[From Bill Powers (2004.06.14.1939 MDT)]
Bill Williams (2004.06.14)]
This is an excerpt from a recent post by you, somewhere just before
06/09/2004 -- I'm on a different computer and don't have the exact date on
hand (forgot to write it down):
The difference between us may be in how this "enviornment" is defined. I
am not sure I understand on first reading your suggestion as to how to
modify the Lattice program-- to do it as you say, "your way." I would,
however, agree that the program as it is now written depends upon an
implicit understanding of the process the program is, in part, modeling.
I am not sure whether or not I can translate the elements which are
implict into code. Writting the lattice program took me about three years.
And, while the result is modestly complex the basic idea involved seems
rather simple. It even appeared to me rather simple to start with, yet
getting to the later versions required an enourmous effort. The problem
you are suggesting, at least initially seems to me to involve elements
that are much more complex than lattice program. So, to start with I don't
have the slighest idea how to go about doining what you suggest. Not,
that I don't see what you suggest as a logical next step.
What I was suggesting was a 10-point outline of a program in which a
perceptual signal was proposed to represent the difference in distances to
neighboring agents in the X and Y directions, and was controlled (at zero
difference, or equality) by varying the agent's position. That is how the
program I called "I5.pas" works, when I finally wrote it and posted it.
You said this to Rick on 4 June 2004, 7:00 PM CST:
There is a good reason you can not see where "the squad members perceive
the distnce." The squad members don't care about "distances." They are
following a rule to position themselves at a spot that is equally distant
from the squad members that they are supposed to form on. The calculation
of this spot is done in the section that begins ...
The squad members in the model don't perceive the equality of distance
from other squad members, either. As you say, the required spot is simply
computed, open-loop, by a separate routine unconnected with the control
systems for position.
I've been looking over some of your previous programs, and it's not clear
at all that you have ever use the concept of control of a perceptual
variable. Most of your programs seem to compute the desired results in the
forward or outgoing part of the system. Your two-component model is an
example, where there is a complex function of prices in the output
functions of four control systems, with no indication of how price
information ever gets into those functions. Price never enters into any
perception. The perceptions that end up being controlled are, as far as I
can see, simply copies of the outputs.
If you will examine I5.pas, you will see that the perceptual signal under
control is computed from the relative distances to the neighbors on the
left and right, or ahead and behind. This is basically what you proposed,
except that there is a perception under control now.
What you dismiss as sophistology is the crux of the principles of negative
feedback control. After 20 years of acquaintance, am I now to conclude that
you have never understood those principles?