orders of vs levels in HPC

[From: Bruce Nevin (Tue 920714 13:14:09)]

To sum up: your argument seems to depend on the emergence of higher levels
from populations of systems of an existing level, without the addition of
any new kinds of physical systems. I would claim that you are relying on
IMPLICIT organization to create new levels. I, on the other hand, argue
that the new levels must be EXPLICIT.

I take it you mean orders of control systems (not to be confused with
levels of control within one hierarchical control system).

Each new order is explicit from a higher-order perspective (the
perspective of the new order or of a yet higher order), but not from the
perspective of any orders out of which the new order is constituted.

For example, a reference signal is explicitly present for a cell as an
electrical potential, ion concentration, whatever, but it is not a
reference signal for the cell. The ion concentration is explicit for
the cell. The reference signal is not explicit, and cannot be, because
reference signals as such do not exist in the cell's universe. As the
cells (by whatever evolutionary process) come to constitute control
systems of a higher order, an ion concentration within a cell can take
on a new identity as a reference signal (or error signal, etc.), in
addition to its value as a variable within the cell. All of that is
invisible to the cell, which is only controlling its own variables in
its own terms.

I am not claiming to demonstrate the existence of social hierarchies. I
am arguing for agnosticism regarding them.

I can flesh this out with more words. Do I need to? You say.

  Bruce
  bn@bbn.com

[From: Bruce Nevin (Wed 920715 13:05:23)]

(Rick Marken (920715.0900) ) --

You are very close to a point that I was trying to communicate.

You claim that the relationship between adjacent orders of control
systems (such that CSs of order n are constituted of CSs of order n-1)
evidently must include the following:

  Variables used to implement control for order n must not be
  controlled variables for order n-1.

The level n variable is not controlled on level n-1, and it is a
cause-effect property of the cell's behavioral outputs on level n-1.
As you say, a neuron cell

generate[s] axon potentials (spikes) at a rate proportional to the
integrated (over time and number of dendritic inputs) electrical charge
at the cell body. This functional property of the neural cell ( rate of
spiking proportional to charge on cell body) is not "controlled" by the
neuron itself; it is a cause-effect property of the cell's activity.

The relationship of ECS function to intracellular control, you claim, is
an incidental byproduct. It is only the physics of the cell's body
(level n-1) that the ECS (level n) uses, cellular metabolism being only
used to maintain the viability of a segment of "wire" in place.

But it is a byproduct of behavioral outputs ("the cell's activity")
which presumably are variable means for achieving uniform results that
matter to the cell. A cell in a given state controls a disturbance of a
given sort reliably with behavioral outputs of a corresponding given
sort. These behavioral outputs (including internal changes) can have
cause-and-effect consequences remote from the controlled variable and
the disturbance, which are not themselves controlled. Thus, it is
possible that these changes in the cell's observed behavioral outputs
are cause-effect byproducts of controlling other internal variables
(such as Na concentrations) against disturbance. Exploring this
possibility might lead to some explanations of how learning and
reorganization work.

I am proposing (920709 09:13:52) that reorganization is carried out in
populations of entities of order n-1. If control of order n results in
chronic error in CSs of order n-1, then the CSs of order n-1 control to
correct the error, with behavioral outputs which for a nerve cell might
include growing new axons, detaching or moving or withdrawing
(atrophying?) existing ones, changing receptor sites around, etc. Local
error in a few order n-1 CSs results in learning. Error in many CSs
results in reorganization.

A neuron, as a cell, is probably busy controlling many variables -- such
as concentrations of K+ and Na- ions, etc. The systems controlling these
variables are made out of cell components (like RNA and DNA molecules)
that may be control systems themselves. But one thing a neuron cell does
That is, the variables involved in this functional relationship . . .
are not (as far as I know) perceived and controlled by the cell. For
example, the cell does not have a preferred (reference) spike rate that
it tries to maintain; it just fires at a rate dependent on the charge at
the cell body . . . . It is this input-output characteristic of the
cell's electrical behavior that makes it a useful component of a control
system. The cell responds to input cell body charge with a certain rate
of firing; this is a "dedicated" cause-effect characteristic of the
cell; the cell cannot change the way it responds (firing rate) to input
(cell body charge) -- there is no control involved in this functional
relationship; that is what I mean by functional specificity. In terms of
it's electrical response to electrical stimulation of the dendrites the
cell functions like a wire in a circuit ( with firing rate the analog of
current and cell body charge the analog of voltage). A control system
must be built out of such "functionally specific" components.

