What is Qi?

[Abbott]

[From Bruce Abbott (2017.02.12.1325 EST)]

Rick Marken (2017.02.10.1845) –

Rick’s comments on various statements I made in my previous post [Bruce Abbott (2017.02.10.1515 EST)] made little sense to me until I finally realized that the main source of our differences relates to how one defines the input quantity, Qi. I had viewed Qi as an environmental variable that stimulates a sensory receptor. The receptor acts as a transducer whose output is a perceptual signal, p, corresponding to Qi. I gave as an example an automotive cruise control, for which Qi is the actual speed of the car and p is a voltage that varies in proportion to the car’s speed.

Below is Figure A.1. from B:CP:

Qi, the input quantity, is depicted as the small circle receiving inputs from the feedback and disturbance functions and connecting via the single arrow to the input function. This diagram is consistent with my description of Qi. In other places, Bill Powers also refers to Qi as the “controlled variable,” the objective variable in the environment that is represented internally to the system as p.

The situation gets a bit more complicated when several environmental variables enter the input function and are combined in some way to produce p. For example, a person’s perception of air temperature might depend both on the actual temperature of the air and the humidity. In this case, what is Qi?

To my mind there would be two Qis, one for air temperature and one for humidity. Separate arrows from each Qi would enter the input function, which would output a perceptual signal p whose value depended in some way (as specified by the input function) on both air temperature and humidity. This representation explicitly shows how p arises from the effects of two environmental variables on the input function. However, one can no longer equate Qi with the “controlled variable” – the observable equivalent of the controlled perception.

Rick, on the other hand, asserted that Qi should continue to represent the controlled variable and pointed to another of Bill Power’s system diagrams, this one from Bill’s (1973) Science paper (Figure 1):

In this diagram, Qi is represented by the large circle containing the individual variables of which Qi is a function. Each individual variable separately enters the input function (labeled “sensor function” here). In this version, there is only one Qi, and this Qi is just a label for the controlled variable.

Before I understood that Rick was defining Qi in this way, I could not see why he kept referring to the observer’s view of the control system. Referring to my cruise control example, he stated:

I thought initially that Rick had descended into madness – he seems to be saying that to work, cruise control requires an observer’s perception of what it is controlling. That, of course, is utter nonsense. After puzzling about this for quite some time it finally dawned on me that Rick wasn’t talking about how cruise control actually works. He was talking about what Qi (a.k.a. the controlled variable) represents in second diagram above, where Qi is defined as the combination of environmental variables that corresponds to p. It is what an observer of the system would infer is the environmental equivalent of p, based perhaps on the test for the controlled variable.

This way of defining and diagramming Qi allows one to talk about an environmental equivalent of the controlled perception, but in so doing it introduces a new set of problems. It seems to suggest that the Qi is a single environmental variable composed of its constituent environmental variables. In fact the environment may present only the individual constituents, which are then combined within the system’s input function to yield p. The taste of lemonade depends on the values of several input variables (sweetness, sourness, etc.) but there is no actual taste of lemonade out there in the environment. (What does exist there are combinations of ingredients that stimulate the right sensory receptors in the right intensities to produce in the taster the taste of lemonade (as that person defines it).

In other cases a single environmental variable result from the influences of two or more environmental variables. How warm the air seems to be depends in part on the temperature of the skin that is exposed to the air. That temperature depends on the air temperature, humidity (which influences the rate of evaporation of sweat, and thus the rate of evaporative cooling, and the wind speed. One might vary these factors and conclude that Qi depends in some complex way on all three. But this does not necessarily mean that each factor is being independently sensed and combined within the input function to create the internal perception p of skin temperature. All three factors directly affect the actual temperature of the skin, and that environmental variable may be the only one actually being sensed and on which p depends.

An additional problem arises from identifying Qi as an observer’s perception of what constitutes the environmental equivalent of the controlled variable. One might think from this that the operation of the control system being observed depends on what the observer perceives Qi to be, as opposed to environmental variables enter the input function and how they combined by the input function to create the perceptual signal. Of course, control systems function perfectly well without the aid of observers.

Returning to Rick’s statement quoted earlier:

RM: What you call the actual speed is the observer’s (in this case probably an engineer’s) perception of the speed, probably derived from instruments (which are also perceptions, of course). It’s the speed (perception) that the observer-engineer wants the cruise controller to control. The speed perception that the cruise controller is actually controlling, as perceived by the observer-engineer, is Qi.

No, what I called the actual speed is the car’s actual speed. It’s the car’s actual speed that affects the speed sensor’s output, not the engineer’s perception of the car’s speed. Qi is an environmental variable (or a composite of several), not a perception. There is a reason why Bill’s second diagram above shows the arrows that go to the input function coming from the individual v’s inside the circle representing Qi, and not from Qi itself. The control system only senses the v’s, not Qi (unless Qi is a single v).

Bruce

[Abbott]

[From Bruce Abbott (2017.02.12.1325 EST)]

Rick Marken (2017.02.10.1845) –

Rick’s comments on various statements I made in my previous post [Bruce Abbott (2017.02.10.1515 EST)] made little sense to me until I finally realized that the main source of our differences relates to how one defines the input quantity, Qi. I had viewed Qi as an environmental variable that stimulates a sensory receptor. The receptor acts as a transducer whose output is a perceptual signal, p, corresponding to Qi. I gave as an example an automotive cruise control, for which Qi is the actual speed of the car and p is a voltage that varies in proportion to the car’s speed.

Below is Figure A.1. from B:CP:

Qi, the input quantity, is depicted as the small circle receiving inputs from the feedback and disturbance functions and connecting via the single arrow to the input function. This diagram is consistent with my description of Qi. In other places, Bill Powers also refers to Qi as the “controlled variable,” the objective variable in the environment that is represented internally to the system as p.

