[From Rick Marken (930617.2000)]
It's hard for me to edit from this computer so I'll just
do this from memory.
First, I'm sorry if it seems like I'm flaming you, Martin.
I don't mean to. I'm kibbitzing with you. I thought my
coin flip reconstructor was kindda funny. I didn't mean
to attack you -- just to make a point about what you
seemed to be saying. I've been reviewing documents for the
last couple of days -- it gets pretty boring and the net
is my only outlet. I guess I've been a little giddy.
Next, Tom and Bill.
Bill Powers hit the nail on the head with his last post, Tom.
I was playing cat and mouse with Martin -- letting him have
more and more of the control system in order to predict the
disturbance, thinking that I would let him get confident
and then administer the coup de grace with a varying, non-linear
feedback function and show that, even if you know EVERYTHING
about the control system, you still can't reconstruct the
disturbance from the perception. I thought it would be obvious
that the control system could not conceivably have any knowledge
of the feedback function, especially if it was nonlinear AND vary-
ing. Obviously, my plan backfired because now Martin is convinced
that he can reconstruct the disturbance -- since he can do it if
he is given knowledge of everything about the control system
except the disturbance itself.
So now there's no more games; if there is information in perception
that "gets through" to the control system then somebody ought to be
able to find it and reconstruct the disturbance from it. If Martin
cannot reconstruct the disturbance from p (as he admits that he cannot)
then it's over; there is nothing (call it information, god or whatever)
in the input quantity or the perceptual signal that can tell a control
system how to produce outputs that keep the perceptual signal at
the reference level. Information theory may have something useful to
contribute to PCT -- but it is not the idea that there is information
about the disturbance in the controlled quantity or the perceptual
signal that represents that quantity.
I'll resume the discussion of information in perception when anyone
sends me back the disturbing variable values associated with
perceptual values that I will post upon request -- no output
function, no feedback function. Just perceptual and reference
values -- the only two conceivable sources of information coming
into a control system. Any takers?
Best
Rick