To this I would add that the cell does not *want* to change the way it
responds. The firing rate per se does not matter to the cell. Indeed,
it probably does not even "know" that it is changing its electrical
potentials, that these changes constitute "spikes," and that they are
occurring at a variable rate. All of that is invisible to the cell,
*and* *must* *remain* *so* for the higher-order function to maintain its
integrity, as you also point out.

Substitute humans for cells:

A person is busy controlling many variables. The systems controlling
these variables are made out of cells, neural structures, and organs of
perception and execution (probably there's a better word, but I'm in a
rush). But one thing a human does is change the color of its aura.
That is, the variables involved in this functional relationship . . .
are not (as far as I know) perceived and controlled by the human. It is
this input-output characteristic of the human's auric behavior that
makes it a useful component of a control system. The human responds to
changes in color of a neighboring human's aura by changes in its own
aura. This is a "dedicated" cause-effect characteristic of the human;
the human cannot change the way it responds (aura color) to input (aura
color) -- there is no control involved in this functional relationship;
that is what I mean by functional specificity. In terms of it's auric
response to auric stimulation of the etheric body the human functions
like a wire in a circuit ( with auric change rate the analog of current
and color the analog of voltage). A control system must be built out of
such "functionally specific" components.

Exploring this possibility might also suggest ways of explaining input
functions and output functions. Must these be separate multi-cellular
structures, or might the metabolism of a single ramified nerve cell be
such that it is not a simple cause-effect "wire" passing neural current
through, but is actually doing the weighting (and the changes of
weighting) of signals? In the case of an input function, the weighting
and changes in weighting of input signals that get combined in the
unified output signal; in the case of an output function, the weighting
and changes in weighting that get applied to the different copies of the
input signal in the process of making them into specific output signals.
In this case, the cell is controlling variables that matter to it, and
as a byproduct differentially weighting electrical potential, which does
not matter to it, in its several branches. The electrical potential in
the dendrite of another cell, on the other side of a synapse, does
matter to it and may disturb variables that it controls; the rate of
peaks "firing" in itself does not matter to it, and is a byproduct of
that control.

People can temporarily arrange themselves so that they function as a
control system -- this is true and I've seen it happen. So in this sense,
social control systems can exist; but these systems are quite transient.

Such things are not control systems of an order above that of humans,
for the simple reason that humans (level n-1) control for creating and
maintaining them (level n). If there are such supra-human organisms,
the variables that matter for them are incidental for us. It may be
that we have a craving to belong to groups and get off on group
participation that is working well (and are frustrated when it is not)
because some aspects of a transhuman organism's function require human
intercommunication as a vehicle. But human intercommunication itself
can be of no import to it, nor can its functioning have any import to
us. Except when it is experiencing error and conflict (chronic error).
That's when we reorganize. Maybe that's when we try to create or change
social institutions.

  Bruce
  bn@bbn.com

[Martin Taylor 920715 15:00]
(Rick Marken 920715.0900, Bruce Nevin 920715 13:05:23)

I posted 920710 14:15 a pointer to an article that to me suggests that the
whole function of an ECS (possibly more than one) can be performed in a
single neuron, plus the control of gain that Bill Powers thinks of as an
unnecessary complication. The reference, once again, is:

"Evidence for a computational distinction between proximal and distal neuronal
inhibition," E.T.Vu and F.B.Krasne, Science, March 27 1992, 255, 1710-1712.

Maybe you don't see in this article as much as I do, but as I said in my earlier
posting, I think all of the elements are there. If this is true, then Bruce's
analogies to human-based control systems take on more force.

