[From Rick Marken (940615.0900)]
Bob Clark (940614.2205 EDT) --
Your demonstration of controlling programs by simply pressing buttons
YOU (not some modelled version of you), pushed a button to select the
program to be displayed.
YOU monitored (observed) the perceptions composing the displayed sequence of
YOU compared these perceptions to the perceptions as YOU imagined them.
I would say that I compared these perceptions to a reference for what I
wanted them to be -- at least, that's what the model says I was doing; I was
only aware of keeping the program (perception) happening. I was not imagining
anything (well, this kind of experiment can get a bit boring so I suppose I
might have been imagining _something_ but it sure wasn't programs;-).
This process could be described as: "a comparison of this [sequence of
numbers]" with a remembered (imagined) preselected [sequence of
Yes. Except that in both cases it is neural signals (not actual sequences)
that are compared and the perceptual signal does not represent a particular
sequence; it represents a _program_. So two very different sequences could
still be examples of the same program, which was "if odd then >5 else =< 5".
So both of the following sequences are the "same" at the program level:
1 10; 2 4
8 3; 3 9
If YOU perceive an exact match, zero error, hence no "disturbance,"
If YOU perceive a mismatch, there is an error, YOU perceive a "disturbance,"
You are making "Martin's mistake". What I perceive is the state of the
(program) perception; if that state is not at the intended level then it's
not; but the system controlling the program perception doesn't know why the
perception is not at the intended level; it could be due to the effects of
the disturbing variable but it could also be due to my own output (pushing
the button) or it could be due to _both_ disturbance and output at the same
time. The disturbance (the disturbing variable or it's effect on the
controlled perception) cannot be perceived by the system controlling the
YOU appear to have controlled your perception of a sequence of numbers by
"generating a program (sequence of perceptions) of outputs used to push the
Not at all. I have controlled the perception of a _program_ of numbers
by generating outputs that are opposed to a disturbance that changes the
program. There is no program of button press outputs; my outputs, o, simply
mirror the disturbance variable, d, so that o = -d. The relationship between
o and d is not a program (though it could be _described_ as one --ie.
"if o then -d"); but the relationship between o and d is really just a
functional relationship that results from the operation of a negative
feedback perceptual control loop.
How could YOU manage to "push a button" without generating a "program
of [button-pushing] outputs?"
Here is a model of what's going on:
S----> |f| -- > p -->|c|<--r
S is the sequence of numbers occuring out there on the computer screen. The
sequence goes through a perceptual function, f, (in me) that determines
whether this sequence is consistent with the program "if odd, then >5 else
=< 5". The function f is the crucial part of the model; I imagine that it's
output is binary, being 1 as long as S is consistent with the program and 0
otherwise. So p (the perceptual signal that indicates whether or not the
program is occurring) is either 1 or 0 . The reference signal is set to 1 (if
I want to see the program) or 0. When the program perception, p, doesn't
match the reference signal there is an output -- ie. a button press. The
actual sequence, S, that appears on the screen depends on both the
disturbance and the output. In my experiment both are binary variables; the
distrubance is either 1 or -1 and the output is either 1 (button not pressed)
or -1 (button pressed). The product o x d determines whether or not the
sequence, S, is being generated by the program "if odd, then >5, else =<5" or
not. If o x d = 1 then S is being generated by the program; if o x d = -1
then it is not.
Actual implementation of this model would probably require some integration
of variables in order to achieve the proper dynamics. But it should work --
and when it does work, it will be keeping the perceptual signal equal to the
reference signal (approximately) and it will be doing so by generating
outputs that oppose any changes in the disturbing variable. Behavior (in this
case, the behavior of producing a particular program of numbers on the
screen) is the control of perception.
I have not yet built this model; I thought that it might be difficult to
build the function f that detects programs. But, now that I think about it,
it's probably pretty easy to just cluge it up. So maybe I will build this
just to show how a control system can control a perceptual variable that
happens to represent a complex environmental event -- a program.
Where do you put all this within HPCT?
At the program level. Obviously, if I actually built the simulation I would
not be simulating the lower level perceptual functions that produce the
perceptions that go into the program detection function. So my cluge "program
detection" function, f, would unquestionably NOT be perceiving the occurance
of the program the way people do. But I think that would be fine; our line
control models don't perceive line position the way people actually do it
And how do YOU account for YOUR independence?
God just made me that way