Is PCT Consistent with the concept of energy & entropy?

CSGnet Archive CSG2008
1

(Gavin Ritz 2008.03.02.20.01NZT)

If PCT is a perception theory, then it should be consistent with the
concepts of entropy and energy. Then is energy and entropy a perception.
(I'm on the programme level now).

How can this be?

I think that this is so; both entropy and energy are perceptions, (at the
systems level). Try measuring energy and entropy directly, it's not
possible.

I argued this with a professor of physics from Caltech I think he was; he
said that of course we can measure energy directly. So I challenged him to
show me how.

He told me by about a semiconductor called Mercury Cadmium Telluride (MCT).
In fact I had started such a company producing MCT some years earlier. And I
told him we only measure the electrical current or voltage and then make a
calculation about the energy. He never did concede though and carried on
about the Copenhagen Interpretation which funnily enough is also a
perception but an abstraction on the category level (actually higher than
the systems level).

There is no way to actually measure energy or entropy directly so, what we
have is that these are perceptions, rather clever ones though that all work
out quite nicely.

Regards
Gavin

2

[From Rick Marken (2008.03.02.1110 PST)]

(Gavin Ritz 2008.03.02.20.01NZT)

If PCT is a perception theory, then it should be consistent with the
concepts of entropy and energy. Then is energy and entropy a perception.
(I'm on the programme level now).

How can this be?

Everything is perception. All we have is perception. Energy and
entropy are perceptions at the category level: they are words that are
labels for many different observations (also perceptions). In physics,
there words are presumed to refer to actual entities in the real
world; the world that is presumed to be the basis of our perceptual
experience. We'll never know whether there is really energy and
entropy out there on the other side of our experience. Science (from
my perspective) is about providing models (also perceptions) of what
might actually be out there, in the reality on the "other side" of our
senses. To the extent that these models accurately predict the
perceptions that occur in carefully defined circumstances
(experiments) then we believe that those models represent (for the
time being) what we think of as "reality".

That's basically my epistemology, anyway.

Best

Rick

···

--
Richard S. Marken PhD
rsmarken@gmail.com

3

(Gavin Ritz 2008.03.03.1403NZT)

[From Rick Marken (2008.03.02.1110 PST)]

(Gavin Ritz 2008.03.02.20.01NZT)

Rick
I totally agree with your epistemology, I have had a hard time convincing
people about it over the years.
Regards
Gavin

If PCT is a perception theory, then it should be consistent with the
concepts of entropy and energy. Then is energy and entropy a perception.
(I'm on the programme level now).

How can this be?

Everything is perception. All we have is perception. Energy and
entropy are perceptions at the category level: they are words that are
labels for many different observations (also perceptions). In physics,
there words are presumed to refer to actual entities in the real
world; the world that is presumed to be the basis of our perceptual
experience. We'll never know whether there is really energy and
entropy out there on the other side of our experience. Science (from
my perspective) is about providing models (also perceptions) of what
might actually be out there, in the reality on the "other side" of our
senses. To the extent that these models accurately predict the
perceptions that occur in carefully defined circumstances
(experiments) then we believe that those models represent (for the
time being) what we think of as "reality".

That's basically my epistemology, anyway.

Best

Rick

···

--
Richard S. Marken PhD
rsmarken@gmail.com

4

[From Martin Taylor 2008.03.02.10.33]

(Gavin Ritz 2008.03.02.20.01NZT)

If PCT is a perception theory, then it should be consistent with the
concepts of entropy and energy.

Absolutely. In fact, I think it is the other way round. My undergraduate background was in Engineering Physics, and I considered doing graduate work in control systems before becoming an Experimental Psychologist, which probably is explains something of my particular approach to PCT.

My take on it is that our current understanding of thermodynamics forces PCT as the underlying basis for all life, whether that life be carbon-based or in some other exotic form nobody has yet imagined. Looked at this way, perceptual control is basically a fundamental thermodynamic concept, as well as a psychological one.

Why should this be?

In a closed environment, entropy increases by equipartition of energy among the various degrees of freedom in the system. If the system starts with some part of it in a structured state, eventually that structure will dissipate into a higher entropy configuration, balncing its environment.

