[From Rick Marken (930319.1300)]
Martin Taylor (930319 14:30)--
Now I am being asked not
only to show that the disturbance is at some point represented in the
perceptual signal but also that a scalar variable is supposed to be
capable of answering "why" questions about the disturbing variable!!!!
No. What is being pointed out is that, although the disturbance may,
at some point, be represented in the perceptual signal (it is,
for example, so represented while p(t) = d(t) + zero) this can be
of no functional significance to the control system, which simply
gets p(t) -- and cannot respond any differently to p(t) values
resulting from the fact that p(t) = d(t) + zero, p(t) = zero + o(t)
or p(t) = d(t) + o(t).
The result is to show that to the extent the
disturbance is represented in the perceptual signal it isn't represented
in the output signal, and vice versa. That's all.
Why is that controversial? More to the point, why is it alarming to you?
If that is all you wanted to say -- which is really just saying "when
p(t) = d(t) it is because o(t) is zero" then there would have
been no controversy or alarm. I am prepared to agree that, if a=b+c
and c = zero then a=b. But you are saying a bit more, as
revealed in the following:
What is "bled off" is the deviation of the perceptual
signal from its reference, which the Engineer can see has been due
to the disturbance.
This is only true during the transport lag (and after some period of
no disturbance since the output is an integral); so it is an unusual
circimstance. The engineer would usually see that what is
continually "bled off" is the deviation between reference signal and a
perceptual signal that is the COMBINED result of the disturbance AND
system outputs. What you imply in the above is that the "disturbance
produced" deviation of perception from reference is what determines
the output. But this is typically not the case; what produces the
deviation of perception from reference is the combined effects of
disturbance and output. You are implying that there is some
important functional significance to the fact that the disturbance
is occasionally represented almost perfectly (limited by noise) in
the perceptual signal. But this is WRONG. Whether or not the disturbance
is EVER represented in the perceptual signal is not important to
the functioning of the control loop -- in fact, in all the control loops
we build, the disturbance is virtually NEVER represented ALONE in the
perceptual signal -- it is only so represented when o(t) hits exactly
zero or is constant for a couple samples -- pretty rare events.
As far as the control system is concerned, that
deviation might well be due to its own output (as when you experiment
with changing the sign of the effect of an action). Allan considered
that this kind of inappropriate (i.e. not contributing to negative
feedback) effect of output ought to be considered as part of the
disturbance. I think that's a semantic matter.
Can we agree that perception is ALMOST ALWAYS a SIMULTANEOUS result
of output and disturbance ? I am beginning to realize that our problems
may result from your apparent belief that it is necessary that
sometimes only the disturbance (or "inappropriate output") cause the
perceptual signal to deviate from the reference signal. You apparently
believe that the only way for disturbance countering output to be
generated is for a disturbance produced error is generated. But this is
NOT TRUE.
NOTA BENE. The size and sign of the difference between perception and
reference is at (virtually) all times the SIMULTANEOUS result of
disturbances AND the effect of the system's own output.
This is why I say that there is no information in the disturbance
that can determine output. The determiner of output is the size
of the error signal; can we agree on that? But the error signal is
virtually NEVER determined by the net effect of only the disturbance to
the perceptual variable; the error signal depends on BOTH disturbance
and output.
So, the reason this topic is controversial is because it seems to
me that what you are saying about the operation of a control system is
incorrect.
Why is this alarming? Because it seems to me that you are trying to
squeeze the beautiful concept of control of perception into the
Procrustean bed of conventional linear cause effect thinking. PCT
demands that we start psychology all over again -- psychology beautifully
reborn. But, if we keep trying to see control in linear cause effect
terms (instead of in terms of the circular causality that is actually
involved) this renaissance will be continually be delayed.
Perhaps we can focus this discussion for a moment on the
main point I made in this post -- about the cause of error in
a control system. Let me ask you the following questions; I think
your answers could really help us move towards a solution
to this apparent disagreement:
1) Do you believe that the disturbance to a CEV (and the perecption
thereof) typically causes the error (discrepency between reference and
perception) that leads to the output that opposes that disturbance?
2) Do you believe that it is necessary that the disturbance at least
occasionally be represented in the CEV?
3) If you answered "yes" to (2), why do you believe this?
By the way, just by using the term CEV I assume we are discussing a variable
that is at least somewhat under control -- like the cursor in a
compensatory tracking task. So let's make these questions tangible and
assume that they relate to that tracking task.
Best
Rick
PS. Gary and Rich -- I appreciate your posts on my research proposal and
I will respond to them ASAP. But my priority at the moment (obviosuly)
is to get to the bottom of this "information about the disturbance"
controversy.