The situation gets a bit more complicated when several environmental variables enter the input function and are combined in some way to produce p. For example, a person’s perception of air temperature might depend both on the actual temperature of the air and the humidity. In this case, what is Qi?

To my mind there would be two Qis, one for air temperature and one for humidity. Separate arrows from each Qi would enter the input function, which would output a perceptual signal p whose value depended in some way (as specified by the input function) on both air temperature and humidity. This representation explicitly shows how p arises from the effects of two environmental variables on the input function. However, one can no longer equate Qi with the “controlled variable” – the observable equivalent of the controlled perception.

Rick, on the other hand, asserted that Qi should continue to represent the controlled variable and pointed to another of Bill Power’s system diagrams, this one from Bill’s (1973) Science paper (Figure 1):

In this diagram, Qi is represented by the large circle containing the individual variables of which Qi is a function. Each individual variable separately enters the input function (labeled “sensor function” here). In this version, there is only one Qi, and this Qi is just a label for the controlled variable.

Before I understood that Rick was defining Qi in this way, I could not see why he kept referring to the observer’s view of the control system. Referring to my cruise control example, he stated:

RM: What you call the actual speed is the observer’s (in this case probably an engineer’s) perception of the speed, probably derived from instruments (which are also perceptions, of course). It’s the speed (perception) that the observer-engineer wants the cruise controller to control. The speed perception that the cruise controller is actually controlling, as perceived by the observer-engineer, is Qi.

I thought initially that Rick had descended into madness – he seems to be saying that to work, cruise control requires an observer’s perception of what it is controlling. That, of course, is utter nonsense. After puzzling about this for quite some time it finally dawned on me that Rick wasn’t talking about how cruise control actually works. He was talking about what Qi (a.k.a. the controlled variable) represents in second diagram above, where Qi is defined as the combination of environmental variables that corresponds to p. It is what an observer of the system would infer is the environmental equivalent of p, based perhaps on the test for the controlled variable.

This way of defining and diagramming Qi allows one to talk about an environmental equivalent of the controlled perception, but in so doing it introduces a new set of problems. It seems to suggest that the Qi is a single environmental variable composed of its constituent environmental variables. In fact the environment may present only the individual constituents, which are then combined within the system’s input function to yield p. The taste of lemonade depends on the values of several input variables (sweetness, sourness, etc.) but there is no actual taste of lemonade out there in the environment. (What does exist there are combinations of ingredients that stimulate the right sensory receptors in the right intensities to produce in the taster the taste of lemonade (as that person defines it).

In other cases a single environmental variable result from the influences of two or more environmental variables. How warm the air seems to be depends in part on the temperature of the skin that is exposed to the air. That temperature depends on the air temperature, humidity (which influences the rate of evaporation of sweat, and thus the rate of evaporative cooling, and the wind speed. One might vary these factors and conclude that Qi depends in some complex way on all three. But this does not necessarily mean that each factor is being independently sensed and combined within the input function to create the internal perception p of skin temperature. All three factors directly affect the actual temperature of the skin, and that environmental variable may be the only one actually being sensed and on which p depends.

An additional problem arises from identifying Qi as an observer’s perception of what constitutes the environmental equivalent of the controlled variable. One might think from this that the operation of the control system being observed depends on what the observer perceives Qi to be, as opposed to environmental variables enter the input function and how they combined by the input function to create the perceptual signal. Of course, control systems function perfectly well without the aid of observers.

Returning to Rick’s statement quoted earlier:

RM: What you call the actual speed is the observer’s (in this case probably an engineer’s) perception of the speed, probably derived from instruments (which are also perceptions, of course). It’s the speed (perception) that the observer-engineer wants the cruise controller to control. The speed perception that the cruise controller is actually controlling, as perceived by the observer-engineer, is Qi.

No, what I called the actual speed is the car’s actual speed. It’s the car’s actual speed that affects the speed sensor’s output, not the engineer’s perception of the car’s speed. Qi is an environmental variable (or a composite of several), not a perception. There is a reason why Bill’s second diagram above shows the arrows that go to the input function coming from the individual v’s inside the circle representing Qi, and not from Qi itself. The control system only senses the v’s, not Qi (unless Qi is a single v).

Bruce

[Huddy_Vyv]

[From Bruce Abbott (2017.02.12.1325 EST)]

Rick Marken (2017.02.10.1845) –

Rick’s comments on various statements I made in my previous post [Bruce Abbott (2017.02.10.1515 EST)] made little sense to me until I finally realized that the main source of our differences relates to how one defines the input quantity,
Qi. I had viewed Qi as an environmental variable that stimulates a sensory receptor. The receptor acts as a transducer whose output is a perceptual signal, p, corresponding to Qi. I gave as an example an automotive cruise control, for which Qi is the actual
speed of the car and p is a voltage that varies in proportion to the car’s speed.

Below is Figure A.1. from B:CP:

<image001.jpg>
Qi, the input quantity, is depicted as the small circle receiving inputs from the feedback and disturbance functions and connecting via the single arrow to the input function. This diagram is consistent with my description of Qi. In other places, Bill Powers
also refers to Qi as the “controlled variable,� the objective variable in the environment that is represented internally to the system as p.

The situation gets a bit more complicated when several environmental variables enter the input function and are combined in some way to produce p. For example, a person’s perception of air temperature might depend both on the actual temperature
of the air and the humidity. In this case, what is Qi?

To my mind there would be two Qis, one for air temperature and one for humidity. Separate arrows from each Qi would enter the input function, which would output a perceptual signal p whose value depended in some way (as specified by the
input function) on both air temperature and humidity. This representation explicitly shows how p arises from the effects of two environmental variables on the input function. However, one can no longer equate Qi with the “controlled variableâ€? – the observvable
equivalent of the controlled perception.

