[From Bill Powers (950322.0640 MST)]
Bruce Abbott (950321.1900 EST)--
Sounds as though you're going to the BAAM meeting loaded for bear. Be
sure to tell the three people who get turned on by the demos how to join
us on CSG-L.
The SD supposedly acts as a sign or signal that a particular set of
relationships between responding and reinforcement/punishment is
now in effect: that particular responses will have certain
predictable consequences.
It's odd, but sometimes I think that EABers are being too cognitive in
their interpretations. I think biologists and evolution-theorists do
this, too, despite the professed claim to be discussing a system with no
goals and a purely mechanistic, environment-controlled nature. The
problem is that it's impossible to talk about behavior without subtly
invoking purpose, meaning, interpretation, and so forth.
An SD is basically just a physical environmental variable that changes
state. Normally it doesn't even have any physical effect on the
relationship between certain actions and their reinforcing/punishing
consequences that it "signals." What it is saying is "Now there would be
a consequence of a certain type _if_ you were to perform a certain act
in the requisite way." The "signal" is really a rather complex message
which requires a cognitive system to understand it; it refers to a
hypothetical situation which may or may not come about.
So it seems to me that the concept of the SD puts a large burden on what
is basically nothing but a simple physical variable. This is the sort of
thing that has led me to accuse behaviorists of "animism." Even the
meaning of SD -- "Discriminative Stimulus" -- attributes what is
basically a perceptual function, discrimination, to inanimate things in
the environment. The observer is projecting his/her own perceptual
processes onto non-neural processes in the outside world.
To say that a stimulus is "discriminative" is really to say that it
_could be used as the basis for discriminating_ by some system that is
capable of making (and has learned to make) the appropriate
discrimination. A flash of light will not give a bacterium on the key an
urge to press it. Likewise, to say that a pellet of food is a
"reinforcer" is to refer to processes inside an organism that convert
what is otherwise only a physical event into something of significance
to a particular organism that happens to be lacking that pellet of food
-- i.e., that wants it and can use it. Implicit in the language of EAB
are properties of organisms, not properties of the environment. But the
language tries to describe organismic properties as if they were
observable properties of the environment.
So even though the professed aim of behaviorism is to analyze behavior
strictly in terms of observables, this aim is not in fact achieved. A
purely physical account of what is observed in an operant-conditioning
experiment would be incomprehensible, for the same reason that a purely
neural account of what a brain does would be incomprehensible. In some
way or another we must at least allude to processes carried on by the
organism if we are to understand behavior.
I have always been troubled by a certain attitude among behavioral
scientists, one that I have had difficulty in characterizing. It is more
or less to the effect that organisms are just pieces of matter, and as
such are subject to all the laws that run the rest of the physical
universe. This idea leads to a kind of self-satisfied yet bitter
attitude toward living systems and particularly toward other people. It
conveys the message that no matter how smart we think we are, no matter
what we think or feel, we are really just mechanisms being pushed around
by external events. Our sense of purpose and existence, our sense that
we pursue goals for our own satisfaction, is just a cruel illusion --
the environment is giving us those impressions, too, and determining
what we think and feel about them. In the long run, therefore, nothing
really matters -- either what we do for ourselves, or what we do for or
to other living systems. Suffering and joy are simply epiphenomena,
side-effects having no physical significance. Life and death are simply
two possible states of matter. If you just convert "the environment" to
"God" and "physical" to "spiritual" you have the basic attitude of the
Dark Ages.
While it is true that organisms are composed of the same kinds of matter
to be found in the inanimate world, and that the laws of physics and
chemistry apply inside them as well as outside, it is not true that the
laws of behavior are the same laws that govern the inanimate world. The
laws of physics and chemistry are _organizational_ laws, laws concerning
the way pieces of matter interact with each other. When we study
inanimate matter, the laws we find are those appropriate to inanimate
matter. But when we study organisms, we find matter organized in ways
that no other assemblages of matter are organized, and we find that new
laws need to be found and understood. A living system, as someone once
said, is a standing wave of organization imposed on a ceaseless flow of
matter and energy; it is not a thing, but a pattern. It is organized as
no nonliving entity is organized.
PCT is basically a mechanistic theory, but the meaning of "mechanism" is
very different from the behavioristic meaning. What makes a mechanism is
not the materials of which it is made, but the organized interactions
that take place between its parts. As we began to appreciate when the
computer revolution came upon us 35 or 40 years ago, a "mechanism" can
carry out functions which had always been thought to be the exclusive
province of living systems, even human beings. The boundary between
"mental" and "physical" has been erased. Even though we do not yet
understand how important aspects of experience can exist in physical
mechanisms, we now have good reason to think that further explorations
of mechanism will lead to such an understanding.
This will by no means diminish the status of a living organism to that
of "just a machine." To think that is to remain stuck in primitive
notions of what a machine is, or could be. A machine is not by
definition something that is operated by the environment. Machines can
be self-determining and creative actors that operate on the environment
to make it deviate from its "natural" course to suit the purposes of the
actor. In short, the distinction between the machine and the living
organism is also in process of being erased. But it is being erased by
elevating our conception of what a machine is, not by reducing an
organism to some less-capable kind of thing. Most of what we have
customarily considered to be the limitations of machines no longer
exist. In fact, the word "mechanism" no longer refers to any special
kind of organization. We can speak of the mechanisms of thought as
easily as the mechanisms of movement. A scientist living only 60 years
ago would have found this way of speaking inexplicable.
Sorry for the irrelevant digression -- but maybe you can use it at the
BAAM meeting. Be sure to extend all our greetings to Dennis Delprato.