Rick, I don't think humans can control other humans by force. All they can
do is to alter the range of control available to the other humans. But humans
CAN fairly reliably get other humans to perform actions that affect the world
in such a way that their (the masters') percepts come closer to their
references. Usually, all one has to do is to ask, but there are more subtle
ways. In a normal PCT-world mirror diagram, one would say that an ECS
in the master is controlling a perception that the slave is doing something,
and that ECS in the master is being provided a reference by another ECS that
is controlling a perception of something affected by the slave. If for master
and slave one reads partner A and partner B, one still comes up with control.

The master does not control the slave, no; I do not control my keyboard. But
both slave and keyboard are aspects of the world that provide me with percepts
I can control. That my actions differ in achieving the same result on different
occasions is fully in accord with normal PCT, and so is the fact that they
sometimes fail because of external world disturbances or barriers that go
beyond my range of control. It is clear that no other human can be part of
MY hierarchy, but the perception of what another human is doing can be the
signal in an ECS that IS in my hierarchy.

In other words, I think that the whole discussion of social control is and
has been following a kipper dragged across the trail.

Martin

[From Rick Marken (920715.0900)]

Bruce Nevin (Tue 920714 13:14:09) says:

As the
cells (by whatever evolutionary process) come to constitute control
systems of a higher order, an ion concentration within a cell can take
on a new identity as a reference signal (or error signal, etc.), in
addition to its value as a variable within the cell. All of that is
invisible to the cell, which is only controlling its own variables in
its own terms.

I think that what is required for the cell to participate as a component
of a control system is what I will call "functional specificity". To avoid
high-falutin' language, let me just say why I think neurons work as
components of control systems and why people don't (even though neurons, like
people, are probably control systems in and of themselves). A neuron, as
a cell, is probably busy controlling many variables -- such as concentrations
of K+ and Na- ions, etc. The systems controlling these variables are made
out of cell components (like RNA and DNA molecules) that may be control
systems themselves. But one thing a neuron cell does is generate axon
potentials (spikes) at a rate proportional to the integrated (over time
and number of dendritic inputs) electrical charge at the cell body. This
functional property of the neural cell ( rate of spiking proportional to
charge on cell body) is not "controlled" by the neuron itself; it is a
cause-effect property of the cell's activity. That is, the variables involved
in this functional relationship (electrical charge on the cell body, spike
rate) are not (as far as I know) perceived and controlled by the cell. For
example, the cell does not have a preferred (reference) spike rate that
it tries to maintain; it just fires at a rate dependent on the charge at
the cell body (up too the limit of saturation -- the cell just physically
cannot produce spikes faster than a certain rate regardless of the input
charge). It is this input-output characteristic of the cell's electrical
behavior that makes it a useful component of a control system. The cell
responds to input cell body charge with a certain rate of firing; this is
a "dedicated" cause-effect characteristic of the cell; the cell cannot
change the way it responds (firing rate) to input (cell body charge) --
there is no control involved in this functional relationship; that is what
I mean by functional specificity. In terms of it's electrical response to
electrical stimulation of the dendrites the cell functions like a wire in
a circuit ( with firing rate the analog of current and cell body charge
the analog of voltage). A control system must be built out of such
"functionally specific" components.

People could also be components of control systems; but the only aspects
of human behavior that could be a component of such a system are those
that are "functionally specific" in the same way that the cell's electrical
behavior is functionally specific. In other words, only cause-effect
aspects of human behavior could be a reliable component of a control system.
Controlled aspects of human behavior could do the job -- but this would
not be reliable because the reference for the controlled variable could be
changed in a way that bombs the function of the control system of which
the person is a component. For example, suppose that a control system
depended on people (as the components) producing a response that was
proportional to some perceptual input. Consider this "social control system";
oneperson (the sensor) responds "yes" when a lion appears and "no" otherwise.
A second person (the comparator) says "help" when he hears "lion" and
nothing otherwise. A third person (the output) shoots a gun into
the air when he hears "help", and does nothing otherwise. People can set
their refereces so that they act this way -- but they are free to change those
references at any time (free in the sense that nothing outside of the
person controls the setting of the reference -- directly, anyway). Most
obviously, any of the people i this "social control system" could decide
to leave (heeding natures call) thus changing, substantially their
input/output function in the control system.

I am not claiming to demonstrate the existence of social hierarchies. I
am arguing for agnosticism regarding them.