Rick, on the other hand, asserted that Qi should continue to represent the controlled variable and pointed to another of Bill Power’s system diagrams, this one from Bill’s (1973)
Science paper (Figure 1):

<image003.jpg>
In this diagram, Qi is represented by the large circle containing the individual variables of which Qi is a function. Each individual variable separately enters the input function (labeled “sensor function� here). In this version, there is only one Qi, and
this Qi is just a label for the controlled variable.

Before I understood that Rick was defining Qi in this way, I could not see why he kept referring to the observer’s view of the control system. Referring to my cruise control example, he stated:

RM: What you call the actual speed is the observer’s (in this case probably an engineer’s) perception of the speed, probably derived from instruments (which are also perceptions, of course). It’s the speed (perception) that the observer-engineer
wants the cruise controller to control. The speed perception that the cruise controller is actually controlling, as perceived by the observer-engineer, is Qi.

I thought initially that Rick had descended into madness – he seemms to be saying that to work, cruise control requires an observer’s perception of what it is controlling. That, of course, is utter nonsense. After puzzling about this for
quite some time it finally dawned on me that Rick wasn’t talking about how cruise control actually works. He was talking about what Qi (a.k.a. the controlled variable) represents in second diagram above, where Qi is defined as the combination of environmental
variables that corresponds to p. It is what an observer of the system would infer is the environmental equivalent of p, based perhaps on the test for the controlled variable.

This way of defining and diagramming Qi allows one to talk about an environmental equivalent of the controlled perception, but in so doing it introduces a new set of problems. It seems to suggest that the Qi is a single environmental
variable composed of its constituent environmental variables. In fact the environment may present only the individual constituents, which are then combined within the system’s input function to yield p. The taste of lemonade depends on the values of several
input variables (sweetness, sourness, etc.) but there is no actual taste of lemonade out there in the environment. (What does exist there are combinations of ingredients that stimulate the right sensory receptors in the right intensities to produce in the
taster the taste of lemonade (as that person defines it).

In other cases a single environmental variable result from the influences of two or more environmental variables. How warm the air seems to be depends in part on the temperature of the skin that is exposed to the air. That temperature
depends on the air temperature, humidity (which influences the rate of evaporation of sweat, and thus the rate of evaporative cooling, and the wind speed. One might vary these factors and conclude that Qi depends in some complex way on all three. But this
does not necessarily mean that each factor is being independently sensed and combined within the input function to create the internal perception p of skin temperature. All three factors directly affect the actual temperature of the skin, and that environmental
variable may be the only one actually being sensed and on which p depends.

An additional problem arises from identifying Qi as an observer’s perception of what constitutes the environmental equivalent of the controlled variable. One might think from this that the operation of the control system being observed
depends on what the observer perceives Qi to be, as opposed to environmental variables enter the input function and how they combined by the input function to create the perceptual signal. Of course, control systems function perfectly well without the aid
of observers.

Returning to Rick’s statement quoted earlier:

RM: What you call the actual speed is the observer’s (in this case probably an engineer’s) perception of the speed, probably derived from instruments (which are also perceptions, of course). It’s the speed (perception) that the observer-engineer
wants the cruise controller to control. The speed perception that the cruise controller is actually controlling, as perceived by the observer-engineer, is Qi.

No, what I called the actual speed is the car’s actual speed. It’s the car’s actual speed that affects the speed sensor’s output, not the engineer’s perception of the car’s speed. Qi is an environmental variable (or a composite of several),
not a perception. There is a reason why Bill’s second diagram above shows the arrows that go to the input function coming from the individual v’s inside the circle representing Qi, and not from Qi itself. The control system only senses the v’s, not Qi (unless
Qi is a single v).

Bruce

[Abbott]

[From Bruce Abbott (2017.02.12.1820 EST)]

[Vyv Huddy 1955.12.02.2017]

VH: This is great set of posts on this thread and the previous one. Particularly helpful to see the posts by rick and bruce describing the figure in the Powers Science paper showing multiple V. I noticed that and missed why it was important.

VH: I don’t have my books with me now but i recall Bill Powers (i think) wrote somewhere that the controlled variable of a thermostat is actually the amount of coil of the strip not the temperature of the room itself. This is partly shown by putting a flame directly under the thermostat - the furnace goes off without the room temperature changing.

BA: The controlled variable is the temperature of the air. The position of the coil (actually, the contact on the free end of the coil) is the thermostat’s perception of that temperature.  The position of the other contact sets the thermostat’s reference level. When the contacts meet, this generates a nonzero error signal indicating that the room temperature has fallen below the reference level. This causes the furnace to switch on (output function), heating the air and thus bringing the room temperature up.

BA: The thermostat is a sensor of room temperature only to the extent that the coil temperature matches the room temperature.  All sensors work this way, sensing a variable through its effect on the sensor.  In this example changing temperatures of the air are conveyed to the coil, producing different expansions on the two sides of the bimetallic strip, thus moving the contact on the free end. In the case of vision, photons striking molecules of photochemical in the photoreceptors cause light-sensitive chemicals within the photoreceptors to break down, which through a series of chemical events causes the “generator potential� of the photoreceptor to change. This in turn alters the rate of firing of associated neurons.

VH: If so the controlled variable (qi) of a cruise control would be the frequency of the counting wheel turns in the speedometer mechanism? If so from that perspective there is no “actual” speed; cruise controls don’t work effectively on a slippery surfaces and drivers are encouraged not to use them then.

