[From Bill Powers (920608.0100)]
My host computer went off the air last Friday just as I was about to send a
message saying that I'd be gone for a couple of days. I'm back.
Chuck Tucker asked me for a copy of my observations about "Coach" in high
school. I tried to find them and couldn't (I searched through four months
of posts and found only that I talk too much). If anyone has a better
filing system than I do (easy to accomplish), would it send a copy to Chuck
at n050024@univscm.bitnet? (Those are zeros, not ohs).
J. Marvin Brown (CT linguist from American School in Thailand) will be
visiting here later today and tomorrow. Messages from other linguists (or
anyone else) welcome. Marvin is not on the net.
···
------------------------------------------------------------------
Ray Allis (920604.1200) --
Penni Sibun (920604.1600) --
I think there are four universes of discourse overlapping here (at least)
in the discussion of modeling, simulation, etc.
Ray is arguing that symbols manipulated in a computer have nothing to do
with reality, in that symbol-manipulations can contain no surprises and the
rules for manipulation contain all possible deductions from them. Penni is
pointing out that the computer itself is a physical reality and that the
symbols themselves are quite real as is the structure (in the computer)
that is realized in terms of symbols. That's "on the one hand."
On the other hand, in the world of CT modeling so far, the computer is not
used to manipulate symbols at all: it's being used as a numerical analog
computer dealing with variable quantities. Furthermore, I haven't yet
pointed out that in many of our simulations one component is a real person
performing real behaviors (moving a control handle and perceiving one or
more moving targets and cursors, or other types of displays) in a
continuous manner while the remainder of the setup runs as a program inside
the computer. So the program has no advance knowledge of what the person is
going to do next, nor does it know how the person will interpret and react
to the display. The "simulation" aspect then amounts to trying to write
another program which, when substituted for the real person in the same
experiment, will produce the same record of behaviors that the real person
produced (or will produce in the future, under new experimental
conditions).
In the case of CT experiments, I object to classifying the processes inside
the computer as "logical deductions" because the same processes could be
carried out by resistors and capacitors and operational amplifiers in a
purely analog setup (or any other means of analog computing such as
pneumatic or mechanical).
On this same hand, Rick Marken has said that a link between the computer
simulation and reality is continually present through the experimenter, who
is comparing what the computer does with what a real person does in the
same experimental situation, with a very critical eye for differences. The
simulation does not run free of influences from reality; in fact the
simulation is set up to preserve critical correspondences with the external
world or the real system. The better simulations attempt to reproduce many
correspondences, not just those at input and output.
On the third hand, there is the fact that either in a digital computer or
in an analog computer, a human being must be present to interpret the
inputs and the results. Digital and electronic analog computers don't
actually manipulate any symbols or perform any logic. They manipulate
voltages according to properties of electronic circuits. These voltages
have no meanings, because neither kind of computer has any experiences
(except other voltages) to become the meanings of the voltages.
The programmer of either kind of computer assigns meanings to arbitrary
voltages used as symbols, these meanings coming from the programmer's
experiences, not the computer's. The computer then manipulates the voltages
according to the physical properties of its components and its programming,
and produces, as output, more voltages. The programmer then looks up the
assigned meanings of the output voltages (preassigned by the programmer)
and interprets the results accordingly.
In digital computing, the inputs and outputs are usually cleverly
programmed so that the assigned meanings of the voltages are entered and
read out as codes displayed to look like words or numbers that remind the
programmer of their intended meanings. These words or numbers have no
meaning in the computer, but they do have meaning for the programmer: the
order in which different words or numbers come out of the computer and the
words or numbers themselves suggest new thoughts and experiences to the
programmer (or, of course, the user).
In analog computing, the input and output voltages (now continuously
variable instead of on-off) are assigned meanings as physical variables:
the inputs might be said to correspond, for example, to temperatures,
pressures, and densities, while the output voltages might be said to
correspond to rates of heat flow in various places. The inputs are set up,
the computer is set in motion, and a tracing of the output voltages is
delivered. The output voltages are plotted against time with the
coordinates labelled as "joules per second, heat exchanger," and so on,
which is the meaning of the output to the human programmer, but not to the
computer. The assignment of meaning by the human programmer is more obvious
in an analog computer, which does not produce outputs in the form of
recognizeable words or numbers.
On the fourth hand, there is the question of the nature of the human
programmer. If we accept the idea that the programmer's brain is a neural
machine, as I do, and that all aspects of experience and action are
embodied in and by this machine, then it follows that both symbols and
meanings exist in this machine as neural signals, trains of impulses
occurring at variable frequencies. "Reality" then becomes a representation
inside this machine, knowable directly to the machine, but corresponding to
something not directly knowable by the machine.
The problem of modeling and simulation then becomes one of creating an
analog of the brain-machine in relationship to a hypothetical physical
reality outside the brain-machine. This hypothetical reality also exists in
the brain, in the form of neural signals being handled by neural functions
that exist as synaptic connections and chemical reactions. The brain
sometimes attempts to mirror itself in artificial machines, giving them
internal operations that make sense to the brain. It naturally constructs
these artefacts to accept inputs and produce outputs that can be perceived
by the brain, that perform operations recognizeable by the brain, that
remind the brain of experiences other than those inputs and outputs. So the
brain, which operates in terms of basically identical signals handled by
processes it cannot perceive directly, tries to discover its own internal
organization by creating external artefacts in which those invisible
operations can be made visible -- and which also operate in terms of
basically identical signals.
There is, finally, a fifth hand, but of a different kind. All these
conceptions of a brain, including that of a brain as a neuronal machine,
exist in the consciousness of an observer-manipulator which is ourselves.
Before we learn anything formal about the physical world or machines, the
experienced world exists just as it exists. Whatever we guess to exist
behind this world of direct experience, it is always direct experience to
which we return in evaluating any guess. Any conjecture that directly
denies direct experience is simply wrong. If you want to say that some
aspect of direct experience is "imaginary," for example, you must then
explain how it is that this particular experience was imagined: if you
conclude that it was not imagined exactly as it appeared, then something is
wrong with your reasoning. No person can do other than rely on direct
experience as the final arbiter. If it seems to you that someone else is
mistaken, then that is the case: that is, in fact, how it seems to you. If
you think the other person is right, then that, too, is how it seems to
you. There is simply no way to deny direct experience. If I think
"therefore I am," (or "therefore I am not") then that is what I did in fact
think. It does not matter whether, in some hypothetical "objective" world
independent of my experience, I "really am." Such "objective facts" and
their implications are conclusions which we test, ultimately, against
experience.
It's not easy to keep straight where we stand in the midst of these
different universes of discourse. I think that we're trying to achieve
consistency among them, so that moving from one to another doesn't create
contradictions among them. No one of these points of view is so important
that we can afford to ignore the others; from each point of view we see a
different facet of the mystery of being conscious in a here and now.
-------------------------------------------------------------------
Best to all,
Bill P.