drugs; reference levels; demo speed

[From Bill Powers (951117.0600 MST)]

Gary Cziko 951116.2245 GMT --

     "Frans Plooij <fxplooij@EURONET.NL> has just subscribed to the
     CSG-L list (Control Systems Group Network (CSGnet))."

That is good news indeed. Welcome to CSG-L, Frans.


Rick Marken (951116.1230) --

Your posts on drug addiction have proposed some reasonable explanations.
I think that Bill Leach has generalized in the right way:

     If we at least agree that the drug is not the controlled variable
     that has a reference but rather something that happens within the
     person that is a result of ingesting the drug then we at least are
     on a start toward a PCT discussion of this subject.

One factor that hasn't been brought up is the effect on errors that
already exist in the system. People who are in conflicts, either
internal or external, and who have not been able to resolve them, are in
a state of continued intrinsic error and reorganization. As long as
they're reorganizing, they have at least a fighting chance of correcting
the errors. But suppose that what (some) drugs do is to make it appear
that one or more, and perhaps all, intrinsic errors have, in one great
rush, gone to zero. This will stop reorganization in its tracks, and
whatever behavioral organization was in existence at that time will
become permanent until reorganization starts again. The normal
connections between behavior, its consequences, intrinsic error, and
reorganization have been broken. There is a drug "habit" simply because
the organization that controls for ingestion of the drug is no longer
susceptible to change.

This idea would seem to explain a number of facts about addictions.
First, the "withdrawal symptoms." As the (reversible) effects of a drug
wear off, all the intrinsic errors that have gone uncorrected begin to
be experienced again. This, naturally, feels terrible; it is the essence
of what "feeling bad" means. Reorganization will start, and if the drug
is still available, taking the drug again will be one of the likely
results (particularly since the person has now learned how to obtain
it). So kicking the habit becomes extremely difficult: one reorganizes
right back into it.

Another effect of zeroing out intrinsic error signals will be that the
normal basis for learning has been destroyed. Starvation, for example,
will no longer cause reorganization that can be ended only by finding
and eating food. Obtaining sex will no longer be a target of
reorganization. Even pain no longer results in reorganization that ends
with learning how to control variables associated with pain. What we end
up with is a system in which reorganization has been almost totally
suppressed, a system with a fixed organization that, from then on, can
only slowly deteriorate. Failures of control do not result in learning
to improve control. They simply continue.

Different "feel-good" drugs must have different degrees of effect on
reducing intrinsic error. Some may selectively affect reorganization at
different levels in the hierarchy. The worst, of course, are those that
have the most general effect on reducing intrinsic errors of all kinds.
A drug that totally suppressed all intrinsic error signals would
probably be lethal -- although the person would die in a state of
tranquility or euphoria, never feeling that anything is wrong.

If this is a good guess about the effects of drugs, then there might be
a direct relationship with the Dumbing of America and the daily national
consumption of tons of feel-good drugs both legal and illegal.

I should point out one positive effect of feel-good drugs, which David
Goldstein convinced me is valid. If a person is in an extreme state of
intrinsic error, reorganization will be going on at such a rate that
potentially successful reorganizations have no chance to reduce
intrinsic error, and are passed over. In this case, judicial
administration of drugs that can reduce intrinsic error might slow the
general reorganization process enough to allow convergence on solutions.
After that has happened, gradual reduction in the drug dosage can allow
the remaining intrinsic errors to be handled more normally as they
become detectable. This assumes, of course, that the biochemical effects
of the drug are reversible by withdrawing the drug.
Richard Thurman (951116.1330) --

     I am left wondering about the term 'reference level.' Do you mean
     that A) a new level of control (as in the control hierarchy) has
     been created? Or B) a new controlled variable has been created? Or
     C) an existing control system now has a new reference signal. Or
     D) an existing control system's reference signal is now injecting a
     new (perhaps higher) 'level' of input into the comparator?

A good call. Long ago I decided to use the term "orders of control"
rather than "levels of control" specifically to avoid this ambiguity
with the meaning of "level" in the term "reference level." But as time
went by I started slipping back into saying "levels of control" and it
didn't seem to cause any problems, so I relaxed on the terminology. Now
thanks to you I realize that there might be lots of people who have
given "reference level" the wrong meaning just because of this laziness
on my part.