BA:  One could design cruise controls to sense the car’s speed in a variety of ways. The typical example is that speed is sensed indirectly by counting the rotational frequency of the car’s drive shaft. Under normal conditions this is proportional to the speed of the car. This relationship breaks down if the wheels slip, as you note. This problem could be avoided by using an optical sensor to read the “optic flow� of the road beneath the car.  However measured, there is still an actual speed that is being measured, whether directly or indirectly, so long as the expected relationships hold. In the absence of wheel slippage, driveshaft rotations map directly onto the car’s speed.  Biological systems experience the same difficulties; sensor reports do not always match reality. Our sensors remain useful, however, because most of the time they provide readings that are good enough to go by.

VH: As i write this though I’m beginning to doubt the examples of thermostats or any other man made control system as clear illustrations of control in living systems. This is because machines input functions come about in a totally different way to those of living things. They can be made to be much simpler; a cruise control can sense velocity via a single variable because of the way it is organised (a turning crank and some frequency counter). There is no way a living system can sense velocity with a single environmental variable (without some sort of bio engineering).

BA: Why not? Optic flow will do it.

VH: I wonder if this difference makes these examples hard for me understand. I find the two domains don’t map onto each other that well.

BA: I don’t see any essential difference between human-engineered control systems and biological ones; in fact it is that correspondence that allowed Bill Powers (and others) to apply control-system principles to biological systems. What is true is that biological systems are products of evolution, a process that involves gradual modification of what is already there. Unlike a human designer, evolution does not begin with a fresh sheet of paper. What results is more like a hodge-podge of alterations and fixes than a clean design, and this can make it extremely difficult to tease out what the parts of a given system are and how they interconnect. Bill Powers notes, for example, that alpha motor neurons in the spinal cord combine the functions of comparator and output function (and even that is probably an oversimplification). So it may be more difficult to figure out the system diagrams in biological systems than in human-engineered systems, but that does not imply that the underlying system principles of the two are necessarily different.  PCT is based on the well-supported assumption that behavior can be explained by control-system models.

VH: If cruise control p is really a single voltage in wire then it would be only intensity controller? For me Intensity control is best explained in the context of the muscle tone example, as in b:cp.

BA: In HPCT, perceptual signals at all levels are equivalent to single voltages in a wire (i.e., average neural current in a nerve). Â Thus, the fact that p in cruise control is embodied as a single voltage in a wire does not make the cruise control an intensity controller.

In fact, all controllers in HPCT, at whatever level, are intensity controllers in the sense that they act to control the intensity of their neural signal p’s.

Bruce

[Huddy_Vyv]

[Vyv Huddy 2130.13.02.2017]

[From Bruce Abbott (2017.02.12.1820 EST)]

BA: The controlled variable is the temperature of the air. The position of the coil (actually, the contact on the free end of the coil) is the thermostat’s perception of that temperature.

VH: Ok great. This is consistent with what Bill Powers says in LCS II p. 134-136, which was the passage I remembered, so I agree that p is the position of the strip.

[Vyv Huddy 1955.12.02.2017]

VH: There is no way a living system can sense velocity with a single environmental variable (without some sort of bio engineering).

BA: Why not? Optic flow will do it.

VH: It might but your comment prompted me to look up stuff on artificial devices that sense optic flow. They seem to record changes in a function of TWO [x, y] co-ordinates of scenes over time. Optic flow is therefore an abstract aspect of the environment. My point was the cruise control is built so it can control frequency of wheel turning (it's p) which is a single variable. I don't think that's possible without the speedometer mechanism. That mechanism is a deliberately designed with an input function that living things don't have. A machine or living thing sensing optic flow would require a higher perceptual level requiring a function of (at least) two variables to create a perception of the rate of flow.

[Vyv Huddy 1955.12.02.2017]

VH: If cruise control p is really a single voltage in wire then it would be only intensity controller? For me Intensity control is best explained in the context of the muscle tone example, as in b:cp.

BA: In HPCT, perceptual signals at all levels are equivalent to single voltages in a wire (i.e., average neural current in a nerve).

VH: Yes I get that. But input of a higher level CS is a function of multitude of parallel sensory receptors. A higher level CS wouldn't ever have one sensory cell (or wire) as an input? Would it?

BA: Thus, the fact that p in cruise control is embodied as a single voltage in a wire does not make the cruise control an intensity controller.

VH: It is no more complex than a single input, like a sensing muscle tension, so is equivalent to an intensity controller. For my understanding. I'm not saying it is an "intensity controller" because that has a specific meaning in PCT, which is the lowest level perception.

BA: In fact, all controllers in HPCT, at whatever level, are intensity controllers in the sense that they act to control the intensity of their neural signal p’s.

VH: Intensity control is a level in PCT, the lowest level, so I prefer the word magnitude for the size of p at all levels.

VH: Thanks Bruce ... this is really helpful stuff.

Bruce

[EetuP]

[eetu pikkarainen 2017-02-14]

Hi Vyv

VH: As i write this though I’m beginning to doubt the examples of thermostats or any other man made control system as clear illustrations of control in living systems. This
is because machines input functions come about in a totally different way to those of living things. They can be made to be much simpler; a cruise control can sense velocity via a single variable because of the way it is organised (a turning crank and some
frequency counter). There is no way a living system can sense velocity with a single environmental variable (without some sort of bio engineering). I wonder if this difference makes these examples hard for me understand. I find the two domains don’t map onto
each other that well.

I feel that doubt, too, (but on the other hand I still enjoy the surprisingly close relation).

The most remarkable difference for me is of course that reference value is determined by an outside force (user, engineer) for the man-made control systems while for LCS the reference in “natural�,
partly innate but anyway suited for its living needs, and they can be further adjusted by systems themselves.

As for the perceptual input, Bruce already answered that a living system could sense velocity with a single environmental variable. But I do not think it is the essential question but rather that
LCS can seldom or never concentrate to perceive only one “thing� like velocity, but because of our needs of living we are always sensing a very large spectrum of variables which are then many ways mixed together and with memories in higher level perceptions.
Those higher level wholes than determine back to the reference levels and probably to input functions too.