There are two ways a structure can survive: be shielded from the environment, or export entropy to the environment as fast as the environment exports entropy to the structure. Both imply a separation between what is and what is not part of the structure. The first implies placing the structure in its own closed universe, and even there, if there is any energy in the small closed universe, the entropy will tend to equipartition, and the structure will dissipate, if slower than it would have without the shell. The second way is more robust, and defines PCT.

What PCT means is simply that something within the structure is more stable (thermodynamically colder) than it would be if the structure did not act on its environment. That "something" is a perceptual signal, but its stability is not the important point. The important point is that by acting on the environment to maintain one or more perceptual signals more stably than they would otherwise be, the structure is itself maintained against entropic decay. The structure can be represented by a set of "intrinsic variables" that are the elements that are important for survival. The controlled perceptions are important only insofar as by stabilizing them at suitable (varying) levels, the system actuallt stabilizes the intrinsic variables. The system imports energy at low entropy from the environment ("food", in the general sense) and exports it at higher entropy ("waste" and heat).

Then is energy and entropy a perception.

They are, in the minds of physicists. That's why I said "our current understanding" above. But if our perceptions given those labels relate to the real world in the way they seem to do, then they are central to the concept of PCT, and PCT is central to all life.

(I'm on the programme level now).

How can this be?

Why not? "Democracy", "lateness for an appointment", "blue", churlish behaviour", "words on a page" and "mis-spellings" are all perceptions. What's the problem with "entropy" and "energy"?

Martin

5

(Gavin Ritz 2008.03.03.21.19NZT)

What's the problem with "entropy" and "energy"

Martin

None at all we are in complete agreement about Entropy and energy, you and
Rick are the first to agree with me about it being a perception. I just call
it a mental construct. My questions were rhetorical by the way?

By the why I am an engineer who drifted into Human resources so probably
that's why I have an affinity to PCT.

Regards
Gavin

[From Martin Taylor 2008.03.02.10.33]

(Gavin Ritz 2008.03.02.20.01NZT)

If PCT is a perception theory, then it should be consistent with the
concepts of entropy and energy.

Absolutely. In fact, I think it is the other way round. My
undergraduate background was in Engineering Physics, and I considered
doing graduate work in control systems before becoming an
Experimental Psychologist, which probably is explains something of my
particular approach to PCT.

My take on it is that our current understanding of thermodynamics
forces PCT as the underlying basis for all life, whether that life be
carbon-based or in some other exotic form nobody has yet imagined.
Looked at this way, perceptual control is basically a fundamental
thermodynamic concept, as well as a psychological one.

Why should this be?

In a closed environment, entropy increases by equipartition of energy
among the various degrees of freedom in the system. If the system
starts with some part of it in a structured state, eventually that
structure will dissipate into a higher entropy configuration,
balncing its environment.

There are two ways a structure can survive: be shielded from the
environment, or export entropy to the environment as fast as the
environment exports entropy to the structure. Both imply a separation
between what is and what is not part of the structure. The first
implies placing the structure in its own closed universe, and even
there, if there is any energy in the small closed universe, the
entropy will tend to equipartition, and the structure will dissipate,
if slower than it would have without the shell. The second way is
more robust, and defines PCT.

What PCT means is simply that something within the structure is more
stable (thermodynamically colder) than it would be if the structure
did not act on its environment. That "something" is a perceptual
signal, but its stability is not the important point. The important
point is that by acting on the environment to maintain one or more
perceptual signals more stably than they would otherwise be, the
structure is itself maintained against entropic decay. The structure
can be represented by a set of "intrinsic variables" that are the
elements that are important for survival. The controlled perceptions
are important only insofar as by stabilizing them at suitable
(varying) levels, the system actuallt stabilizes the intrinsic
variables. The system imports energy at low entropy from the
environment ("food", in the general sense) and exports it at higher
entropy ("waste" and heat).

Then is energy and entropy a perception.

They are, in the minds of physicists. That's why I said "our current
understanding" above. But if our perceptions given those labels
relate to the real world in the way they seem to do, then they are
central to the concept of PCT, and PCT is central to all life.

(I'm on the programme level now).

How can this be?

