[From Bill Powers (930518.0800 MDT)]
Bob Clark (930515.2115) --
It seems my attempt to condense a complex series of events,
that is familiar to ordinary people, has omitted parts that are
"taken for granted" in ordinary conversation.
This implies that the PCT view is well-known, but that people
simply neglect to mention all its details in ordinary
conversation. I think this is far from true, even though informal
common sense often comes closer to the PCT picture than
scientific views do.
Thus, the process, as I view it, begins with the perceptual
variables involved in riding a bike along a street. This
includes "paying attention" to all the variables required to
maintain the bike upright, moving as desired, etc.
This is a non-explanation; it alludes to the existence of
explanatory terms without actually saying what they are. To say
"the perceptual variables involved in riding a bike along a
street" is only to say that "variables" are "involved" in some
unspecified way with riding a bike. It makes "riding a bike along
a street" into an objective process that is to be explained in
terms of an unspecified set of perceptual variables that is in
some way involved.
I would far prefer to be completely explicit: riding a bike along
a street IS a set of perceptual variables: a visual picture of
the world from the rider's position, the feel of the handlebars
and the feet on the pedals, the sense of effort that holds the
body up and presses down on the pedals, the sense of velocity and
wind, and so forth. The experience of riding a bicycle consists
of perceptions, some of which we control and others of which are
simply consequential. Our view of someone else riding a bicycle
omits practically everything of importance in riding a bicycle
ourselves.
The current perceptions are compared with the memories to
determine whether any form of "corrective action" is
appropriate.
This way of putting it seems, at least to me, to imply an
inappropriate amount of reasoning, and makes the control process
seem needlessly complex. If the current perceptions can be
compared with memories, the comparison process itself immediately
yields the error vector. The control system isn't designed, as I
visualize it, to have any choice about whether acting on the
basis of error would be "appropriate." The discrepancy between
the perception and the reference condition in memory, once
realized as an error signal, is itself what drives the action,
directly.
You're presenting this picture as if a whole rational conscious
person operating strictly at the logical level were in charge of
riding a bicycle. My objection isn't that what you're saying is
wrong; you understand the relationships involved in control just
as well as I do. It's that the viewpoint from which the processes
are described imputes capabilities to the control systems that
are not appropriate to their level. In your terms, you're
treating the lower-level processes as if they were "modes" of
your own sixth order. Your description is appropriate for the way
a person would experience riding a bicycle at the level of
symbol-manipulating logic, but not for the way the systems
actually involved work.
Remember Karl Pribram? We communicated with him several times
during the late 1950s trying to get him interested in feedback
theory. The only result was _Plans and the organization of
behavior_ and the infamous TOTE unit, which seems to be the only
publication of 1960 on this subject that anyone remembers, ours
having been mostly ignored.
The TOTE unit -- Test, Operate, Test, Exit -- represents all
behaviors as if they were little computer programs, no matter
what the level of the behavior. As I acknowledged in BCP, perhaps
the TOTE unit might have some applicability at the program level
of the brain's organization. Perhaps this picture shows how a
person who is being aware from that level exclusively would
experience the operation of a lower-level control process. But it
does not in fact describe the actual processes at other levels --
either lower or higher -- in a way that represents how those
levels really work.
You go on:
Among these are variables related to possible obstacles and
their avoidance. When the current set of perceptual variables
includes some item calling for evaluation, attention is
attracted. That is, related memories come to awareness for
examination.
How does an item "call for evaluation?" Isn't that itself an
evaluation? This passage underlines what I'm saying: that the
viewpoint is too rational and aware for the process being
considered. You make it sound as though there's someone with a
clipboard noting down items in the environment and evaluating
them as possible obstacles, and calling to awareness all the
associated memories so that the subject of whether or not each
item should be avoided can be rationally examined.
If you think of obstacle avoidance as an analogue process, there
is no need to classify items as obstacles (vs. non-obstacles).
The point is not to run into ANYTHING no matter how it is
classified. The reference level for any object approaching from
the frontal zone is some low degree of proximity, and one can
learn a quite automatic control system for keeping all
proximities under control by changing the path of the bicycle.
