[From James Nord (960702. 5p.m. Japan]
me (960628) --
It seems quite acceptable to most everyone on the list to say that
"behavior controls perceptions," but since we are dealing with a
closed loop, and circularity, why isn't it equally correct to say
"perceptions (reference condition) control behavior"?
> Bill Powers (960628.0920 MDT)
>Neither reference signals nor perceptions control behavior, if by
>behavior you mean the output actions of the control system -- what we
>observe it doing.
Of course that's NOT what I'm referring to when I talk about
behavior. You've just described an open loop system. I'm talking about
behavior in a closed loop control system i.e. goal oriented behavior
purposeful behavior.
me 960630,2 p.m Japan, to Bill Powers --
As I understand you, you are saying that those people who see human
behavior as part of a cybernetic closed loop control sytem, but who
manipulate the perceptual feedback portion, do not fit into the PCT model.
>Bruce Abbott (960630. 1020 EST)
In all these cases, what is being controlled is perception -- what the
speech should sound like (or feel like). One can suggest that the actions
required to produce those sounds or feelings are also being controlled, but
this is only an illusion fostered by the fact that in most cases there is
little disturbance acting on the vocal apparatus to distort the sound. When
disturbances appear, the speaker must alter the actions to recover
proper-sounding speech; thus it is the perception that is controlled
(defended against disturbance) and not the actions of the speech apparatus.
All of this seems to me to be perfectly compatible with PCT. Would you
argue otherwise? If so, on what ground?
>Rick Marken (960630. 1130)
I think you are assuming that the changed outputs (changed voicing,accents,
etc.) that are seen in response to "changed perceptual feedback" show
how perception guides or "controls" behavior. What you are actually seeing
is the result of a change in output that is required to keep a perception
under control after a change in the feedback connection between output and
controlled perception. This is beautifully demonstrated in the Tomatis study.
me
Well I can tell that at least some of the people on this list hold
views about control systems, if not PCT, similar to my own, but since I
have not heard directly from Bill after my second posting, I still can not
be sure of his position. One of the reasons for my concern is that, I have
not really seen any reason for all of the verbal snipping that seems to be
going on over what I consider irrelevant issues i.e. verbal
misunderstandings. I would rather get over these issues and get on with
some major concerns of how PCT can be used to help people learn to control
their lives more effectively and efficiently. Let me see if I can lay out
as clearly as possible my position and why many of the issues seem
pseudo-issues (linguistic confusions) to me.
First, the issue of behavior controlling perception vs perception
controlling behavior. If we are truly involved with a closed loop and
causal circularity, I can not see the distinction, except as a point of
view. Sometimes one point of view is useful, and sometimes the other point
of view is useful, but neither in my mind is "incorrect" if we maintain the
closed loop circularity as central. Wiener's original work in this area
involved perception controlling behavior in the form of homing torpedoes.
By consultation with his physician colleague Rosenblueth, he came up with
the closed loop feedback control principle, but applied it first to the
control of goal oriented purposeful behavior, i.e. hitting the target. In
almost all of the arguments I have heard which reject the premise that
perception also controls behavior, consciously or subconsciously, the
writers forget the closed loop aspect and reintroduce the open loop
phenomenon. Of course, perceptions do not control behavior if we are
talking about an open loop. But if we maintain the closed loop aspect, why
can't we see it from two different points of view?
Next! For me there are two fundamental types of systems we are
dealing with: power systems and communication systems. Power systems
involve matter-energy events, and communication systems involve information
events. I still hold to the distinction Wiener made between these two
systems, i.e. when you couple two power systems, the system with the
greatest energy dominates the system with the lesser energy. (Big bullies
can physically man handle little guys).
Communication systems, on the other hand, are the reverse. When
two communication systems are coupled, the receiver is always in control.
This is because the concept of information (the "stuff" of communication
systems) is receiver (Observer) defined. I will get into that whole issue
later, but I wanted to get this far in order to establish what I see as
control systems, which are for me, a coupling of a power system and a
communication system.
A heat control system consists of a thermometer sensor
(communication system) and a furnance (power system). The thermometer
sensor can detect a difference, that makes it a sensor. But the
thermometer is coupled with a furnance such that it requires a specific
difference to actually trigger any action by the furnance (either on or
off), and that difference, "the difference that makes a difference" is
information in the view of Bateson (one of the members of Wiener's Think
Tank group) The thermometer can not detect electrical voltage, it can not
detect dust in the atmosphere, it can not detect darkness or light,
radiation or humidity although all of these may be in its vicinity. The
information is defined by the thermometer sensor in terms of the heat
difference which is detectable by it, and by the physical coupling to the
furnance which determines the "amount of heat difference" that makes a
differences (triggers the change in the furnace state).
The human body is a control system because its sensors (eyes, ears,
nose, touch, etc.) are all coupled to the physical matter-energy system
consisting of bone, muscle, oxygen supply, glucose, etc. The system
evolved over the millenium to maintain certain critical variables (such as
body heat, oxygen flow, etc) in a relative steady state. The ultimate
purpose of the control system was survival. Because it is a control system
consisting of a communication aspect (the perceptual sensor, reference
condition aspect) and a physical power system, (the physical human body) it
must operate both according to the laws of physics and according to the
"laws of communication" (information theory). This means that there are
predictions made within the communication system, and there is ballistic
inertia which must be accounted for within the physical system. In order
to deal with the world, we must have perceptual foresight, not hindsight.