Why not? "Democracy", "lateness for an appointment", "blue", churlish
behaviour", "words on a page" and "mis-spellings" are all
perceptions. What's the problem with "entropy" and "energy"?

Martin

6

[From Bill Powers (2008.03.04.0644 MST)]

Martin Taylor 2008.03.02.10.33 --

My take on it is that our current understanding of thermodynamics forces PCT as the underlying basis for all life, whether that life be carbon-based or in some other exotic form nobody has yet imagined. Looked at this way, perceptual control is basically a fundamental thermodynamic concept, as well as a psychological one.

Our take on the role of thermodynamics is quite different, though what you say about the conceptual relationship of "thermo" to PCT is interesting. As I see it, neither thermodynamics nor PCT "forces" anything in nature (other than our understanding) to be the way it is. They are both mental constructs invented to help explain what we experience. Experience is primary, not theories about experience. The universe, in my opinion (if it cares), does not operate by laws and principles, which exist only in human brains. It operates as it does. We try to explain what we observe of its operations by making up things like entropy and energy, which follow mathematical rules that we also made up and which act as they do (in our imaginations) because of the way we have defined them and because of what those definitions, given the rules we use, imply.

We are good at making up rules and systems of rules that serve as predictive models, but those models are still the results of iteratively guessing and testing our guesses against observations. So we can easily build castles in the air, thinking that because our models predict observations better and better, we are getting close and closer to the truth. We forget that we made up the very entities that we are trying to explain by using the models: distance, space, time, energy, entropy, charge, valence, force, acceleration, and so on down the endless list of scientific concepts. We forget that many things happen inside our models that we never actually observe. We sometimes find that we are explaining things that we no longer believe even exist.

Once in a while nature reminds us about how far off the track we can go while feeling that we are right on the verge of understanding everything. Scientists, such as they were then, believed in phlogiston for 150 years. Elaborate explanations of phenomena were developed, becoming closer and closer to quantitative. Priestley, who discovered oxygen and thereby provided Lavoisier with the ammunition he needed to shoot down phlogiston, believed in phlogiston until he died. Phlogiston had become as real to him as energy and entropy are to a thermodynamicist, or as God is to a devout worshiper. I think it is much the same process that leads to the devout PCTer, the devout thermodynamicist, the devout phlogisticator, and the devout worshiper. It is the process that creates devout belief, which puts an end to certain kinds of enquiry.

Best,

Bill P.

···

--
No virus found in this outgoing message.
Checked by AVG Free Edition. Version: 7.5.516 / Virus Database: 269.21.3/1308 - Release Date: 3/3/2008 10:01 AM

7

[From Martin Taylor 2008.03.04.09.50]

[Bill Powers (2008.03.04.0644 MST)]

Martin Taylor 2008.03.02.10.33 --

My take on it is that our current understanding of thermodynamics forces PCT as the underlying basis for all life, whether that life be carbon-based or in some other exotic form nobody has yet imagined. Looked at this way, perceptual control is basically a fundamental thermodynamic concept, as well as a psychological one.

Our take on the role of thermodynamics is quite different, though what you say about the conceptual relationship of "thermo" to PCT is interesting. As I see it, neither thermodynamics nor PCT "forces" anything in nature (other than our understanding) to be the way it is. They are both mental constructs invented to help explain what we experience.

That is precisely why I said "our current understanding" rather than saying that "thermodynamics forces". You will note that I never said anything about either thermodynamics or PCT forcing anything in nature. As for the rest of your message, it expands on my reasons.

So I don't know that our take on the issue is so different, unless you deny that our current understanding of thermodynamics requires PCT for the existence of life, where life is taken to mean the persistence of certain structures that would be expected decay if they failed to act on their environment.

Martin

8

(Gavin Ritz 2008.03.05.9.14NZT)

[From Bill Powers (2008.03.04.0644 MST)]

Martin Taylor 2008.03.02.10.33 --

They are both mental constructs invented to help explain what we
experience.

You would be one of the few persons that has ever agreed with me on this
issue.

They are by all means very clever mental constructions.