Such a control system (at least one example of which I include in
the "crowd" program) contains no mechanisms for representing the
situation logically, or reasoning about the appropriateness of
any action. It simply converts a left or right proximity error
directly into a curvature of the path. An observer might
characterize the particular curvature as "appropriate," but the
control system itself contains no such evaluation.
Of course while that is going on, it's possible that the bike
rider, operating consciously from his program level, might be
thinking " Ah, a possible obstable; yes, definitely a threat to
my continued progress. How cleverly and swiftly I have evaluated
it and responded with the appropriate decision to alter my path,
thus avoiding a collision." By the time that thought is finished
the rider is already well past the obstacle. The program level
actually had nothing to do with the action; it was simply
affirming to itself that it was in charge.
I think that you and Avery Andrews both tend to complicate
control processes by casting them in logical terms. If the
control process is itself a logical one, this is not
inappropriate. If the problem is to deliver Mrs. Jones' milk on
alternate days of the week except when the upstairs curtains are
pulled and there's no car in the garage, this logical condition
can be maintained only through use of the milkman's program level
(shades of the distant past). But if the problem is to throw Mrs.
Jones' newspaper onto the front porch while riding past on a
bicycle, the newspaper deliverer can devote the logic level to
something more interesting, like how to get a friend to trade a
particular baseball card.
As you know, Bill, I consider accuracy in language very
important -- but I find I must begin with words/concepts that
are included within the vocabularies of my audience. Perhaps
that comes, in part, from my professorial experience where one
begins with the student's existing information and proceeds to
add to and modify that material. To do this, one must, to some
degree, "enter" the student's universe and work with it.
Yes, I know. Control theory, however, contains concepts which are
not already in the vocabularies of many audiences, and the
existing words usually mean something that has to be overcome
before the wanted meaning can be communicated. By trying to make
PCT concepts seem TOO familiar, in the hope of getting a friendly
reaction from the audience, one can end up convincing them only
that there's nothing new in it.
ยทยทยท
------------------------
RE: stimulus-response.
But there are other significant viewpoints: the experimenter
applies a "disturbance" to the organism, but he thinks of of it
as a "stimulus." It was his decision, and his act, intended to
control some set of his internal perceptual variables. To him,
it is ("merely") a matter of terminology. On the short time
scale he (still the experimenter) has chosen, the subject's ref
level has not changed. Thus the experimenter observes the
subject's action to control his (the subject's) perceptual
variable, and calls it a "response" instead of "preventing the
controlled variable from changing.
The problem with this approach is that it assumes that the
controlled variable is just as evident to the observer as the
disturbance and the response are. This is seldom true. The
observer alters something in the environment of the actor, and
the actor produces some behavior or change in behavior. To the
observer, it seems that the environmental change constituted a
stimulus, to which the actor responded. There is nothing wrong
with observing that this S-R relationship holds true. But the
impression is that the actor must have been sensing the disturbed
aspect of the environment, while in fact the actor may have been
(and probably still is) sensing and controlling something quite
different from the variable that the observer thinks of as the
stimulus.
Neither is it self-evident that by acting, the actor has
_prevented_ the changed aspect of the environment from materially
altering the thing actually being sensed and controlled. From the
standpoint of the observer, the action has been explained as a
consequence of the antecedent manipulation of the environment.
But the same test for statistical significance that justifies
this conclusion efficiently discards the actual controlled
variable (if it is even noticed), because the controlled variable
does NOT show a significant correlation with either the
disturbance or the action.
... in my 930513 Post, I suggested an alternative:
"If I can't reach the shaker, 'Please pass the salt.'"
As you note, this can easily be interpreted as a stimulus that
results in the response of passing the salt. However, the naive
observer is not likely to understand what has been maintained the
same when the Companion follows the verbal request with the
physical action of passing the salt. The Companion has maintained
the appearance of compliance or good manners, an appearance that
the Companion could bring to quite a different state by replying,
"Get it yourself, you lazy bum."