An example to illustrate my meaning. When a baseball players
stands at the plate with his bat poised, he watches the pitcher wind up and
throw. He makes a calculation (prediction) as to the probable perceptual
path of the ball, and then swings to met the ball where he expects it to be
in the time it takes to get his bat there. The laws of physics combined
with the perceptual reaction time means that the bat will be following a
ballistic path (open loop, uncontrolled) for at least a short period of
time. This is clearly evident when the umpire calls a strike. This
experience should be common to most.
The experience that etched this need to recognize both
communication and power system variables came to me as I watch the 16" gun
turret of the USS Missouri shake back and forth almost tearing the ship
apart because the gun fire control officer was swinging his perceptual
sensors too quickly, (as part of a closed loop control system) creating
signal mismatches, positive feedback loops, hunting and oscillation not
easily forgotten. This was a mechanical control system, and humans have
evolved a much smoothly running control system, so we don't notice these
problems until we are faced with control disfunction in the human such as
with stuttering or palsy.
Humans are hierarchical control systems. Indeed all systems can be
viewed as part of a hierarchy of systems, meaning that at any level,
including at the organism level we are dealing with a holon, a system which
has split loyalities, i.e. one part wants to be an independent system,
while the other part wants to be a part of a larger whole. There is a
mini-max game played out within all systems, (independence vs cooperation)
and because the stable mini-max points are almost never the same, there is
a range of "playing room" within which each system operates. Americans
tend to operate toward the independent side, while the Japanese tend to
operate toward the cooperative end of the mini-max. A cancer is a cell
system which has gambled everything on being an independent autonomous
system.
I bring this up here because most of what I have seen in the
discussions on this list, seems to see control systems as autonomous
systems. I believe that at the higher levels within humans, certainly at
the system levels, there is a reference condition which is oriented to the
survival of the larger unit, the family, the community, the nation, etc. as
well as the survival of the individual organism. These higher order levels
take a long time to develop. According to Piaget, they go through certain
growth "reorganizations" (acommodation) as they add higher and higher
levels of hierarchical control through maturation and learning. Most of
this "growth" takes place in my view through the coupling of communication
systems, i.e. two people talking to each other. The problem with this is,
however, clearly expressed by von Foerster, when he said,
"We seem to be brought up in a world seen through descriptions by
others rather than through our own perceptions. This has the consequence
that instead of using language as a tool with which to express our thoughts
and experience, we accept language as a tool that determines our thoughts
and experiences."
I want to express my experiences, so don't accept anything I say
that doesn't fit yours. But for me, almost all of the dispute on
"information" seems to be a linguistic tendency to accept nouns as
substances rather than as processes. We have seen this happen with the
word "calorie" constantly. People are told that they should only take X
number of calories if they want to lose weight. We are continually
informed that there are X number of calories in a unit of food. All of
this is linguistic nonsense because a calorie is not a substance, but a
measure of heat potential, and that heat potential does not become
actualized heat until the substance, i.e. the food, is burned. Even then
"initial potential" is only that, a potential. The actual measure of heat
which results, depends not only upon the "initial potential" in the food
substance, but also upon the furnance in which it is burned, the amount of
oxygen available at the time of burning and the effeciency of heat
transfer, etc. It may be a handy shorthand, to consider a calorie as a
substance, just as we find the expressions "sunrise" and "sunset" as handy
shortcuts. But it only works as long as we as receivers continually remind
ourselves of its real meaning. After all, meaning is in people not in
words. And thus, meaning is under the control of the receiver, not the
sender. But because senders have habitually used the nominal term
"calorie" as a substance which can be moved, ingested, transfered etc. we
receivers sometimes tend to forget the original process meaning of the
word. This is precisely, in my view what has happened to the discussion
on information. In his challenge, Bill refers to information as a
substance.
"2. The information directly available from the state of the essential
variable is denied to the regulating person."
Correct me if you believe I am wrong, but for me, the technical
use of the term information is as a measure, not as a substance. It is
therefore an observer's term, i.e. it is a perceptual result, not something
to be perceived. It is a measure of a difference and that measure requires
a sensor capable of measuring that difference. One thing is clear, at
least to me, information (or variety, which is Ashby's term) is not a
physical substance. It can not be carried, moved, weighed, bought used or
sold. It doesn't exist, except to a sensor capable of detecting the
difference. The capability of the sensor to detect differences is an
inherent part of the definition of information. Hence in Bill's challange,
if the "information" is denied to the receiver, then obviously there IS no
information. Hence there is no challenge to the the information theorists.
In this regard, the discussion about the thermostat makes me ask
the question, can a thermostat detect a difference in temperature. If so,
it involves information, i.e. the detection and measurement of differences.