Regards
Gavin

9

[From Richard Kennaway (2008.03.04.2113 GMT)]

[From Martin Taylor 2008.03.02.10.33]
There are two ways a structure can survive: be shielded from the environment, or export entropy to the environment as fast as the environment exports entropy to the structure. Both imply a separation between what is and what is not part of the structure. The first implies placing the structure in its own closed universe, and even there, if there is any energy in the small closed universe, the entropy will tend to equipartition, and the structure will dissipate, if slower than it would have without the shell. The second way is more robust, and defines PCT.

A control system is only one possible way in which that can happen. A heat pump can keep the inside of a box cold without any control systems.

···

--
Richard Kennaway, jrk@cmp.uea.ac.uk, Richard Kennaway
School of Computing Sciences,
University of East Anglia, Norwich NR4 7TJ, U.K.

10

[Martin Taylor 2008.03.04.16.43]

[From Richard Kennaway (2008.03.04.2113 GMT)]

[From Martin Taylor 2008.03.02.10.33]
There are two ways a structure can survive: be shielded from the environment, or export entropy to the environment as fast as the environment exports entropy to the structure. Both imply a separation between what is and what is not part of the structure. The first implies placing the structure in its own closed universe, and even there, if there is any energy in the small closed universe, the entropy will tend to equipartition, and the structure will dissipate, if slower than it would have without the shell. The second way is more robust, and defines PCT.

A control system is only one possible way in which that can happen. A heat pump can keep the inside of a box cold without any control systems.

A control system IS a heat pump. I use the refrigerator as an example when I'm trying to describe what's going on. The interesting thing about the normal refrigerator is that it controls a degree of freedom that is a function of a large number of identifiable environmental degrees of freedom, namely their average momentum.

Martin

11

[From Bill Powers (2008.03.04.1707 MST)]

Martin Taylor 2008.03.04.16.43 --

[Kennbaway]A control system is only one possible way in which that can happen. A heat pump can keep the inside of a box cold without any control systems.

[Taylor]A control system IS a heat pump. I use the refrigerator as an example when I'm trying to describe what's going on. The interesting thing about the normal refrigerator is that it controls a degree of freedom that is a function of a large number of identifiable environmental degrees of freedom, namely their average momentum.

A refrigerator doesn't control anything. There is a control system in most refrigerators that senses temperature and turns the motor on and off to keep temperature at the set point, but refrigerators do not require that control system in order to cool something.

A heat pump that keeps a room warm does so by extracting heat from its surroundings and concentrating it in the heated area. It can be used as part of a control system, of course. Are you saying that this is also a refrigerator, and that what is refrigerated is still the controlled variable? This would be a refrigerator that works by heating, not cooling, so it's hard to see how you get refrigeration out of that, except in terms of some loose metaphor.

Best,

Bill P.

···

--
No virus found in this outgoing message.
Checked by AVG Free Edition. Version: 7.5.516 / Virus Database: 269.21.4/1310 - Release Date: 3/4/2008 8:35 AM

12

[From Martin Taylor 2008.03.04.22.51]

[From Bill Powers (2008.03.04.1707 MST)]

A refrigerator doesn't control anything.

No, It IS a control system which controls (stabilizes) the average momentum of the molecules in the neighbourhood of its temperature sensor.

There is a control system in most refrigerators that senses temperature and turns the motor on and off to keep temperature at the set point, but refrigerators do not require that control system in order to cool something.

That's not the control system in question, is it?

A heat pump that keeps a room warm does so by extracting heat from its surroundings and concentrating it in the heated area.

I think Richard was talking about the heat pump that keeps a room cool, actually. There's no control system involved in heating things.

A normal elementary control unit, the kind used as an element in HPCT, works thermodynamically exactly the same way as a refrigerator. An energy flow from a lower-entropy source to a higher entropy sink is used to do work, which in the case of a refrigerator is to increase the stability of the molecules in the volume cooled. In the case of the "standard" control system, the work is used to increase the stability of the controlled variable.

This would be a refrigerator that works by heating, not cooling, so it's hard to see how you get refrigeration out of that, except in terms of some loose metaphor.

All refrigerators work by heating their environment, and its not a metaphor, tight or loose. It's just a statement of the way things work.

Am I to take it that you are denying that a control system exports entropy from the controlled variable to the environment by using an energy flow from and to the environment?

