[From Bill Powers (931112.0815 MST)]
Martin Taylor (931112.1750) --
Continuing the conversation:
We can set up a situation that looks as though
it HAS to be open-loop, but if the critical variable proves
resistant to disturbance, we have misconceived what is going
on.
Yes, but I think this sentence misconceives the issue under
discussion, at least in respect of the Lang-Ham configuration.
Their presumption is that the critical variable is under
feedback control, but that changes in the reference value
initially cause some effects in an open-loop manner.
The Lang-Ham way of partitioning control between a negative
feedback system and a feedforward connection is sort of like
vector addition: you can resolve a force into east and west
components, or any other pair, but that doesn't mean there are
any such components in the physical system. You can resolve a
control behavior into a feedforward and a feedback component, but
if various resolutions (including omission of the feedforward)
fit behavior equally well, then it's best to stick with the
simplest model. You can always put the feedforward connection in
again if you ever need it to explain a deviation from the
feedback model -- provided it will actually do so.
The main issue we have to clear up is whether a purely open-loop
model is EVER a plausible representation of the mechanisms of
behavior. Combining feedforward with feedback only obscures the
main issue, and in a way that is needless considering the present
state of our knowledge. It encourages people to think that if
feedforward can be part of a control behavior, maybe sometimes it
accounts for the entirety of the behavior. It can even encourage
people to believe that open-loop feedforward is the _usual_ mode
of behavior, with feedback occurring only now and then.
I am trying to get people to think more carefully about behaviors
that they describe in haste. When we talk about "walking in a
dark room" the term "walking" seems to wrap up one aspect of the
behavior quite neatly; all we have to do is decide to walk and
off we go, right? But walking is not an action: it is an outcome
of a complex series of actions all occurring at the same time, in
combination with inertial effects, gravity, and other
disturbances. A command to "walk" has to enter a system capable
of turning that command into exactly the detailed actions that
will result in walking rather than falling down, running,
skipping, hopping, and so forth. To treat walking as an open loop
behavior is to name the behavior in terms of its effects instead
of in terms of the mechanisms needed to create those effects. And
when you name behavior in terms of its effects, you completely
lose sight of all the hidden mechanisms required to accomplish
it. It is only through concealing hidden mechanisms in that way
that we can make open-loop behavior seem simpler than closed-loop
behavior.
ยทยทยท
-----------------------
When J. B. Watson was a graduate student, he had a tremendous
insight about purposive behavior. He was holding a bird which
seemed to be struggling to escape; its struggles, when allowed to
succeed, always carried it toward its nest. For some reason, I
belief this epiphany took place in the Galapogos Islands.
Watson's great insight was that these struggles _were caused by
the sight of the nest_. The sight of the nest induced the bird's
muscles to exert efforts that would take it toward the nest: what
else needed to be explained?
Moving toward the nest was indeed an outcome of the struggles.
But if you asked how this outcome could be achieved, it would be
clear immediately that there was a great problem in the phrase
"toward the nest." Why did the sight of the nest not induce
struggles in a direction 15 degrees to the right or left of the
nest, or even at right angles to it or away from it? If you
carried the bird around to the other side of the nest, would it
still struggle in the same compass direction? No; now its
struggles would be aimed the other way by the compass, although
still in the direction of the nest.
The explanation that might have been offered later by B. F.
Skinner was that the bird had been reinforced for struggling
toward the nest. Evolutionists might have explained it by saying
that birds which did not struggle toward their nests have become
extinct through failing to take care of their eggs. But none of
these glib explanations answers the basic question of how a bird
can possibly struggle "toward" a nest. All other kinds of
explanations are simply attempts to appear wise without actually
knowing anything about behavior.
By pursuing this question to its end, Watson could have come up
with control theory. He could have realized that there must be
something about the perceived direction of the nest that
influenced the direction of struggling; that if the bgird moved
to the right of the direction of the nest, the struggles would
aim more to the left, and so on. And then, if he had not stopped
thinking at that point, he would have realized that as soon as
the direction of the struggles carried the bird more in the
direction of the nest, the perceived deviation would necessarily
decrease and possibly even reverse, decreasing or reversing the
tendency to alter the direction of the struggles. The closed-loop
nature of the situation would have become perfectly clear, and we
would have been spared 80 years of nonsense.
But Watson stayed at the level of his first insight; he thought
he had explained the behavior simply because he had found a way
to describe it that put the cause into the environment. He was
guided by the structure of language instead of by careful
exploration of the phenomenon. He wanted to believe in his great
insight: why go any further when you have found the answer?
I have always found this aspect of psychology particularly
exasperating: the widespread habit of offering explanations that
don't explain. When Skinner said that a certain behavior was
produced because the animal had been reinforced to produce it, he
sat back with a smug attitude assuming that he had just closed
the subject. Evolutionists -- sociobiologists especially -- do
the same thing. This peculiar way of explaining things manages to
avoid the very heart of the problem, which is the nature of the
system that is producing the behavior that we see. Perhaps
reinforcement or evolution would explain the etiology of that
mechanism, but if you don't know what the mechanism is, how can
you know the nature of the burden you are putting on the
explanation? It's as though scientists had given up on
understanding the HOW of behavior, and had decided that they
would henceforth explain it in terms of superficial apparent
relationships -- basically, supersititions and prejudices.
When I look at the behavior of an organism, I don't look around
for events in its environment, or into its history, to explain
what I see. I wonder "How can this organism possibly be doing
what I see it doing right here in front of me?" I don't care how
it got its organization. I just want to know what that
organization is. Neither do I care about general categories of
which this behavior is just an instance; I want to explain THIS
behavior, not ALL POSSIBLE behaviors. If I can explain THIS
behavior, and THAT behavior, and still another behavior, maybe I
will begin to see that the explanations are alike in some basic
respect, and then we will have a theory. I don't want to keep the
phenomenon at arm's length; I want to be up to my elbows in it.
Explanations that are nothing more than naming behaviors by their
outcomes, that explain outcomes in terms of reinforcement,
evolution, social influences, childhood trauma, chemical
imbalances, mental quirks or conditions, or general scientific
principles like thermodynamics all beg the only question about
organisms worth answering: what is behavior and how does it work?
What goes on inside an organism that it can do what it is doing?
The meaning of all other questions depends on the answer to that
one. If we can't answer that question, we're trying to answer
other questions of no importance: we're trying to sound as if we
were answering the basic question, but we're not doing it.
Martin, I hope you can tell when I'm talking over your shoulder
to onlookers. And when I'm not.
-------------------------------------------------------------
Best,
Bill P.