(Perhaps Chad was referring to something like this in his last message about controlling wholeness.)

···

Eetu

[chadtgreen]

[Chad Green (2017.02.14.0936 EST)]

EP: (Perhaps Chad was referring to something like this in his last message about controlling wholeness.)

CG: Wholeness to my mind is a reflection of unconscious logic at work, e.g., see:
https://en.wikipedia.org/wiki/Ignacio_Matte_Blanco . Why not leverage it?

Best,

Chad

···

From: Eetu Pikkarainen [mailto:eetu.pikkarainen@oulu.fi]
Sent: Tuesday, February 14, 2017 6:56 AM
To: csgnet@lists.illinois.edu
Subject: RE: What is Qi?

[eetu pikkarainen 2017-02-14]

Hi Vyv

VH:
As i write this though I’m beginning to doubt the examples of thermostats or any other man made control system as clear illustrations of control in living systems. This is because machines input functions
come about in a totally different way to those of living things. They can be made to be much simpler; a cruise control can sense velocity via a single variable because of the way it is organised (a turning crank and some frequency counter). There is no way
a living system can sense velocity with a single environmental variable (without some sort of bio engineering). I wonder if this difference makes these examples hard for me understand. I find the two domains don’t map onto each other that well.

I feel that doubt, too, (but on the other hand I still enjoy the surprisingly close relation).

The most remarkable difference for me is of course that reference value is determined by an outside force (user, engineer) for the man-made control systems while for LCS the reference in “naturalâ€?, partly innate
but anyway suited for its living needs, and they can be further adjusted by systems themselves.

As for the perceptual input, Bruce already answered that a living system could sense velocity with a single environmental variable. But I do not think it is the essential question but rather that LCS can seldom
or never concentrate to perceive only one “thingâ€? like velocity, but because of our needs of living we are always sensing a very large spectrum of variables which are then many ways mixed together and with memories in higher level perceptions. Those higher
level wholes than determine back to the reference levels and probably to input functions too.

(Perhaps Chad was referring to something like this in his last message about controlling wholeness.)

Eetu

[MartinT]

[Martin Taylor 2017.02.13.17.57]

[Vyv Huddy 1955.12.02.2017]

    This is great set of posts on this thread and the previous

one. Particularly helpful to see the posts by rick and bruce
describing the figure in the Powers Science paper showing
multiple V. I noticed that and missed why it was important.

It isn't important. It's just a consequence of the fact that if you

have a perception p1, it’s evolutionarily pointless to have a
perceptual function that just takes only p1 as input and creates p2
as a function of p1. Why pointless? because if you control p1 you
would also be controlling p2, and vice-versa. Every perception you
have depends on several variables, without exception, unless you
count as perceptions the myriads of individual sensors (retinal rods
and cones, auditory hair cells, and so forth). Even they are
influenced by what their neighbours are and have been doing. So when
you talk about a perception, you are always talking about “multiple
V” in the environment. I learned about levels of perception that
were built one on the other from a Children’s Encyclopedia when I
was about 10, so for me this idea that every perception is built
from a lot of others is just a given.

The important point is about degrees of freedom. In the Powers

version of Perceptual Control Theory, each perception has only one
degree of freedom, its value. That creates a bottleneck in the loop,
which means that no matter how many variables there are at different
parts of the loop, there’s only one degree of freedom anywhere in
the loop that is related to the perception. That degree of freedom
may be distributed over many “wires”, but no matter how many wires,
there’s only one degree of freedom. Specifically, for the purpose of
this discussion, the CEV has only one degree of freedom. That’s the
important point to keep in focus.

If you are unfamiliar with the "degrees of freedom" concept, the

basic idea is very simple, though the nooks and crannies of it an
get rather arcane. Basically, a construct has as many degree of
freedom as it has variables that can be independently changed. A
point on a plane can be moved in x and y, but once you have
specified them, you can’t move the point without changing one or
other of them. If you now try to describe the location as radius and
angle theta (polar coordinates), you find you can’t vary either of
them without changing one or both of x and y. You could describe the
location of the point as x and radius, (which would be ambiguous)
but when you specified those, then you would have fixed theta and y
(apart from sign). A triangle has three degrees of freedom if you
ignore its location. You can specify the lengths of its three sides,
but then you can’t independently specify any of its angle. Or you
can specify two sides and an angle, or one side and two angles.
Three angles won’t do in this case, because two of them determine
the value of the third, and if you try to specify only angles, you
have only two degrees of freedom to work with. You need three, and
the third could be, say, the distance from the triangle centroid to
the nearest side.

In the case of the CEV and the perception, you can specify one of

them, but then you can’t independently specify the other without
changing something else. They have one degree of freedom between
them, but you can locate that degree of freedom anywhere around the
loop, such as in the values of the component variables that
contribute to the value of the CEV.

You talk about the diagrams of the loop in which the CEV is shown as

a circle containing components, and multiple lines lead from the
action output to the components in the CEV circle. Those diagrams
are all valid, if sometimes a little misleading, but equally valid
is the type of diagram in which there’s just one line from output to
a little circle where the output meets the disturbance to create the
sensor input. The “one-line” diagram correctly shows the single
degree of freedom around the loop, but fails to show any of the
complexity of the processes. It’s your choice which kind of diagram
you prefer, to illustrate a particular point you want to make.

The HPCT diagram usually shows layers of control units inside the

organism with lots of cross links from systems at one level to
systems at the other. But you can do the same sort of hierarchic
visualisation with the “dots in a circle CEV” view.