Martin

13

[From Richard Kennaway (2008.03.05.0843 GMT)]

[Martin Taylor 2008.03.04.16.43]

[From Richard Kennaway (2008.03.04.2113 GMT)]
A control system is only one possible way in which that can happen. A heat pump can keep the inside of a box cold without any control systems.

A control system IS a heat pump.

Not all heat pumps are control systems. As Bill points out, take the thermostat out of the refrigerator and run the heat pump full on, and it will keep the refrigerator cold without a control system. Colder, in fact.

There are plenty of natural processes that result in local decreases of entropy without control systems. Evaporative cooling, for example.

···

--
Richard Kennaway, jrk@cmp.uea.ac.uk, Richard Kennaway
School of Computing Sciences,
University of East Anglia, Norwich NR4 7TJ, U.K.

14

[Martin Taylor 2008.03.05.10.11]

[From Richard Kennaway (2008.03.05.0843 GMT)]

[Martin Taylor 2008.03.04.16.43]

[From Richard Kennaway (2008.03.04.2113 GMT)]
A control system is only one possible way in which that can happen. A heat pump can keep the inside of a box cold without any control systems.

A control system IS a heat pump.

Not all heat pumps are control systems. As Bill points out, take the thermostat out of the refrigerator and run the heat pump full on, and it will keep the refrigerator cold without a control system. Colder, in fact.

There are plenty of natural processes that result in local decreases of entropy without control systems. Evaporative cooling, for example.

Aristotle is a man.

Not all men are Aristotle.

Evaporative cooling is the way most heat pumps work.

I think both you and Bill miss the point. Simply running a heat pump full on doesn't stabilize anything. Without an external energy source, you run out of matter to evaporate, or else you simply reach a condition in entropic equilibrium with the environment. With an external energy source, and an ongoing supply of stuff to evaporate, you just keep on increasing/decreasing the temperature. There's no stabilization there.

Control systems stabilize a variable. Stabilization is cooling. The variation in the stabilized variable is less than it would be without the action of the control system. Control systems export entropy by means of the energy flow from their fuel/food source to their waste products, and are refrigerators.

Martin

15

[From Richard Kennaway (2008.03.05.1754 GMT)]

[Martin Taylor 2008.03.05.10.11]
I think both you and Bill miss the point. Simply running a heat pump full on doesn't stabilize anything.

But it does reduce the temperature of the side it's pumping heat from.

Without an external energy source, you run out of matter to evaporate, or else you simply reach a condition in entropic equilibrium with the environment. With an external energy source, and an ongoing supply of stuff to evaporate, you just keep on increasing/decreasing the temperature. There's no stabilization there.

Leakage back in will eventually balance heat pumped out and an equilibrium will be reached. A ball-in-a-bowl type of equilibrium rather than a controlled one, but still an equilibrium.

Control systems stabilize a variable. Stabilization is cooling. The variation in the stabilized variable is less than it would be without the action of the control system. Control systems export entropy by means of the energy flow from their fuel/food source to their waste products, and are refrigerators.

You appeared to be claiming the two were the same, that not only do control systems export entropy, but anything that exports entropy must be a control system. The latter is simply not true. And if control systems are only a subset of the entropy-exporting mechanisms, I don't see any way of deriving them from fundamental thermodynamics in the sort of way that one can derive the Carnot heat engine cycle.

···

--
Richard Kennaway, jrk@cmp.uea.ac.uk, Richard Kennaway
School of Computing Sciences,
University of East Anglia, Norwich NR4 7TJ, U.K.

16

[Martin Taylor 2008.03.05.13.48]

[From Richard Kennaway (2008.03.05.1754 GMT)]

[Martin Taylor 2008.03.05.10.11]

With an external energy source, and an ongoing supply of stuff to evaporate, you just keep on increasing/decreasing the temperature. There's no stabilization there.

Leakage back in will eventually balance heat pumped out and an equilibrium will be reached. A ball-in-a-bowl type of equilibrium rather than a controlled one, but still an equilibrium.