Does a control device have a sensor which detects and measures a
difference between a reference condition and another signal? If so, then
it involve information. While Shannon worked primarily with discrete
units, and was therefore more into number theory and counting, rather than
measuring which is required for analog data, the theory, i.e. information
theory can be applied to anything which detects and measures differences.
And, if done very carefully, it can often be useful. Information theory
provides a metric where in we can talk about the number of alternatives
from which the specific single choice was selected. As Ashby puts it,
informationally we as observers always operate in sets of possible
alternatives, our senses then inFORM us as to which one actual occurred.
Some people refer to this as a reduction of uncertainty.
For example, as English speakers (listeners), we have been taught
to detect 26 distinctly different letters of the alphabet. The detection
of any one of those letters is then measured as a choice from a set of a
total of 26 alternatives. Shannon used the log scale to provide the
measuring scale. Each letter (for native speakers of English) is then said
to be potentially less than 5 bits of information. Since the Japanese
have to know over 2000 kanji, each kanji has a potential information
measure of around 11 bits. Learning new languages, increases our capacity
to detect differences. It increases our total "information" or our
variety.
Sheldon body types, heighth, weight, blood types and when we were
born all effect our sensory information processing capacities and hence our
control systems. There are physical differences in the way we process
oxygen. "Information" from our genes determine whether we can be marathon
runners or not. This "information" effects our control system, but it is
not information we "use". These are shorthand expressions. They are useful
only in recognizing that if we want, we can get more details about how
these differences effect the different control systems. In practical
terms, information measures are simply potentials of processes and the
actual information is determined by the communication system detecting
these signals.
The power of information theory however is that it is a
mathematical theory. Mathematics is a way of making simple concepts
complex in order to allow us to manipulate the data in ways we could not do
otherwise. But just as in number theory, manipulation is one thing,
practical results is another. 2 + 2 = 4 is only an informational
(communication) abstract manipulation. We should not confuse numbers and
numerals or the manipulation process with the final usage. When we finish
our calculations, we must put a dimension on the process (two chairs plus
two chairs) to make it useful in the physical matter/energy world. I
repeat, information is a process and therefore always involves a receiver,
an observer, a detector, a sensor. It is not a substance that can be
stored or transferred inspite of its nominal status in our language. If we
keep this in mind, I can see no conflict with using information theory
manipulations (where appropriate) in PCT.
My belief is that as one goes up the hierarchy of control, as we
learn more, as we develop and use instrumentations to detect new
differences, there is a greater and greater variety of detectable
differences we are potentially aware of. Complex human control systems
thus have more information (capacity to detect and process differences)
[variety] than simple single cell amoebas. The work of Tomatis and others
who manipulate the feedback variable are activities in which the
information variable of the communication side of a closed loop control
system is being varied. This has a tendency to change over time, the
perceptual capacity to process information, and hence a capacity over time
to change behavioral patterns. The people who use the acoustical filters,
do not just change their behavior temporially to match the perception they
want to hear. The use of the acoustical filters, can over a period of time
teach them to perceive new sounds that they were unable to perceive before.
Once learned, this new perceptual capacity allow them to speak a new
language without accent without the aid of an acoustical filter. Thus by
varying the informational quality of the sensors, new behaviors can be
generated. In other words, by looking at the closed loop from a perception
to behavior perspective, we can get a better understanding of the learning
process. If we have no concern for learning, changing behavioral patterns
or the like and are only concerned with "proving" that behavior controls
existing perceptions, then we can look at the system by holding the
perceptual side constant. If the sensor side, the perceptual side is held
constant, and the behavior varies, as in "The Test" then there may well be
very little information involved i.e. differences detected by the
controller.
However I repeat, is that the only way to look at a closed loop
control system? Is it always the best way? Is there not some benefits
from shifting our viewpoint as long as we keep the loop closed? To
summarize, whenever the arguments about information come up, there seems to
be a tendency on the part of some to treat it as a substance, which
automatically although unconsciously causes a switch to an open loop
process, or to a communication-communication couplings. This prevents the
viewer from seeing information as the metric of the perceptual
(communication) side of a closed loop information/energy control system.
Why can't we can look at a control system from both the behavioral
(matter-energy physical) side and the perceptual
(informational-communicational) side, as long as we keep the loop tighly
closed. I personally think there is an advantage in being able to see a
system from more than one point of view.
I have already expressed far more than I ever intended to say in
this discussion group, and I am sure it will evoke some negative reactions
from some, but I also hope that it will provide some "food for thought"
(tested against your own experiences). I have expressed what I have
expressed as one communication system to another. I have used the words to
try to express my experience as best I understand them. The final
responsibilty for understanding their meaning is however yours, the
receiver, since you control your own input. In conclusion, I believe the
members of the discussion group would gain by from holding very tight to
the idea of control as a closed loop circular process, but loosing up on
the vantage point from which it is viewed, and remembering that final
control over the communication of ideas is in the heads of the receivers.
. And don't forget to use the outstairs instead of the elevator, its
better for your health:-)
James R. Nord
Nanzan University
Nagoya, Japan
NORD@axia.ic.nanzan-u.ac.jp