![NestedCEVs.jpg|1529x720](upload://7ywxZtNLMxGRolFdznP7cJrGaIY.jpeg)

In this diagram, the little inset at the top left shows one complete

control loop with an incoming disturbance signal. The rest shows
only the environmental parts of the loops at different levels of the
hierarchy. At the left we have the “dots in a circle” view at a
level I call “Level N”. The heavy circle and incoming line from
below show respectively the single degree of freedom CEV and
disturbance.

Inside the circle are shown component variables of the CEV, each

being affected by separate branches of the output and of the
disturbance. If you ignore the value of the CEV, these three have
one degree of freedom each, making three in all. But if you vary one
of them, you have to also vary one or both of the others if you are
not going to change the value of the CEV. Among the four values (CEV
and its three components) there is only one degree of freedom,
because the CEV is a function of its components.

In the middle part, the diagram shows the same thing one level

higher in the hierarchy, except that in this case the individual
internal components (the level N CEVs) each are explicitly shown to
be of the type illustrated at the left. Again, each component at
that level has its individual “wire” from the output and from the
disturbance, so that if level N had been the base level, the action
output would now be influencing 9 independent
components, and there would be nine affected sensory inputs. At the
next level there would be 27, and so forth, assuming always three
component CEVs for each next-level CEV. Likewise going to lower
levels, several perceptions at level N-1 become the components at
level N. But each CEV, considered by itself, has exactly one degree
of freedom, no matter how many components and subcomponents it may
have.

(Since not all perceptions are controlled, the components could

simply be the environmental variables corresponding to uncontrolled
perceptions, in which case one would omit its “action wires” in the
diagram.)

We ask about the "reality" of the CEV represented by the heavy

circle. Since at every level, the heavy circle represents a function
of the little circles, the “reality” question has to be given the
same answer at every level. Furthermore, since even at the lowest
perceptual level, there are always multiple sensors (e.g. retinal
rods, auditory hair cells, touch receptors. etc) involved, the same
“reality” question applies before you even get to the lowest level
of perceptual control. I think logically, one has to say either that
there are NO environmental variables corresponding to ANY perceptual
signal, or that ALL the CEVs corresponding to ALL the perceptions
that are based purely on sensory inputs are equally real. And each
CEV has exactly one degree of freedom if the corresponding
perception does, no matter how many components it can be proved to
depend on.

If all the CEVs at a level are equally real -- either all real or

all unreal – what should we say about “virtual reality” and about
the apparently real scenes in movies? Firstly, you cannot control
anything you see in a movie. All the components in the diagram are
devoid of “action wires”. If that had been true all your life, you
probably would never have reorganized to produce perceptual
functions that produced those perceptions. But it hasn’t been true
for you or for your ancestors. You have most of those perceptions
because they or something like them have been usefully controlled
for long enough to establish the corresponding perceptual functions.
The movie uses the perceptual functions you have developed through
active control. Their usefulness in selecting their specific
functions out of the environment in which you have lived has helped
you or your ancestors to “live long and prosper”. As one might say,
imagination may caress, but reality bites.

Now there's always the question as to whether the CEV is actually in

the environment. It’s not a question one can ever answer assuredly
affirmatively. As with a scientific hypothesis, it can be disproven,
but not proven. So what about “Virtual Reality”? In VR you can
control quite a few perceptions, and the more perceptions you can
control, the more real (on average) the experience. But you know
consciously that you are not in a world in which virtual shopping
gets real food that will allow you to control your perception of
satiety-hunger. I would be surprised if that difference were not
also part of the perceptions that you control in VR – at least at
high enough levels.

So far in this thread, we have been assuming that the perceptual

input comes only by way of the sensory apparatus, which means the
CEV is indeed in the environment. But it’s quite possible for some
of the input to a perceptual function to come from imagination. In
that case, the environmental portion of the input is no longer
constrained to be a single degree of freedom, because changes in one
or more of the environmental variables can be compensated by changes
in the part that comes from imagination.

No Test for the Controlled Variable could find a CEV that is partly

imagination. The CEV corresponding to the perception is only partly
in the environment, and the best a TCV could do if the imagination
part changes is to find control to be poor at best. Perception might
be controlled very well, but the corresponding perception in the
external environment might not be. So we have to ask for each
perception the degree to which its value is influenced by
imagination. This is a tough problem both for an outside observer
and for an observer in the same body as the perception under
examination. So we look elsewhere, elsewhere being any other
observer who can access the CEV only through the environment.
Introspection won’t work to tell us whether something is in the
environment or is an illusion or mirage.

If you control for X to be at a reference level R, and vary R

appreciably, and someone else observing your environment says that
they perceive something changing that they would also identify as X,
that’s evidence that X or something very like it is in the
environment. If the other disturbs what they see as X and you have
to vary your output to bring your perception of it to R, that’s more
evidence. If lots of people equally can perceive what they think of
as X and can also seem to influence it so that you have to vary your
output to bring it back near R, the weight of evidence increases.
But it’s never proof. All these other people might be subject to the
same illusion as you. But if you and they can control other
perceptions using X as though it was real, that’s better evidence.

Proof that something is not in the environment can often be

achieved. The example of the Ames Room has been brought up. The Ames
room looks from one specific viewpoint like a normal rectangular
room, in which people and objects change size as they move around
the room. Is the room really in the environment? Possibly. Is there
a real rectangular room in the place where you perceive one to
exist? Not according to people who view it from different places.
They see that the room exists, but is not rectangular. The
rectangularity is not in the environment, though the room may be. A
lake you perceive in the distance may not have any water if it is a
mirage. These properties are subject to test. You can’t control your
perception of your thirst level my drinking from an illusory lake.

