non-revolutionary two-step

[From: Bruce Nevin (Wed 93016 12:54:24)]

It seems to me that problems and solutions are theory-dependent.

What constitutes a "problem" depends one's ignorance that one is
aware of. Ignorance of which one is unaware is the blissful kind
(even if error results that is not blissful--no connection to the
ignorance can be made, and pain is often misattributed).

Awareness of ignorance has to do with gaps in the constructs
(themselves perceptions, of course) by which we try to organize
our perceptions (mostly remembered and imagined perceptions).

This is how I put it in (Fri 921218 13:24:41):

ยทยทยท

-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-

The perceptions to which we pay attention are more clearly
defined and longer lasting in memory than those we disregard. In
this way, we create reference levels for selected perceptions
(selected by having attended to them). When we use these in
imagination, unforeseen perceptions come up as ramifications and
consequences. By reasoning about these, we develop/impose order
and structure in them. In these perceptions of order and
structure there are gaps. A gap of this sort provides a context
for recognizing a perception ("real" or imagined) for what it
"really is," and by that I mean perceiving it as a filler of that
gap. (Related perhaps to Gibson's notion of affordances.)
This we call intuition or insight.

The things we have conscious control over are: what we pay
attention to, how well we pay attention to them, and how well we
reason about them. The rest is on automatic pilot.

The part for which we have some conscious responsibility includes
how we interpret perceptions as to what they "really are" (what
they constitute at higher levels). And of course how attached we
are to our conclusions as imagined perceptions. We know that one
false premise puts the conclusions at random, but we often forget
or ignore this in practice. When we use as premises conclusions taken
from prior lines of reasoning from premises based on authority,
etc., the house of cards looks pretty shaky.

-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-

If we use the term "theory" for constructs of that sort, then
"problems" are always perceived relative to one's theory.

If this is true, then how can Perceptual Control Theory address a
problem as identified by mainstream psychologists (sociologists,
control theorists, etc.) within their respective theories? PCT
can address only problems as identified within PCT.

That way of putting it is fairly succinct, but implies that the
difficulty is due to a deficiency of PCT (or of the would-be
problem-solver, whatever the pair of theories). Rather,
one finds it very hard trying to understand the PCT treatment of
the problem simultaneously in terms of the other theory (as
"problem") and in terms of PCT (as "solution").

This leads to the non-revolutionary two-step:

        1. Problem in other-theory terms.
        2. Solution in PCT terms.
        1'. Non-solution in other-theory terms.

We have seen how it does not work to frame the solution in terms
that are compatible with the other theory. The whole spectrum of
predictable responses ensues, from "So what?" to "Old hat!" to
"Huh?"

Perhaps another step is needed in the process:

        1. Problem in other-theory terms.
        2. Same problem in PCT terms.
        3. Solution in PCT terms.
        1'. Contrast with non-solution in other-theory terms.

In many cases, step (2) is to show that there is no problem in
PCT terms, and then steps (3) and (1') show why this is really OK
as a resolution.

It might be useful actually to show how statistical results of
the sort typically found in the literature follow from a
statistical treatment of models of control systems, identically
to the same statistical treatment of the modelled control systems
-- but how the models and modelling data are much more
interesting and useful for the purposes to which the statistical
results are typically (and inappropriately) applied.

In other cases, a real re-framing of the problem may be possible.
Then (1') becomes crucial for demonstrating the legitimacy of the
re-framing and the solution.

        Bruce
        bn@bbn.com

[Martin Taylor 930106 15:30]
(Bruce Nevin 930106 12:54)

It seems to me that problems and solutions are theory-dependent.

Yes; I think that is a generally agreed position.

What constitutes a "problem" depends one's ignorance that one is
aware of. Ignorance of which one is unaware is the blissful kind
(even if error results that is not blissful--no connection to the
ignorance can be made, and pain is often misattributed).

This latter is central to the Powers form of reorganization. One
determines that there IS error in something that matters (i.e. an
intrinsic variable), but one knows not what to do about it. So a
random change somewhere in the behaving hierarchy alters some aspect
of control, and maybe that fixes the problem. This method should always
work, if slowly, and it demands no apparent connection between the place
where uncorrectable error is noted and the place in which a modification
fixes the error in the intrinsic variable.

When there is a known "problem"--ignorance that one is aware of--my local
type of reorganization may be more appropriate. An ECS in the hierarchy
cannot control its error, and some local modification may be appropriate
to fix the problem. Perhaps the Perceptual Input Function (PIF) is
modified (seeing the situation differently), perhaps some output connection
has its sign switched (doing something different about it), perhaps
a new output link is made or one eliminated. These changes are local to
the ECS, based on a local determination that it is not in good control.

Awareness of ignorance has to do with gaps in the constructs
(themselves perceptions, of course) by which we try to organize
our perceptions (mostly remembered and imagined perceptions).

I would call these constructs PIFs rather than perceptions. The result
of the PIF depends on the incoming sensory data, and is a perception.

This is how I put it in (Fri 921218 13:24:41):

-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-=+=-

The perceptions to which we pay attention are more clearly
defined and longer lasting in memory than those we disregard.

There is a phenomenon psychologists call "the Zeigarnik effect" (or
used to, when I was in graduate school). What is remembered best is
an unfinished task. Now knowing about PCT, I would translate that to
say that what is remembered is a perception that continues to differ
from its reference. I'm not clear whether this has anything to do with
"attention," or even what "attention" means within PCT. I have my own
notion of what it means, but that is within the construct of the alerting
system, which is (so far) not accepted as part of orthodox PCT.

As a side-note, I don't like the use of the term "aware" in all of this,
but I can't really think of a better term. "Aware" invokes the notion
of consciousness, which isn't relevant to this discussion. What is
necessary is that appropriate information about the existence of an
uncorrectable error be available at suitable points in the main hierarchy
or in the reorganizing system. To say that the person is "aware" of this
information is to ask the homunculus to get working. And that's someone
I'd like to exclude.

But subjectively, much of what you say, using "aware" and similar terms,
is true of conscious experience (mine, anyway). How that aspect works is
part of the mystery of teaching as opposed to learning. In your self-quoted
passage, you refer to this, and I'm not trying to criticize that part of your
posting. I'm only noting that the issue exists without bringing consciousness
into it.

Martin

[From Rick Marken (930106.1500)]

Bruce Nevin (Wed 93016 12:54:24) --

It seems to me that problems and solutions are theory-dependent.

Yes.

This leads to the non-revolutionary two-step:

       1. Problem in other-theory terms.
      2. Solution in PCT terms.
      1'. Non-solution in other-theory terms.

I love it! And a one and a ...

Perhaps another step is needed in the process:

       1. Problem in other-theory terms.
       2. Same problem in PCT terms.
       3. Solution in PCT terms.
      1'. Contrast with non-solution in other-theory terms.

?Very nice Bruce. I hope we all can pick up your beat so that
PCT can waltz past the opposition.

It might be useful actually to show how statistical results of
the sort typically found in the literature follow from a
statistical treatment of models of control systems

Bill Powers has a wonderful demo of just this. It is described
in his chapter in the American Behavioral Scientist "Purposeful
Behavior" issue; a must read for those who still harbor the
hope of understanding control systems using statistical methods.

Best

Rick