NetNews Usenet Archive 1992 #26/NN_1992_26.iso/spool/bit/listserv/csgl/1440

Text File | 1992-11-10 | 8.0 KB | 167 lines

Comments: Gated by [e-mail]
Path: sparky!uunet!stanford.edu!bcm!convex!darwin.sura.net!paladin.american.edu!auvm!VAXF.COLORADO.EDU!POWERS_W
X-Envelope-to: [e-mail]
X-VMS-To: @CSG
MIME-version: 1.0
Content-transfer-encoding: 7BIT
Message-ID: <[e-mail]>
Newsgroups: bit.listserv.csg-l
Date:         Tue, 10 Nov 1992 15:12:07 -0700
Sender:       "Control Systems Group Network (CSGnet)" <[e-mail]>
From:         "William T. Powers" <[e-mail]>
Subject:      Hamming distance; belief;association;fundamentals
Lines: 152

[From Bill Powers (921110.1430)]

John Gabriel (921110) --

Let me echo Rick Marken's and Bruce Nevin's comments on Hamming
distance.

I think your concept will be most useful in a model of the way an
individual compares the meaning of a communication that is under
construction with the intended meaning. This has come up repeatedly;
Avery Andrews is particular has commented several times that
implementing my idea of closed-loop sentence construction is difficult
because of not knowing how to judge an error in meaning, and how to
transform the error into an appropriate kind of change in the sentence
or utterance (Penni has made me self-conscious about saying "sentence").

But I agree that it's not very useful as a judgment of the difference
between two people's perceptions, either of sentences or of meanings. A
comparison has to be made inside one of the systems. Two people may
perceive different differences between their perceptions (ugh, what a
sentence). We can't use that old disembodied omniscient observer ploy.
The FACT that two people perceive the same thing is unknowable unless
someone knows it.
----------------------------------------------------------------
Martin Taylor (921110.1415) --

Rick: >Belief is the enemy of control.

Martin: >I would have thought belief to be the essence of control.
>Belief, to me, is a short word for "perception of the current state
>of whatever is being controlled for by this ECS."

You're using belief in the sense of perception; Rick was using it in the
sense of a reference signal. I think that Rick's use has to be qualified
to read

Unchangeable belief is the enemy of control.

A fixed reference signal anywhere in the hierarchy is bad news for the
organism. If it's not at the highest level, it makes the associated
control system unusable by higher systems -- they can't vary the
reference signal when they need to.

I think system concepts -- or whatever the highest level is when
posterity reads this -- have to be a special case. The reference levels
at the top level can't be varied freely by a higher system because
there's no higher system. It's my impression that system concepts are
extremely hard to change. Hard to learn, and hard to change. They come
along, they grab you, and then they have you. I don't know how that
works. But you can see that it DOES work just by looking at Islam, or
any other system concept shared by a lot of people (or so they believe).
I don't think anyone can say "I'm bored, I think I'll change a system
concept." I don't think anyone changes a system concept without turmoil,
pain, and confusion, and the result is not likely to be predictable.
This is why PCT is slow to be appreciated as a new system concept. Mucheasier to
 see it as a couple of new principles, or a set of clever
calculations.
-----------------------------------------------------------------
Bruce: > In PCT, meanings are remembered or imagined
>perceptions associated with a given perception.

MArtin:>Sorry? What construct in PCT is an association?

There isn't one. We're fooling around with the phenomenon of
association, which I think we have to admit exists, trying to find a
good place in the model for it. Your suggestion that it could be part of
perceptual functions is good. It could also form a link BETWEEN
perceptual functions, which is pretty much what Bruce is thinking of, I
think. I agree that there may be other more PCTish ideas that can
explain SOME phenomena of association. But I don't see anything yet that
can explain my being reminded of a Volkswagen by a certain engine sound.
Nothing, that is, but content addressing of memory. I think of
association as a phenomenon waiting to be put in the model.

There could even be a specific kind of perceptual transformation that
involves inputs of one kind and a perception of another kind, where the
link is memory association and not an ordinary computing function. Keep
thinking of ideas. Maybe the right one will show up.
-------------------------------------------------------------------
Oded Maler (921110) --

RE: Fundamentals of PCT

>I wonder how much of the above holds if you replace "organism"
>by ECS.

Some but not all, because I think that section was concerned with social
principles. In social situations, there is no superordinate control
system that senses and sets reference signals for lower level control
systems. Inside one organism, a higher-level system is responsible for
conflict; it is specifying incompatible reference signals for ECS inside
one person. As a result, the conflict can be resolved by reorganizing
the higher-level system. This isn't possible in a social conflict, where
there is no higher-order system in charge of both persons.

>Can you elaborate a little on the coexistence of "close contact" in
>one hand "control of linearily independent variables" on the other.
>I can see that if 2 systems perceive the same point in the plane,
>one controlling for the x dimension and one for the y, then they
>can achieve their goals simultaneously, but in what sense are they
>then in "close contact"?

When the controlled variables are really orthogonal, close contact
doesn't exist, of course, even though the two systems may be controlling
something associated with the same object. But that's an end-state;
normally we don't start out interacting with others orthogonally, and
even after learning we seldom achieve perfect orthogonality. I meant the
term "close contact" to be a contrast with the situation where
completely different objects are involved, so no attention has to be
paid to orthogonality.
Put it this way: "close contact" implies the likelihood of conflict,
which can be resolved by achieving orthogonality of controlled
variables.

It's always nice to see evidence that someone understands what I'm
talking about even when I don't say it very well.
--------------------------------------------------------------------
Bruce Nevin (92110.1259) --

>It seems clear that we control for a given traversal of "adjacent"
>perceptions having been talked about: instead of repeating the
>traversal in words, we use one of the reduced forms

Sounds good. Do you agree that we also control for unwanted meanings?
I.e., some traversals lead to words that have unwanted meanings that
stick out enough to cause us to switch to a different traversal, or to
insert a specific denial of the unwanted meaning, or to loop back and
repeat part of the traversal in different terms.

>It seems to me that we use the socially learned structure of
>language as an framework for organizing how our attention traverses
>our perceptions (the combination of perceptions from the
>environment with perceptions from memory and imagination).

Yes, I agree that when we think in words (not always the case) we use
the socially (well, individually) learned structure as a way of guiding
the traversal of meanings -- for creating a scenario to go with the
words. This also applies to what happens when we listen to another
person speaking, or read, doesn't it? The different in listening or
reading is that we have to supply in imagination a lot of the details
that the words don't evoke. We don't often misunderstand ourselves, but
it's easy to do when listening to others.

>Failure of social agreements is probably occasion for intrinsic >error
in mammals (possibly in other creatures as well).

I can think of lots of ways in which such failure would lead to
intrinsic error without the failure itself being an intrinsic error.
-------------------------------------------------------------------
Best to all,

Bill P.