···
---------------------------------------------------------------------
Dag Forssell (950321 1600) --
What a magnificent job you're doing with the demo disks! Take heart --
in three years or so there will be affordable drives for making CDs, and
your space limitations will disappear. We will miss you very much at the
next CSG meeting, but of course wouldn't dream of wishing that the
joyous reason were otherwise.
I, too, have the entire archive (except for a few posts at the very
beginning) on my hard drive. If you would like a copy of a shareware
indexing program for it (full Boolean search) just say the word. In
fact, I refer to it so seldom that I may follow your lead and stick it
onto a backup tape, keeping only one year of it for quick reference.
With Gary Cziko working on a WWW resource for PCT, we may be able to
dispense with the disks eventually. But the infobahn has to get a lot
further along. Unfortunately, WWW isn't available yet to us users in the
boonies working through VAX mainframes without any Unix capabilities.
-----------------------------------------------------------------------
Martin Taylor (950321 11:30)--
Sorry about the tangle -- it was Bill Leach's "circuit" you were
referring to, not mine.
I don't think that ["or"] handles "category" as I perceive
categories, for two reasons. One is that an analogue "or" is a
very funny beast, known to fuzzy logic as something one might call
maximized partial membership. The other is that categories have
boundaries. Something is or is not a member of the category, and
if it is not, it is usually a member of a contrasting category.
It may or may not be a member of a contrasting category -- but it will
not include ALL other categories. I'm not trying to deny the particular
case in which limited sets of categories are mutually-exclusive, with
hysteresis and all that. I'm just making the claim that this is only one
kind of categorization. If you think of the category "All things with
three legs" you can see that it includes a very large set of objects
with nothing in common but possessing three legs (a three-legged dog,
for instance). But this category does not exclude four-legged dogs or
other things like centipedes, and it is irrelevant to things with two or
fewer legs, like algebra.
Your version
is basically the conventional one that treats category perception as a
box receiving multiple inputs and capable of producing multiple
outputs (or different codings of a single output), one for each
possible "recognition" of a category.
How do you read this into either my words or my diagrams? Both the
diagrams and the words are quite explicit in "one perception, one
perceptual function." There is no concept of the kind of multiple
output box you impute.
When you say that we are given a thing and must answer the question
"Which is it, X or Y?", you are implying that the perception must be
either X or Y: the question itself forces this choice. So from the point
of view of the asker of the question, the answer is found in a single
channel which can carry the message "X" or the message "Y", implying a
perceptual function that can deliver two different messages.
If the "channel" is a higher-level system's perceptual function, then it
is that function that is deciding which category-signal to accept. You
can, if you like, put some of the decision into a category level
organized as a flip-flop so that only one category signal at a time can
be emitted. In that case the higher system needs to identify which
lower-level signal source is active in order to detect the answer. The
same result could be obtained if the higher-level system received both
category signals but through a perceptual function that contained a
flip-flop so that only one signal could result. However, that would mean
that _either_ X or Y would satisfy the higher system, by producing the
same perceptual signal. So there would be no discrimination between X
and Y.
your cross-connected flip-flop scheme is designed to assure that there
will be only one answer.
That is correct when the cross links have strong negative weight.
It is also perceptually correct when a pattern can be seen as one
thing or another but not both at once.
I am having a problem with seeing every cross-connection as an example
of category perception. The reciprocal innervation in the spinal cord
brings opposing muscle systems into a single two-way control system, but
it's hard for me to see this as categorization.
However, my only reason for introducing categories in the first place
was to provide a way to name things, so a symbol could be used to stand
for some collection of analog perceptions. Beyond that, I really don't
have much to say about categories beyond some examples that seem
problematic to me. I can see a lot of different ways to achieve
category-ness, but no way to pick one over another. No matter what
circuit I (or you) think of, I can think of examples that seem like
categories to me but don't fit the circuit. And given one circuit for
achieving a category-like result, I can think of others that would
accomplish the same end in a different way.
Your universally-adjustable cross-connections will, of course, allow for
a great many types of categorization, including what I call the simple
"or" type. I have a problem with how the system doing the adjustments on
these cross-connections knows what it is doing -- it seems to me that it
would already have to know what a category is in order to tell whether
it is making the adjustment correctly. But this whole subject confuses
me; I'm not ready to understand it in any complex way. Your proposal
suggests some immense complexities when you think of all the possible
combinations of cross-connections and strengths; I'm not up to grasping
them.
... it is not relevant, so far as I can see, to the control of
category perceptions. All that demands is that we be able to bring
about the perception of something being in the reference category.
I'd just as soon leave it at that. But you can carry it as far as you
like, of course.
I think that categorization is essential BEFORE there can be any
logic.
Yes, that's what I have said, too. You need to convert the analog world
into signals that are basically symbols, so the logic level can treat
them as "present" or "not present."
What were the observations that led you to postulate the existence
of a "category level?" What mechanism might account for them?
Only, as I said above, that we do name or symbolize things, and then
treat the synmbols as if they stood for a category of things, ignoring
the differences between them. "A dog" is any dog, without regard for the
infinite variety of individual dogs.
----------------------------------------------------------------------
Avery Andrews (950322)--
Great examples about behavioral thermoregulation. I wonder if the Test
could be formally introduced to herpetologopolists -- uh --.
----------------------------------------------------------------------
Paul Stokes (950321.1045 GMT)--
I was hoping that our sociologists would be responding to your original
message. We have a very active subgroup in sociology, including Kent
McClelland, Dan Miller, Clark McPhail, and Charles (Chuck) Tucker.
This is a hint, fellows.
---------------------------------------------------------------------
Best to all,
Bill P.