I was typing on line to MCI mail and a part got left off in reformatting
/ sending. Here is corrected retransmission.

From Dag Forssell [920326 - 2]

Gary asked about "Tightly coupled". Imagine that you are playing the
rubber band demonstration with a strong machine, programmed to go through
a set pattern of motion.

You have no difficulty, since the machine only influences the position
of the knot, through the tension in the rubber band.

Now "tightly couple" the knot to the machine by substituting a stick (or
a rope as I like to do when demonstrating conflict between two
"pullers"). If you still are connected to the knot by a rubber band on
your side, you will pull in vain.

If the stick is extended to your hand, you will be pulled along,
powerless to do otherwise. You are being controlled by the machine with
overwhelming physical force, the only way Bill says you can be

I believe it is important to remember one of the hallmarks of control
systems: amplification. This term does not communicate well. I am
shifting my language to "the direction of resources" or something like
that, with emphasis on resources. Your heating system at home opens a
valve to release (and ignite) a stream of natural gas. The stream is not
finely calibrated, but has a powerful influence on the air temperature.

If you have an air conditioner working at the same time, set at 68
degrees, while the heater is set at 75, the two will pull (with tight
coupling) on the air temperature knot with as much influence as each is
capable of.

If the gas line has the capability to release more resources to raise the
temp than the air cond has resources to lower it, then the air
temperature will stay at 75 degrees.

The rubber band is such a marvelous tool, because it shows influence
without tight coupling. Try the demonstration with a rope, and two dots.
One towards the left as a target for the left person and one a little to
the right (one foot apart if you are at a blackboard with a 4 foot rope,
which works best, one inch if you are on a paper with a short string) as
a target for the person on the right. See which person is willing tu pull
hardest. This person will pull the knot to his dot and keep it there.
This will illustrate the heater / air cond conflict.

A detail: The heater and the air cond are separate control systems, as
in wall heater and window air cond, with separate thermostats.

Hope this helps. I enjoy your questions.


[From Rick Marken (920601 15:30)]

I wrote k.e when I meant k.f. So it should read:

K = k.o * k.f

Then k.o is in units of output/input; k.f is in units of input/output.

This makes me realize that the main source of disturbance to high level
variables is probably the system itself! Maybe it's because I make and then
correct so many mistakes when generating intended sequences, programs, etc
that I am acutely aware that these must be controlled variables -- and not
generated outputs?

Best regards




Richard S. Marken USMail: 10459 Holman Ave
The Aerospace Corporation Los Angeles, CA 90024
(310) 336-6214 (day)
(310) 474-0313 (evening)

In re Michael Fehling 930909 8:46 AM PDT --

In hurriedly typing my short reply to Tom Bourbon I garbled the last paragraph.
I apologize for that. Here's what I meant to say.

...I urge you and others on CSG-L to look up my commentary (and that of others
in that issue of AI journal.) Incidentally, my article also contrasts
Newell's S-R view of how to explain behavior with the Noam Chomsky's idea of a
"competence" account. Since Avery Andrews is a linguist he might find this
worth reading.