effects of our own behavior.
BP: I hope you really concentrate on this idea and see what’s wrong with
it. I’ve been trying to get this point across for so many years that I’ve
lost count. Repeating behavior does not, NOT, normally cause the same
consequences to repeat. That happens only under special circumstances.
Normally what has to be done is to vary the behavior in just the way
needed to generate the same consequence as before. And that is what
organisms can do. Only a negative feedback control system can do that. A
stimulus-response system can’t. A system that is reinforced for repeating
a behavior can’t. There is no one “perceptible effect” of any
given behavior. The environment is always there, varying in itself and in
our relationship to it. That means that to produce the same effect twice,
behavior must alter in just the way required to cancel out any changes in
the environmental influence on the effect. If you repeat the behavior,
the environmental influences will cause a different effect to appear.
That isn’t what we observe. We observe that the same effect repeats in
spite of the environmental changes, because the behavior changes just the
way it has to to preserve the same NET effect.
BP earlier: Terms like
expectation are essentially useless to us unless you can express the same
meaning in PCT terms.
BA: Is it useless to ask whether the psychological phenomena such terms
refer to are real? If they are judged to be real, do they not need to be
explained? Is it then useless to ask whether they can be accounted for
within PCT?
BP: All right, tell me what the phenomenon is that is indicated by the
word “expectation”. Expectation is the name of a phenomenon.
What phenomenon? What do you experience that you call
“expecting” something? If you can tell me that, we can look for
a PCT description of it, without having to go through the word
“expectation” on the way.
BP: What is in the model is a
perceptual input function, a comparator with a reference level, and an
output function. In a hierarchical model there are many of these things,
connected in a specific way. Unless you can connect
“expectation” to something in this model, you’d be better off
finding out what the term indicates, and starting at that level. Just
saying “expectation” doesn’t explain anything.
O.K., what you refer to as the model above is only that portion of the
system that does not loop through the environment. Youre free to do
that, of course, but for what purpose?
Because that is the only part that is in the organism. What is outside it
is the ENVIRONMENTAL feedback function. The model of the environment is
easy to construct because we can do that without the organism
present.
To deny me the point I was
making? The models we construct always include the effect of the systems
output on the controlled variable. The model system doesnt need to
expect what effect its behavior will have on the controlled variable,
its behavior just has a certain effect, which was built into the system
by the programmer. The system doesnt need to speculate about how its
behavior might affect a certain perception, and is never surprised on
those occasions when something unexpected happens instead. Ricks models
to date havent needed to include expectation because the expectation
part of the model takes place inside Rick: He expects the models actions
to closely follow the actions of the participant during the experimental
run, and if the results violate that expectation, he revises the model.
The final, successful version has the correct relationships built into
it.
BP: If the model includes reorganization, it doesn’t need to know
anything about the feedback function, not even its sign. Because we start
the weightings at zero and make parameter changes small, starting in the
positive-feedback direction makes the errors increase, so reorganization
comes into play immediately, before fatal runaway can happen. Only
negative feedback will make the errors smaller, and that is what the
final result is. In the ArmControlReorg demo that you’re very familiar
with, it often happens that there are moments of instability, but the
reorganizing process never seems to have any problem getting through
them.
When reorganization is not included in the model, then of course the
designer has to specify the environmental feedback function.
BA previously: So what would
distinguish a system that developed expectations from one that did not?
Perhaps a crucial test would be to observe what the system did if the
expectation were violated.
BP also previously: I wouldn’t start there, because I wouldn’t know how
to tell if there were an expectation at all. Maybe systems don’t ever
develop expectations – how would you know? The only way to find out what
you’re talking about is to settle down and look at something you expect,
and take the experience apart into its components.
BA: You’ve made a prediction,
based on the information you have at hand (including relevant past
experience), about when the train will arrive. I dont know that you have
necessarily imagined the train arriving on time, although of course you
might. Thats one way of expressing the prediction. You might also
express it in words. Its a perception, one way or the other, of a
relationship between the clock and the trains arrival, although not a
controlled perception.
BP: That’s not what I had in mind. How about taking an actual
circumstance you remember, or finding one going on right now, and looking
at the experience you call “expecting” something? What would
you say you are “expecting” right now? How does that activity
appear in your conscious experience? Is the expected thing happening, or
is it being imagined? Maybe you’d have to wait until you’re clearly
expecting something to occur, but the point is to observe what the
phenomenon of expecting is while it’s happening. There’s no need to
theorize about it if you observe it.
BA previously: If you suddenly
reversed the relationship
between mouse and cursor movements, a system without an expectation
would
simply continue to act as it did before, and control would simply
fail.
BP: There’s a partial definition of expectation. What is the expectation,
such that when it’s missing, control would fail?
The recognition by the system that its actions are not having the
required effect on the CV.
BP: All right, I’ll change the question: what form does this
“recognition” take?
BA previously: A system that
“expected” the cursor to move as before (based on previous
experience) would find its expectation violated and presumably take
action to sort out the problem.
BP: So how does it know that the expectation has been violated? Is the
expected result present in some form, being perceived or at least having
an effect? Is the perception of the actual action being compared with the
expected action?
BP: Is that how control systems
change their behavior to counteract errors? If you venture a guess as to
how this expectation thing results in taking action, and what kind of
action it would take, and what the problem is that needs to be sorted
out, you would have a useful model, perhaps, in which the term
expectation wouldn’t even appear.
Youre referring to reorganization, of course. I covered that possibility
below.
BP: No, I thought I was referring to ordinary control.