And one that still involves the use of an ongoing energy flow to maintain it. It's not a ball-in-the-bowl type of equilibrium. It's the kind of equilibrium where someone tries to push a boulder up an ever-steepening slope, and gets to a point where the effort required is all the person can muster. As soon as the person stops pushing, the boulder falls back. It's the same kind of equilibrium that in most practical situations stops a positive feedback loop from going into an explosive runaway.

Control systems stabilize a variable. Stabilization is cooling. The variation in the stabilized variable is less than it would be without the action of the control system. Control systems export entropy by means of the energy flow from their fuel/food source to their waste products, and are refrigerators.

You appeared to be claiming the two were the same, that not only do control systems export entropy, but anything that exports entropy must be a control system. The latter is simply not true.

No. Only ones that stabilize a variable at some level not in equilibrium with the environment and not at some physical limiting condition.

  And if control systems are only a subset of the entropy-exporting mechanisms, I don't see any way of deriving them from fundamental thermodynamics in the sort of way that one can derive the Carnot heat engine cycle.

OK. That's a different issue. I thought I had done the derivation. You don't think I have (in any of my writings on the matter?). So I will have to think about whether you are right, and consider the question further.

There are two issues:

1. Is the top-level stabilizing loop in the classical HPCT structure a "control system" (i.e. must a "control system" have an explicit reference signal input)? Bill P. considers it to be a control system with an implicit reference value for the controlled (stabilized) variable. So do I. If you do not consider such a feedback loop to be a control system then you win the argument, because it is that kind of control system that I think defines life within the context of thermodynamics.

2. Given that a structure is by definition a low-entropy part of a larger system, can a structure be maintained in the absence of a control system to sustain its low-entropy configuration?

I believe I have argued that it cannot. There are only two ways to avoid the import of entropy (i.e. the decay of the structure). One is to shield the structure -- partition it away from the rest of the universe so that it becomes its own closed universe and is at a maximum entropy state for that closed universe; the other is to act as a heat engine and continually export entropy by way of a flow from a low-entropy input energy source to a high-entropy waste flow.

Apart from issue 1, which is an issue of definition, the question is whether a structure can be maintained in its low entropy state without using a feedback loop through its environment that acts as a dynamic shield. The fact that I can see no alternative does, not, I grant, mean that no alternative exists. If you know of one, we must consider it.

Martin

17

[From Bill Powers (2008.03.05.1439 MST)]

Martin Taylor 2008.03.04.22.51 --

I think Richard was talking about the heat pump that keeps a room cool, actually. There's no control system involved in heating things.

But there is a heat pump, so you can't say that a heat pump is automatically a control system or that a control system is always a refrigerator. Whether the heat pump is heating something or cooling it depends only on which side of it you're looking at.

A normal elementary control unit, the kind used as an element in HPCT, works thermodynamically exactly the same way as a refrigerator. An energy flow from a lower-entropy source to a higher entropy sink is used to do work, which in the case of a refrigerator is to increase the stability of the molecules in the volume cooled. In the case of the "standard" control system, the work is used to increase the stability of the controlled variable.

By "increase the stability of molecules," all you mean is slowing down their random motions. That is not control. Control could equally well mean increasing the random motions to a desired level and then acting to prevent either increases or decreases in the motions. But that's not the only problem with your metaphor. There is also the issue of loop gain.

The net random motion results from the sum of heat inputs tending to increase the motion and heat outputs which tend to decrease the motion. The final random motion -- temperature -- is simply the equilibrium point which depends on the temperature differences between heat sources and heat sinks. If an independent agency now inserts or removes heat, the heat sink will not alter its rate of removing heat in such a way as to prevent the motion of the molecules from changing -- that is, to make the motion change less than it would simply from the passive effects of temperature differences. This supposed control system would turn out to have a loop gain of one or less, which makes it into a passive equilibrium system.

As I said above, a heat pump can use used to hold the temperature of a set of molecules higher than the ambient temperature, using an output device that receives energy from an external source (a power plant) and uses some of it to raise the temperature of the molecules. So control can involve heating something as well as cooling it. Moreover, because of amplification, the loop gain of a control system can be in the hundreds or even thousands, while that of a passive thermodynamic system can never be greater than 1. To turn actual heating into metaphorical cooling would be quite a trick, and only a trick.