The S-R viewpoint considers only the large visible phenomena that
show relationships to each other. But at the center of these
phenomena, between the disturbance of the local environment and
the action that ensues, there is a controlled variable that is
not evident to the S-R observer. So the S-R description, while
pragmatically useful as you say, is not just an alternative to
the PCT view; it is incorrect. It proposes a causal role for a
disturbance and omits the central fact of the whole situation,
the controlled variable that is the actual stimulus.
And the controlled variable is not the central variable just from
the Engineer's point of view. It is the central variable from the
actor's point of view. When the neurologist jabs the actor in the
elbow-joint with a pin, the neurologist notes that the correct
response to this stimulus has occurred. The jabbee, however,
notices the reason for this response: it hurts like hell. The
controlled variable is pain, with a reference level of zero. The
only reason for jerking away is to stop the pain, no matter what
caused it. But the neurologist knows nothing of that. The pain is
not observable.
The S-R view, by misidentifying the stimulus and omitting all
mention of control (by the recipient of the stimulus), is led to
an incorrect interpretation of the situation that will break down
as soon as the situation changes enough. The S-R interpreter
notes that by opening a window, the observer can cause the
subject to execute a behavior from a small list: move to another
chair, put on a sweater or a coat, get up and stand by the
fireplace, unfold a blanket and wrap up in it, and so forth.
Having verified this antecedent-consequent relationship, the
observer then thinks he knows a fact about the subject. Six
months later, the observer tries to demonstrate this fact to a
skeptical audience: flinging the window open, the observer
indicates the subject and waits confidently for one of the list
of actions to occur. The subject says, "Ah, that's better. It was
getting too hot in here."
However, the observer can salvage something from the wreckage by
doing the experiment over and over for whole year. It can be
shown with high statistical reliability that the subject TENDS to
put on a sweater, move toward the fireplace, etc., when window is
opened. In this statistical approach, the results of a single
trial are not significant; only the means matter. On at least
half of the trials in this year-long experiment, the expected
response to the stimulus occurs. So it must be a real effect. On
the average, the subject puts a sweater halfway on, moves halfway
toward the fireplace, and so forth.
When dealing with "real people" in ordinary situations, I not
only avoid detailed "engineering descriptions" but also S-R
descriptions. Instead I tend to use "Cause/Effect" language, or
whatever language my friend prefers. Otherwise, my audience
disappears with incredible speed, approaching Warp Ten at
times!
If you tell people only what they want to hear, they will hear
nothing new. In that case, why bother at all?
--------------------------------------------------------------
Greg Williams (930517) --
Free will:
I think you asked earlier whether free will connected with HPCT,
or should be considered as an unconnected phenomenon. I think the
latter is right. It's possible to do something just to see what
will happen. This would seem to have something to do with
reorganization, but I don't know what. It's just a phenomenon
that we have to stick on the shelf along with awareness until
someone has a useful idea about it. We can still count these
things as real aspects of experience, but for the time being they
remain unexplained. Like you, I'm unimpressed by the stories
people have told so far in attempting to account for them.
I think it is important ETHICALLY to have a well-tested theory
that I am not the sole creator of the universe I perceive. Some
who believe THEY create it ALL will have no qualms about
attempting to DESTROY it all.
I'm not sure I see what ethics has to do with it. Is there an
objective standard for ethics that says destroying it all is _a
priori_ unethical?
I prefer just to go with the evidence and make sense of it the
best way I can. The best model I can think of that represents
both experience and physical models is the one that says we
experience neural signals, but that the neural signals come from
outside the brain that we live in. Without assuming a regular
external world, I can't see any reason for the regularities we
experience and I can't explain why we have to learn how to
control our perceptions.
I've never seen any arguments for solipsism that go much beyond
just blurting out that we're making it all up as we go and nobody
can prove different, so there. I wouldn't accept arguments of
that quality in any other field; why should I make an exception
for philosophy?
--------------------------------------------------------------
Best to all,
Bill P.