[From Bill Powers (930913.2000 MDT)]
Michael Fehling (930912.1218 PDT) --
A few comments, some redundant, on your remarks to Rick Marken:
RE: behaviorism
Usually ["behaviorism"] is reserved for approaches seeking to
describe/predict/explain _concrete_instances_ of behavior,
rather than idealized and abstracted classes of instances.
The distinction concrete-abstract is not the one I make. Try this
distinction between behaviorism and PCT:
Behaviorism is the study of the perceived consequences of motor
outputs from a point of view other than that of the behaving
organism. PCT is the study of the perceived consequences of motor
outputs from the point of view of the behaving organism.
RE: Chomsky's generative models
Note that "generation" is _not_ processing or behavior, it is a
logical relationship between [a] rule's input structure and its
output structure.
For me rules don't exist apart from the mind that is conceiving
and following them. A physical structure is required to support
logical operations, either perceiving them or carrying them out.
When you describe rules, you are describing something being
perceived by an organization in a brain, not something that
exists in a disembodied space. A person who reasons about rules
and abstract structures is using a particular faculty of the
mind/brain, even if this fact is not in awareness at the time.
The processing that is occurring while a rule is apprehended or
used may be unconscious, so it seems that the rule simply exists
in a separate universe, but I don't accept that appearance as
real, nor could I model it that way.
If there is a logical relationship between a rule's input
structure and its output structure, then the output of the output
structure is a function of the inputs to the input structure, and
the whole relationship can be described as the operation of a
physical device.
I know that Chomsky didn't think of his scheme as a model of
operations in a brain (he told me so once, explicitly, in a
letter), but it is one, nevertheless. He was describing something
his own brain was doing.
RE: causality in the operation of closed loops:
... traversals of this loop occur in time, Bill's static
analysis notwithstanding.
I want to expand on Rick's brief response to this point. If you
focus on a perturbation in a control loop, you can follow "it"
(that is, a succession of further perturbations at points farther
and farther around the loop) all the way back to the starting
point. This has led many people to think of the closed-loop
process as a succession of events. First, perception; then
comparison; then error; then action; then an effect on input that
tends to correct the original perception toward the reference
state.
This could be called the "mousetrap" image of the closed loop:
each mousetrap triggers the next one in line.
The proper image, as suggested by David Goldstein, is the "wheel"
image. When a wheel turns, all parts of the rim move at the same
time. If you apply a tangential force to the wheel, this force
propagates immediately all the way around to the point of
application, and from then on the whole wheel turns as a unit.
Another image is that of blood circulation. We can trace a blood-
cell from the heart, through the arteries, through the
capillaries, through the veins, and back to the heart. But we
can't forget that there is a steady stream of blood-cells both
leaving the heart and returning to it, all at the same instant,
all of the time.
Neither circulation nor the turning of a wheel involves any sort
of cycle of activities. An apparent cycle, an artifact of
observation, can be created by an observer who focusses on one
feature of the closed circle -- one blood cell, one point on the
rim of the wheel -- and mentally follows it around and around the
loop. But the process itself is continuous, not cyclical, and all
parts of it go on concurrently. This is the case for control
processes, too.
As Rick indicated, ordinary causality does hold in the input-
output relationships that define the individual functions
comprising a control system and its external feedback path. But
in the loop as a whole, there is no causation of that kind,
because each variable in the loop is a function of itself,
delayed in time but still continuous. The perceptual signal is
not only a function of itself, via the output effects through the
environment, but also of independent environmental disturbances.
It's the action of the whole loop that brings the perceptual
signal toward its reference level despite the disturbances, and
keeps it there.
Disturbances SEEM to be related to the actions of control systems
in the ordinary sense of causation. It's this sense of causation
in S-R (or S-O-R) theory that is rejected under PCT, because this
is not the direct causal relationship it seems to be. The action
is not simply caused by the disturbance, because the reference
signal also, at the same time, determines what will be considered
a disturbance.
Likewise, changes in reference signals SEEM to cause the actions
of a control system, because a sudden change in reference signal
will produce a sudden change in output activity. In actuality, it
is the difference between the perception and the reference signal
that produces the output, as becomes evident when the changes are
followed further through time or continuously. Cognitive theories
assume a command path like the path from the reference signal
through the comparator to the output, and treat it as a causal
path: the command causes the behavior. That causal concept, too,
is rejected under PCT.
When Rick refers to other theories as "causal" theories, he is
rejecting the two senses of causation above, not the basic
concept of physical causation or its application to individual
functions in a control loop.
The reason for rejecting the two concepts of causation is neither
religious nor abstract. In any control system, the output actions
can change over their entire range with no change in the
reference signal (because of disturbances); the output actions
can change over their whole range with no change in the
independent disturbances (because of changes in the reference
signal). Neither the S-R nor the cognitive concepts of the causes
of behavior can account for these facts. The S-R interpretation
seems to work but only with a constant reference signal; the
cognitive-command interpretation seems to work but only in an
environment free of disturbing influences. In real environments
with real organisms, neither interpretation works.
RE:
My interest in a cybernetic model of human performance (at
_all_ levels) comes from my suspicion that it is the only type
of performance model that can accomplish what I have just
sketched.
Amen.
···
----------------------------------------------------------------
Best,
Bill P.