…

    As i write this though I'm beginning to doubt the examples of

thermostats or any other man made control system as clear
illustrations of control in living systems. This is because
machines input functions come about in a totally different way
to those of living things. They can be made to be much simpler;
a cruise control can sense velocity via a single variable
because of the way it is organised (a turning crank and some
frequency counter). There is no way a living system can sense
velocity with a single environmental variable (without some sort
of bio engineering).

I'm not clear why you say this. Velocity and direction sensing is

the job of “complex cells” in the primary visual cortex, which is
presumably before any possibility of being incorporated in a control
loop. At least the little neurophysiology that I read suggests that
their inputs are strictly bottom-up. I wouldn’t be at all surprised
if directional velocity rather than intensity were eventually found
to be at level 0 of the visual part of the control hierarchy. Bill
often used velocity below position in his models, so he wasn’t at
all dedicated to the hierarchy as we usually list it. He was usually
careful to point out that those eleven levels came from his own
introspection, and were individually unsupported by experiment. As
for intensity being at the base, relatively few (if any) sensor
cells give outputs that are functions of current intensity. Most
preferentially report changes both over time and with respect to
their neighbours, with some late resting level that might have a
relationship to intensity.

Visual velocity sensors function as I would imagine an engineered

velocity sensor would do if it had to rely on visual input alone.
Indeed, to me this seems to be true of a lot of systems. Evolution
has found a lot of solutions for problems addressed by engineers,
and there’s a lot of feedback between physiologists and engineers.
An engineer has a problem that biology seems to have solved; how
does the biological system do it? Maybe we could try that. Or, the
biological system seems to be doing something funny (such as the
frequency sweep of a hunting bat’s squeak); Why does it do that? Oh,
if we do that we can make our sonars more informative. The sensor
systems may be physically different, but there’s often a close
functional correspondence.

    I wonder if this difference makes these examples hard for me

understand. I find the two domains don’t map onto each other
that well.

    If cruise control p is really a single voltage in wire then

it would be only intensity controller?

Why? Every perceptual signal at every level in Powers's PCT is

carried on a single wire, isn’t it?

Sorry again for the length of this. All I really want to get across

is (1) that every perception and every corresponding environmental
variable (the new expansion of “CEV” agreed by Kent and me) has only
one degree of freedom, at least in the Powers version of PCT, which
means it is effectively “carried on a single wire”, and (2) that in
the absence of specific evidence in respect of a particular
perception and its CEV, all CEVs at every level have exactly the
same likelihood of being really in the environment. Either all of
them may be, or none of them can be. Philosophically I don’t think
there is a third possibility.

Martin

[wmansell]

Hi Martin,

This is a very helpful thread. One important element I take from it among many is the use of the imagination connection on a sliding scale, presumably at multiple levels, rather than as an all or none ‘mode’ of the entire system as I get stuck into describing it.

All the best,

Warren

···

On 15 Feb 2017, at 04:13, Martin Taylor mmt-csg@mmtaylor.net wrote:

[Martin Taylor 2017.02.13.17.57]

[Vyv Huddy 1955.12.02.2017]

    This is great set of posts on this thread and the previous

one. Particularly helpful to see the posts by rick and bruce
describing the figure in the Powers Science paper showing
multiple V. I noticed that and missed why it was important.

It isn't important. It's just a consequence of the fact that if you

have a perception p1, it’s evolutionarily pointless to have a
perceptual function that just takes only p1 as input and creates p2
as a function of p1. Why pointless? because if you control p1 you
would also be controlling p2, and vice-versa. Every perception you
have depends on several variables, without exception, unless you
count as perceptions the myriads of individual sensors (retinal rods
and cones, auditory hair cells, and so forth). Even they are
influenced by what their neighbours are and have been doing. So when
you talk about a perception, you are always talking about “multiple
V” in the environment. I learned about levels of perception that
were built one on the other from a Children’s Encyclopedia when I
was about 10, so for me this idea that every perception is built
from a lot of others is just a given.

The important point is about degrees of freedom. In the Powers

version of Perceptual Control Theory, each perception has only one
degree of freedom, its value. That creates a bottleneck in the loop,
which means that no matter how many variables there are at different
parts of the loop, there’s only one degree of freedom anywhere in
the loop that is related to the perception. That degree of freedom
may be distributed over many “wires”, but no matter how many wires,
there’s only one degree of freedom. Specifically, for the purpose of
this discussion, the CEV has only one degree of freedom. That’s the
important point to keep in focus.

If you are unfamiliar with the "degrees of freedom" concept, the

basic idea is very simple, though the nooks and crannies of it an
get rather arcane. Basically, a construct has as many degree of
freedom as it has variables that can be independently changed. A
point on a plane can be moved in x and y, but once you have
specified them, you can’t move the point without changing one or
other of them. If you now try to describe the location as radius and
angle theta (polar coordinates), you find you can’t vary either of
them without changing one or both of x and y. You could describe the
location of the point as x and radius, (which would be ambiguous)
but when you specified those, then you would have fixed theta and y
(apart from sign). A triangle has three degrees of freedom if you
ignore its location. You can specify the lengths of its three sides,
but then you can’t independently specify any of its angle. Or you
can specify two sides and an angle, or one side and two angles.
Three angles won’t do in this case, because two of them determine
the value of the third, and if you try to specify only angles, you
have only two degrees of freedom to work with. You need three, and
the third could be, say, the distance from the triangle centroid to
the nearest side.

In the case of the CEV and the perception, you can specify one of

them, but then you can’t independently specify the other without
changing something else. They have one degree of freedom between
them, but you can locate that degree of freedom anywhere around the
loop, such as in the values of the component variables that
contribute to the value of the CEV.