I think your application of thermodynamics to control theory is based on a narrow set of examples which leaves out cases that negate the general thesis. Being a generalization, it can always be applied when the particular case is known, but it's not possible to work from the general case to the particular and end up with just one case. You can certainly analyze a machine and calculate its properties using energy conservation laws, but if you start with only the energy conservation laws, there is nothing to tell you what kind of machine is demonstrating them. The same happens with the relationship between control systems and thermodyamics. If you know a system is a control system, you can show how the laws of thermodynamics are consistent with the properties of the control system, but you can't go the other way. Given the laws of thermodynamics, there is no way to derive the design of a control system, or any particular kind of system, from them.

William Ross Ashby made the same mistake with his "law of requisite variety." He proved, or thought he had proven, that the "variety" in the actions of a viable control system must always match the "variety" of the environment. His measure of "variety" was something like a measure of information or information flow. But while it is probably true, the law of requisite variety is useless for telling us how to build a control system. It is strictly an after-the-fact description. It is nothing more than a very complicated and fuzzy way of describing what we mean by control: the output varies equally and oppositely to variations in disturbances. But since "variety" is a statistical measure it can't distinguish the direction of a variation, so it can't even say "equal and opposite." All it can say is that the RMS variation in the action is about the same as the RMS variation in the disturbance. That could be true even if the output were not canceling the effect of the disturbance. So Ashby's law missed the critical feature of control.

Best,

Bill P.

···

--
No virus found in this outgoing message.
Checked by AVG Free Edition. Version: 7.5.516 / Virus Database: 269.21.4/1313 - Release Date: 3/5/2008 9:50 AM

18

[From Martin Taylor 2008.03.05.22.48]

Bill Powers (2008.03.05.1439 MST)]

Martin Taylor 2008.03.04.22.51 --

I think Richard was talking about the heat pump that keeps a room cool, actually. There's no control system involved in heating things.

But there is a heat pump, so you can't say that a heat pump is automatically a control system or that a control system is always a refrigerator.

I repeat what I said to Richard: Aristotle is a man, but not all men are Aristotle.

All control systems are refrigerators. Not all heat pumps are control systems.

A normal elementary control unit, the kind used as an element in HPCT, works thermodynamically exactly the same way as a refrigerator. An energy flow from a lower-entropy source to a higher entropy sink is used to do work, which in the case of a refrigerator is to increase the stability of the molecules in the volume cooled. In the case of the "standard" control system, the work is used to increase the stability of the controlled variable.

By "increase the stability of molecules," all you mean is slowing down their random motions. That is not control.

No. It is refrigeration. Actually, it's not the slowing down that's the point. That's only a side effect. The point is that their RMS variation is reduced.

I think your application of thermodynamics to control theory is based on a narrow set of examples which leaves out cases that negate the general thesis.

I haven't seen one yet. An example that would negate the general thesis would be a control system in which the controlled variable has more variation than it would have had if the output of the control system were cut off from the perceptual input. There may be others, but I can't think what they would look like.

If you know a system is a control system, you can show how the laws of thermodynamics are consistent with the properties of the control system, but you can't go the other way. Given the laws of thermodynamics, there is no way to derive the design of a control system, or any particular kind of system, from them.

True. You need one other starting point -- that there exists a structure of some kind that is to be maintained in its low-entropy configuration longer than would be the case if it was left passively to equilibrate with its environment.

Not all thermodynamic systems are alive. Not all control systems are alive. The issue is what, thermodynamically, is required to allow something to be alive.

All control systems, whether they are alive or not, are refrigerators, and I repeat what I said to Richard: If you can think of some way that would allow a low-entropy state to persist in a higher entropy environment other than an isolating shield or a control system, then you will be in a position to assert that thermoduynamic considerations do not require PCT as a basis for life.

Martin

19

[From Bill Powers (2008.03.06.0306 MST)]

Martin Taylor 2008.03.05.22.48 --

All control systems are refrigerators. Not all heat pumps are control systems.

Are you replying to the post in which I reminded you of control systems that act by raising the temperature of air and maintaining it in this elevated state despite either heating or cooling disturbances? Not all control systems are refrigerators. If the variation in RMS molecular velocity is too little, this control system increases it. It both heats and cools, though it relies on an external heat sink for the cooling (a furnace can only heat).