All perceptions can vary from zero to some maximum amount. This applies
even to categorical or logical variables. An animal is either a dog or
not a dog, but at the same time we recognize that some dogs are more
doggy than others. Logically, a proposition is either true or not true,
but as we perceive the world, the applicability of a proposition such as
"all men are equal" varies over a continuous scale. Perhaps this is what
fuzzy logic is trying to capture.

In technical fields, "level" is used to refer to a measure of magnitude.
We can hear a low-level sound or a high-level sound; the water level in
a boiler can be high or low. People can tolerate different levels of
pain. The light-level may be too low or too high for comfort. A bank
requires a high level of honesty from its tellers. So the term "level"
is widely used in the sense that is intended in PCT.

A _reference_ level is simply one of the possible magnitudes of a
perception or the external variable it represents. If the level of a
perception can range from 0 to 100 units, then the reference level might
be any fixed number in that range, like 53. This reference level is
specified by a reference signal that has a fixed magnitude of 53 in the
same units in which the perception is measured. The comparator in the
associated control system would thus receive one signal with a magnitude
of 53, and another signal, the perceptual signal, with some other
magnitude, 0 or 25 or 95. The difference, the reference level of 53
minus the actual level of the perceptual signal, is the error signal.

The reference signal for a control system is set by the output of a
higher-order control system (that is, a control system at a higher level
in the hierarchy, as we have been ambiguously saying). The reference
signal can be set to any level at all, and can be varied over time. The
action of the lower-order control system will cause the level of the
perceptual signal to track the level of the reference signal. If the
reference signal is changing in magnitude, then the lower-order
perceptual signal will be made to change in almost the same way, by
actions that affect the environment and thus the inputs to the lower-
order control system.

It almost begins to feel natural again to speak of "orders of control."

One other matter. I've been maintaining a distinction between a
reference _level_ and a reference _signal_. As I've tried to use the
term, a reference _level_ means the state near which an observable
environmental variable is maintained by the observable behavior of the
system. So when we talk about the reference level of a controlled
variable, we're talking about a physical measurement in the environment,
a publicly-observable phenomenon. The reference _signal_, on the other
hand, is part of a model of the internal operation of the system, a
theoretical entity unless directly measured by instruments. Presumably,
when the external variable is at its reference level, the perceptual
signal has some corresponding magnitude matching the magnitude of a
reference _signal_. The reference _signal_ inside the system has a
magnitude that determines the reference _level_ of the observable
controlled variable. This distinction isn't normally important, but if
we speak consistently in this way we can avoid more ambiguities, in this
case the ambiguity between an actual observation and a theoretical
Gary Cziko --

What is Java?

From other comments, I would agree that any ability to upload an

executable program to run on someone else's machine is too dangerous to
allow. The sickies out there who enjoy destruction would pounce
immediately. All you have to do is include a data section which is
really a virus program, and pass control to it.

Anyway, live interaction on the Web is practical only for those whose
bills are paid by someone else, or who are very, very rich. Use of the
internet, contrary to some impressions, is NOT free.
Bill Leach 951116.22:11 U.S. Eastern Time Zone --

     Ok, that explains what is happening then, the vert. frame rate is
     90 Hz on this machine I believe. That would be pretty fast I
     should think.

In VGA 640 x 480 16-color mode, the frame rate should be around 60 Hz.
In the Demos, the normal duration of a run is 30 sec. The last couple of
interactive demos on the Demo 2 disk should take 1 minute. In Little Man
version 2, the time counter should indicate real elapsed time, pretty
near. So you can check to see just how much too fast the program is
Best to all,

Bill P.

<[Bill Leach 951117.21:36 U.S. Eastern Time Zone]

[Bill Powers (951117.0600 MST)]

Alas, the machine went "out the door" only moments ago (actually it did
not go out under it's own power -- my PCT programming skills have not
reached that level).

The machine has been giving me "fits" whose description or explaination
would only lead me into another "Why I HATE PC-DOS box junk"!

When the machine and I again occupy the same general area again (probably
in a little over a week), I run LM again and watch the clock.


BTW, I do not have any programming languages for that machine other than
REXX (I don't even have gcc for it yet).