Agnosticism is what science is all about. I'm certainly willing to believe
that social control exists if it is demonstarted to me -- otherwise, I go
with the null hypothesis, which is based on my understanding of how
control works and how people work as control systems. This understanding
leads me to believe that the only way that people can be components of
a control system is in terms of cause-effect aspects of their behavior.
And there is not much that you can just cause people to do without
the use of extreme physical force -- and force is the typical way that
people are made to function as components of a social control system. And
even force doesn't succeed for long.

People can temporarily arrange themselves so that they function as a
control system -- this is true and I've seen it happen. So in this sense,
social control systems can exist; but these systems are quite transient.
I don't think you would want your nervous system to work the way a social
control system works.

Best regards

Rick

···

**************************************************************

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

[From Rick Marken (920715.1300)]

Bruce Nevin (Wed 920715 13:05:23) says:

Substitute humans for cells:

A person is busy controlling many variables. The systems controlling
these variables are made out of cells, neural structures, and organs of
perception and execution (probably there's a better word, but I'm in a
rush). But one thing a human does is change the color of its aura.
That is, the variables involved in this functional relationship . . .
are not (as far as I know) perceived and controlled by the human. It is
this input-output characteristic of the human's auric behavior that
makes it a useful component of a control system. The human responds to
changes in color of a neighboring human's aura by changes in its own
aura. This is a "dedicated" cause-effect characteristic of the human;
the human cannot change the way it responds (aura color) to input (aura
color) -- there is no control involved in this functional relationship;
that is what I mean by functional specificity. In terms of it's auric
response to auric stimulation of the etheric body the human functions
like a wire in a circuit ( with auric change rate the analog of current
and color the analog of voltage). A control system must be built out of
such "functionally specific" components.

I agree with this analogy. If, while controlling, people give off an aura
output that is lawfully related to an aura input (where aura is some
unperceived variable and where the lawful function relating auric input
to output in some way depends on the structural and functional charateristics
of the human control system) then this auric function could be used as part
of a meta control system, with "people" as its components. But this control
system would be pretty much invisible to us.

Actually, if the neuron analogy is carried through, then aura is equivalent
to cell potential -- and this should be perceiveable by the system itself;
people should be able to perceive (possibly with instruments) the aura just
as the neuron (if it had any brains) could perceive its cell potential. The
neuron could even measure the cell body charge to spike rate transfer
function; and I think people would be able to measure an auric transfer
function if it existed (indeed, this auric function might be a relationship
between any ol' physical variables, as long as it was a cause effect relation-
ship that depended on the presence of a person (as the cell charge/spike
relationship depends on the presence of the neuron. The neuron
might even be able to figure out that it was a part of a large control
system (the nervous system) just as people might be able to figure out
(or, at least fantasize) that they are components (or pawns) in a higher
order control structure. But I don't see any way that the neuron could figure
out what the control system of which it is a part was controlling; the
variables we control would be "cognitively impenetrable" to the individual
cells that make up the nervous system. A "perceptual" neuron would never
be able to tell that its firing rate was proportional to the degree of
"squareness", say, of an image on the retina ("what the hell is a retina?",
says the neuron). So if people are part of a higher order control system
(which uses uncontrolled "auric" relationships are components of control,
we MIGHT be able to figure out that this is the case (A BIG MAYBE) but
even if we do I don't think we could figure out what the hell that
control system was about; what would it matter, really, anyway. Why would
a neuron care that it's part of a control system that controls what
its owner (us) calls "the real world". All it cares about is whatever
variables a neuron cares about.

So I think the kind of "meta control" that Bruce is talking about here is
a bit deeper (and more impenetrable) then "social control". Meta control
would involve control of variables that we (humans) will never know about--
placing them outside the realm of science. Social control, to me, implies
a system that controls social variables -- like relationships between
people, programs of action involving several people, etc --and these
variables are easy to perceive. It is also easy to test to see whether these
variables are controlled by anything other than the people involved (in
which case there would be evidence of some kind of superordinate "social"
controller). I think the evidence suggests that there is no such social
controller; consideration of the existence of a meta controller I leave
to the Pope (when he feels better).

Regards

Rick

···

**************************************************************

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