By "increase the stability of molecules," all you mean is slowing down their random motions. That is not control.

No. It is refrigeration. Actually, it's not the slowing down that's the point. That's only a side effect. The point is that their RMS variation is reduced.

But that is precisely what distinguishes thermodynamic equilibrium systems from control systems. In a control system positive fluctuations are resisted by actions in the negative direction; negative fluctuations by actions in the positive direction. There is no one action that will reduce the RMS fluctuations. If all you know about are RMS variations, there is no directional information, so there is no way to create systematic opposition to changes knowing only the magnitude of the RMS variations.

I think it's really stretching a point to see RMS variations as primary and molecular motions as side-effects. RMS variations are a calculation, not a phenomenon. Can you imagine the RMS value of variations being other than zero in the absence of molecular motion, or not increasing if molecular motion increases?

I think your application of thermodynamics to control theory is based on a narrow set of examples which leaves out cases that negate the general thesis.

I haven't seen one yet.

Evidently you hadn't read my example of the controller that raises temperature.

An example that would negate the general thesis would be a control system in which the controlled variable has more variation than it would have had if the output of the control system were cut off from the perceptual input. There may be others, but I can't think what they would look like.

See above. You're reasoning by analogy, not deriving rigorous relationships.

All control systems, whether they are alive or not, are refrigerators, and I repeat what I said to Richard: If you can think of some way that would allow a low-entropy state to persist in a higher entropy environment other than an isolating shield or a control system, then you will be in a position to assert that thermoduynamic considerations do not require PCT as a basis for life.

Not all control systems create and maintain low-entropy states. For example, a dancer deliberately leaps, twirls, lifts, and runs in controlled patterns, creating motion where there was none before. All control of dynamic variables involves generation of variation, not reduction in variation. Walking is a higher-entropy state than standing still; standing up is a higher-entropy state than remaining lying down. What you see as higher and lower entropy depends on what you think the direction of disorganization is. If you see walking as more organized than standing still, you will say that walking decreases entropy and corresponds to cooling; if you see it as you see the motion of molecules, you'll see the increased activity as heating.

The only way you can say that control always involves decreasing entropy is to ignore the physical nature of the controlled variable and look only at whether it is varying more or less because of the control. Only then could you say that maintaining a room at a comfortable temperature in the winter is refrigeration: by using a thermostat you have decreased the fluctuations in temperature, so you have "refrigerated" those fluctuations -- while increasing the fluctuations in molecular positions. But if you ignore the physical nature of the controlled variable, you have converted a systematic analysis into a metaphor, and have lost the rigorous connection to the world of physics. If you decide to use the thermostat to run the temperature up and down more than it would fluctuate naturally, you have to change the basis of the metaphor again to make this seem to be refrigeration -- now you have to define "less" fluctuation as "varying more according to the pattern of change that I want to see."

You have looked at control only as reducing variations due to disturbances, which is a narrow view of control. The actions of a control system can just as easily increase the variations in the physicaL world if that is required by the nature of the perception under control. I can agree that control does not violate the laws of thermodynamics, because nothing can or does violate them. But it is thermodynamics that is explained by real system, not the other way around.

Best,

Bill P.

···

--
No virus found in this outgoing message.
Checked by AVG Free Edition. Version: 7.5.516 / Virus Database: 269.21.4/1313 - Release Date: 3/5/2008 9:50 AM

20

[Martin Taylor 2008.03.06.09.25]

[From Bill Powers (2008.03.06.0306 MST)]

Martin Taylor 2008.03.05.22.48 --

All control systems are refrigerators. Not all heat pumps are control systems.

Are you replying to the post in which I reminded you of control systems that act by raising the temperature of air and maintaining it in this elevated state despite either heating or cooling disturbances? Not all control systems are refrigerators.

I can't tell whether you are ribbing me, are doing the Lord Nelson thing of putting a telescope deliberately to the blind eye, are trying to pull the wool over the eyes of the lurkers, or truly don't understand. Charity leads me to want to believe the last, but it's a position I find ever harder to sustain in the face of increasing levels of disturbance.

Martin