I’m beginning to suspect that the phrase “change their
behavior” is being used one way by you and a different way by me. I
mean, for example, changing 10 pounds of pull to 15 pounds of pull, or to
-10 pounds of pull meaning 10 pounds of push. I mean changing the same
behavior to different states. You seem, here and there, to mean changing
from one category of behavior, like pushing and pulling, to a different
one like twisting. Is that why my reference appeared to involve
reorganization?
BA previously: Although this
seems like a simple enough test for expectation, one might have
difficulty distinguishing between true expectation and reorganization. As
in the case of expectancy, in reorganization the violation of the usual
relation between mouse movement and cursor movement would bring about a
change in the system’s organization; if successful, reorganization would
restore the negative feedback relation and control over the CV would
recover.
Well, we won’t know if any of that is relevant until you describe the
phenomenon we’re referring to with the word “expect”. To test
whether expectation is occurring you first have to say what phenomenon
you’re talking about. Pretend I’m the man from Mars who has no idea what
the word expect, or any of its synonyms, means. Just look at what you
experience when you’re expecting something, and describe what you find.
I’m not asking for generalizations, just observations.
BP: Did you ask your
participants what thoughts they had when they first encountered the
reversal? How does reorganization target only the system whose control
has broken down? Attention seems to have something to do with that, but
its still an undeveloped aspect of the PCT model.
BP: Since I was one of the participants I don’t have to ask. When I was
trying to prepare for an expected reversal, I controlled much worse, with
lots of false compensations. I did the best when I just carried out the
tracking and reacted calmly but quickly when I found the cursor running
away, reversing my own system. This happened quickly enough and
consistently enough that Rick and I agreed that this must have been an
example of control through changing parameters in a lower system (the
sign of the output function gain, perhaps), rather than changing a
reference signal. I don’t know what I actually did to produce the
compensating internal reversal. A typical reversal episode started with a
very good positive-exponential runaway, followed by an abrupt return of
the tracking error to a low value. The change wasn’t random, so that
would rule out reorganization as we now think of it.
Interestingly enough, even though I am a highly experienced tracker, when
I try very long runs in a simple tracking task, spontaneous reversals
show up, which have the same form as when the program reverses the sign
of the effect of the mouse – only of course, both reversals originated
in me. The spontaneous reversals are disconcerting; it takes a
perceptible time to figure out that something is wrong and fix it, which
implies a higher-order process. And I don’t sense how I fix it. I just
do.
BA previously: Expectation may
be a high-level process involved in planning actions, drawing on
means-end relations learned during previous experience, worked out
logically, or perhaps communicated to us by others. (“You want to
get to the bank? Take Third Street to Maple and turn left.” You then
follow those directions because you expect that they will take you to the
bank.)
BP: That’s more like it. I would say you follow the directions as the
only means you know of getting to the bank, and in the background are
hoping that you’re remembering them right or they were given right. There
might be some sense of expectancy, but I don’t know how that would change
if the destination is a bank or a grocery store. A little more work and
we could just drop the term expectancy, except as a description of a
side-effect of doing all this.
BA: What is the point of having a sense of expectancy, if it plays no
role in behavior?
I didn’t say that. I said that the word expectancy (or the phrase
“sense of expectancy”) refers to a phenomenon and that we need
to examine the phenomenon, not the word, to see what it is and represent
it in PCT terms. If we understand what the phenomenon is and can describe
it directly, we can stop just alluding to it with the word
“expect.” We still haven’t done that.
BP: Did you really say
“planning actions”?
BA: Did you really say actions? (See my comment near the beginning of
this post) (;->
Thats still problematic for me. Following the directions given is an
action of a higher-level system,
Au contraire, following the directions given is a consequence of some as
yet unnamed action. In order for following directions to happen, what
must the behaving system be able to do?
but carrying out that
action is done by setting references for a set of controlled variables,
which ultimately are carried out via a complex set of variable means. In
the past Ive tried to distinguish between behavioral acts, which are
controlled performances, from actions, which are the variable means by
which such acts are produced. Drawing a circle is a behavioral act,
carried out by variable means.
The distinction between “act” and “action” is too
confusing and when it was first invented, had nothing to do with
controlled variables. It’s much clearer to call controlled variable
controlled variables. Look at the mess Skinner made of this idea with his
concept of the operant.
BA previously: Expectation seems
less likely to be involved in habitual activities, although then we do
behave “as if” we had them.
BP: The “as if” part is in the observer’s imagination. Throw it
out.
BA: In the case described, thats my point: its unnecessary. This seems
to be a place where Bruce Gregorys stories we tell ourselves comes
into play.
BP: We wouldn’t need to disbelieve the claim of an expectation if we knew
what the phenomenon was well enough to propose a model of it. All I’m
saying is that we don’t need the “as if” part. If an
expectation exists, it does; if not, it doesn’t. Same idea as testing for
a controlled variable. If the variable proves to be under control, it’s a
controlled variable. Otherwise it isn’t.
BA: But then there are those
other cases, where real expectation may be involved. Lets not throw the
baby out with the bathwater, even if its a rather unwelcome
baby.
BP: If you’ll just try doing what I’m suggesting there will be no need to
throw the poor baby away. To what does the term “real
expectation” refer? I’m not doubting that you can find it, or them,
I’m only saying you haven’t done that yet. When you do it, I think you
will find out what part of PCT it corresponds to. I don’t want to suggest
what part that is.
Best,
Bill P.
P.S. This might also help. When I say A affects or influences B, I am
proposing that there is a direct causal link from A to B. Both terms mean
that there are other influences acting at the same time, so we can’t
predict the outcome without knowing what all the other influences are. If
there are no other influences, we say A determines B. And when we say A
controls B, we mean still something different.