[From Bill Powers (920326.0900)]
Avery Andrews (920326) --
But shouldn't we distinguish cases where the action (output = O)'s
effect on the input (I) flows thru to effect further changes in O from
those where it doesn't? E.g., in the infamous plummeting moth, it is
perhaps the case that the change in position of the moth caused by
folding up (O) causes some change to I, but there will not typically be
any resulting further change to O ...
I assume that control systems always pass their inputs continuously
through the comparator to their outputs as long as they're turned on at
all. My picture of the moth's plummeting response is that the moth
normally controls for bat-sound intensity by flying away from it (so the
inverse-square law reduces the intensity below the reference level --
this would be a one-way system that controls only for excesses of the
input over the reference level and does nothing about smaller amounts of
input). "Away" might also mean "down," and the control system might have
an evolutionary bias for "down" because of the protection afforded by
leaf litter etc.
If my concept of the moth's hookup is right, then the plummeting
response results when the bat is near enough to cause a very loud sonar
sound relative to the reference level, creating a very large error
signal. The plummeting response represents the fastest downward velocity
that the moth can produce. If the sound is loud enough, however, this
response won't be enough to bring the sound intensity below the
reference level. The error signal will remain large and the sound input
will be uncontrolled because the "output" opposing it can't get any
larger. So the moth is in an open-loop situation, even though it is
still organized as a control system. It is simply faced with a
disturbance larger than it can handle.
This imaginary picture supposes that there are levels of bat-sound
intensity below which the plummeting response won't be seen; the moth
may descend to the ground or fly away from the sound, but this will
suffice to keep the sound below the reference level. This wouldn't be a
very dramatic response, because the moth would still be controlling for
all the other inputs with which it's concerned and all you'd see would
be a bias in the flying patterns away from the sound. With small
excesses of sound intensity, you might see the moth descend normally to
the ground for a while, then rise again and go about its business. There
wouldn't be any big interesting "response" sticking out to draw
attention to itself. But you'd be seeing the same control system working
in its normal range of operation. This is what I mean by saying that if
you only notice extremes of behavior, you'll miss most of what's going
on.
I have no problem at all with the belief that (sub-)systems that are
supposed to run open loop (my sense) are pretty rare, but I think it's
quite important to accept them without a fuss if they stand up to
careful scrutiny. If you don't, people are likely to get the idea that
PCT is some kind of religion rather than an actual insight into what is
usually going on with living things.
That's what we're trying to do, subject the behavior to careful
scrutiny. Unfortunately we have to imagine a lot of the data, but that
could be remedied if someone were to give the moths a closer scrutiny.
Maybe they have -- if so I have yet to hear about it. My suspicion is
that observers of these moths don't believe that the moth can hear a
distant bat sound, and flee from it, unless the moth plummets.
I completely agree with not treating CT as a religion -- forcing the
appearance of control onto every situation and rejecting every piece of
evidence to the contrary. No matter how much I disbelieve in open-loop
behavior in organisms, it can still happen, and I wouldn't automatically
reject evidence that says "we looked for control and there wasn't any."
If I'm skeptical, it's because I never see data that includes a check to
see if control was going on.
···
-----------------------------------------------------------
Best
Bill P.e2