[From Bill Powers (2007.03.12.0555 MST)]
[Note to Ted Cloak – I’m replying via CSGnet, and the above salutation
is our conventional way of showing who the post is from at the beginning.
This is in answer to your query:
Would you mind if we
share this conversation (I hope it will continue) with the rest of the
Group?
Ted Cloak(2007.03.11.xxxx) –
[And that is how we show to whom we are replying – normally. it’s just a
copy of the sender’s opening time-date stamp. I will include below the
entire text of our previous posts, though normally we delete all but what
we’re directly replying to, since everyone on CSGnet has already seen the
complete posts.]
From: Bill Powers
[
mailto:powers_w@frontier.net]
Sent: Sunday, March 11, 2007 9:33 AM
To: Ted Cloak
Subject: Re: PCT Presentation Attached
Hello, Ted –
You’ll need Pando to
open this attachment.
Get
Pando free here.
Gary suggested I try Pando,
of which I had never before heard. Hope this works!
I had no trouble downloading Pando or playing your presentation. I think
your technique is admirable. Wouldn’t it have been great to have a video
camera?
** The
filming was done in 68 or 69. No video cameras then! But
video cameras could make this work a lot quicker and easier.
You speak of a “hierarchy” but you don’t really show one.
** Cant
follow you there. I think I show lots of hierarchies, although not
complete ones if that were possible.
I think the
matter of the “cues” would become clearer if you did. The cues
are perceptions, too, and they are part of a program that depends on the
perceived state of the environment. The environment, which contains the
feedback functions for all the control loops, does not cause any of the
behavior: it simply is what it is, and the hierarchy of control systems
(and goals) inside the person determines what will happen as a result. If
you modeled everything as control systems, you wouldn’t need all those
“stimuli” and “responses,” which of course belong to
another theoretical framework entirely.
** I think
Im missing something about PCT here. Suppose youre walking down
the street thinking about PCT or what youre going to have for dinner,
and you hear somebody yell (what sounds like) Hey, Bill!, and then you
turn your head, etc. How would you diagram
that?\
First of all, the sound has to enter your ears and produce perceptual
signals of the first order, intensities. There is no other way for
sound-stimuli to enter the hierarchy, and the entry is always at the
bottom level. Then inside the brain the intensities are perceived at the
second order as sensations of hiss and buzz and whatever else there is.
Then we get phonemes at the third order, configurations, then
vowels at the fourth order, transitions, and consonants at the
fifth order, events. Then according to Bruce Neven, we should expect
sixth-order contrasts (relationships) and finally seventh-order
categories (words, symbols, or other perceptions coupled with collections
of lower-order perceptions). Then come sequences or
words/symbols/categories (sentences) and at the logic level, the first
level that actually works like a computer, programs which consist of
networks of choice-points connecting sequences. Control of principles and
system concepts still lie above, but this takes us to the level you
designate at the highest in your model. It’s also the only level, since
you treat everything as programs and subroutines in programs. This story,
of course, is just my approximation – no real work has been done to pin
down the levels of perception in linguistics or any other field. No
volunteers yet. However, without evidence yet to the contrary, this is
the “official” form of the PCT model as of publication of
“Making Sense of Behavior” in 1998. You may notice that two
levels have been added since 1973, a result of discussions and
suggestions on CSGnet.
A change in a perception that is not caused by your own actions via the
external feedback loop is caused by some independent variable, which we
call a disturbance. So if somebody yells “Hey, Dill” (or Gill
or Phil), the “Bill” recognizing input function at the seventh
level, supposedly, will put out a signal (not as large as the signal
caused by “Bill”), which may or may not disturb perceptions
that are under control at the moment. If no controlled perception is
perturbed, there will be no change in an error signal, and hence no
change in any output, since all outputs, in PCT, are produced by error
signals. If there is a change in some perceptual signal that is under
control, an action will result that tends in some way to diminish the
resulting error signal. Of course this looks to an outside onlooker
unfamiliar with control theory like a response to a verbal stimulus or
“cue.” However, in PCT there are neither “stimuli”
nor “cues” capable of causing outputs directly.
Control systems at levels higher than the first do not produce physical
actions; they produce reference signals that tell lower systems how much
of their perceptions to create, not what to do. Or, as you correctly show
in the one control-system diagram in your presentation, they produce
addresses that select memories of past experiences as reference signals
for what to perceive – still not what to do.
So, a long answer to a short question, which seems to be my habit lately
(there are no “habits” in PCT but the word is evocative). You
show “cues” entering systems of intermediate level and causing
output independent of the reference signals received from higher systems.
That makes the cues into disturbances, and the actions that result are an
attempt by the control system at that level to counteract the effect of
the disturbance on whatever perception it is perturbing.
The Control Systems Group had its first meeting in 1985 (CSGnet started
in 1990), so there have been ongoing discussions and refinements of PCT
for the past 22 years, 17 of them on the internet. Your presentation
describes the hierarchy as consisting entirely of programs and
subroutines in programs, which puts it all at just the eighth level in
the current model, or the sixth level in the model as of two years before
the publication of your papers. I think that if you want to say your
model is based on PCT, it should include more that that. This is one
reason I asked what took you so long to find us. It would have
helped.
It’s not clear from the way
you draw the cues and the lower control processes what is in the
environment and what is inside the person.
** Whats in
the boxes is in the person. Whats between large square brackets is
in the environment.
The entire person is in each box? But that makes this a flow diagram, not
a system diagram – that is, a diagram showing different actions by the
same system, rather then the component parts of one system and the way
they simultaneously interact with one another during any action.
Could it be that you have mistaken the diagram of a single subsystem for
the entire model? I’m not saying you have, just wondering.
Early in the presentation you
say something about avoiding “mentalisms.” I’m not sure what
you mean by that. Aren’t mentalisms like perceptions, goals, desires,
intentions and so on simply neural signals in the brain? What’s the
problem with that?
** Did I say
mentalisms (plural)? Its supposed to be Mentalism (singular),
i.e. the position of psychophysical dualism, that there is mental stuff
independent of physical stuff; the default position of 99% or more of
humanity, including social scientists.
I think you have shown how it pays off to be patient and attend to the
details of behavior. A lot can be done using your method, and I hope it
is taken up and used by others. It looks like a good tool for
investigating higher-level control processes that we have trouble
modeling on a computer. I especially appreciate your willingness to Keep
It Simple and not set up experiments that are far more complicated than
they need to be. There is plenty to learn from watching people nail two
boards together.
** Right we
(ok, they the working anthropologists) have to learn to walk before
we can run.
What took you so long in finding the CSG after all these years?
** Thats a
long story. Ill save it for later.
Best,
Bill P.
End of previous posts.
Best,
Bill P.
You’ll need Pando to