You talk about the diagrams of the loop in which the CEV is shown as

a circle containing components, and multiple lines lead from the
action output to the components in the CEV circle. Those diagrams
are all valid, if sometimes a little misleading, but equally valid
is the type of diagram in which there’s just one line from output to
a little circle where the output meets the disturbance to create the
sensor input. The “one-line” diagram correctly shows the single
degree of freedom around the loop, but fails to show any of the
complexity of the processes. It’s your choice which kind of diagram
you prefer, to illustrate a particular point you want to make.

The HPCT diagram usually shows layers of control units inside the

organism with lots of cross links from systems at one level to
systems at the other. But you can do the same sort of hierarchic
visualisation with the “dots in a circle CEV” view.

<NestedCEVs.jpg>



In this diagram, the little inset at the top left shows one complete

control loop with an incoming disturbance signal. The rest shows
only the environmental parts of the loops at different levels of the
hierarchy. At the left we have the “dots in a circle” view at a
level I call “Level N”. The heavy circle and incoming line from
below show respectively the single degree of freedom CEV and
disturbance.

Inside the circle are shown component variables of the CEV, each

being affected by separate branches of the output and of the
disturbance. If you ignore the value of the CEV, these three have
one degree of freedom each, making three in all. But if you vary one
of them, you have to also vary one or both of the others if you are
not going to change the value of the CEV. Among the four values (CEV
and its three components) there is only one degree of freedom,
because the CEV is a function of its components.

In the middle part, the diagram shows the same thing one level

higher in the hierarchy, except that in this case the individual
internal components (the level N CEVs) each are explicitly shown to
be of the type illustrated at the left. Again, each component at
that level has its individual “wire” from the output and from the
disturbance, so that if level N had been the base level, the action
output would now be influencing 9 independent
components, and there would be nine affected sensory inputs. At the
next level there would be 27, and so forth, assuming always three
component CEVs for each next-level CEV. Likewise going to lower
levels, several perceptions at level N-1 become the components at
level N. But each CEV, considered by itself, has exactly one degree
of freedom, no matter how many components and subcomponents it may
have.

(Since not all perceptions are controlled, the components could

simply be the environmental variables corresponding to uncontrolled
perceptions, in which case one would omit its “action wires” in the
diagram.)

We ask about the "reality" of the CEV represented by the heavy

circle. Since at every level, the heavy circle represents a function
of the little circles, the “reality” question has to be given the
same answer at every level. Furthermore, since even at the lowest
perceptual level, there are always multiple sensors (e.g. retinal
rods, auditory hair cells, touch receptors. etc) involved, the same
“reality” question applies before you even get to the lowest level
of perceptual control. I think logically, one has to say either that
there are NO environmental variables corresponding to ANY perceptual
signal, or that ALL the CEVs corresponding to ALL the perceptions
that are based purely on sensory inputs are equally real. And each
CEV has exactly one degree of freedom if the corresponding
perception does, no matter how many components it can be proved to
depend on.

If all the CEVs at a level are equally real -- either all real or

all unreal – what should we say about “virtual reality” and about
the apparently real scenes in movies? Firstly, you cannot control
anything you see in a movie. All the components in the diagram are
devoid of “action wires”. If that had been true all your life, you
probably would never have reorganized to produce perceptual
functions that produced those perceptions. But it hasn’t been true
for you or for your ancestors. You have most of those perceptions
because they or something like them have been usefully controlled
for long enough to establish the corresponding perceptual functions.
The movie uses the perceptual functions you have developed through
active control. Their usefulness in selecting their specific
functions out of the environment in which you have lived has helped
you or your ancestors to “live long and prosper”. As one might say,
imagination may caress, but reality bites.

Now there's always the question as to whether the CEV is actually in

the environment. It’s not a question one can ever answer assuredly
affirmatively. As with a scientific hypothesis, it can be disproven,
but not proven. So what about “Virtual Reality”? In VR you can
control quite a few perceptions, and the more perceptions you can
control, the more real (on average) the experience. But you know
consciously that you are not in a world in which virtual shopping
gets real food that will allow you to control your perception of
satiety-hunger. I would be surprised if that difference were not
also part of the perceptions that you control in VR – at least at
high enough levels.

So far in this thread, we have been assuming that the perceptual

input comes only by way of the sensory apparatus, which means the
CEV is indeed in the environment. But it’s quite possible for some
of the input to a perceptual function to come from imagination. In
that case, the environmental portion of the input is no longer
constrained to be a single degree of freedom, because changes in one
or more of the environmental variables can be compensated by changes
in the part that comes from imagination.

No Test for the Controlled Variable could find a CEV that is partly

imagination. The CEV corresponding to the perception is only partly
in the environment, and the best a TCV could do if the imagination
part changes is to find control to be poor at best. Perception might
be controlled very well, but the corresponding perception in the
external environment might not be. So we have to ask for each
perception the degree to which its value is influenced by
imagination. This is a tough problem both for an outside observer
and for an observer in the same body as the perception under
examination. So we look elsewhere, elsewhere being any other
observer who can access the CEV only through the environment.
Introspection won’t work to tell us whether something is in the
environment or is an illusion or mirage.

If you control for X to be at a reference level R, and vary R

appreciably, and someone else observing your environment says that
they perceive something changing that they would also identify as X,
that’s evidence that X or something very like it is in the
environment. If the other disturbs what they see as X and you have
to vary your output to bring your perception of it to R, that’s more
evidence. If lots of people equally can perceive what they think of
as X and can also seem to influence it so that you have to vary your
output to bring it back near R, the weight of evidence increases.
But it’s never proof. All these other people might be subject to the
same illusion as you. But if you and they can control other
perceptions using X as though it was real, that’s better evidence.

Proof that something is not in the environment can often be

achieved. The example of the Ames Room has been brought up. The Ames
room looks from one specific viewpoint like a normal rectangular
room, in which people and objects change size as they move around
the room. Is the room really in the environment? Possibly. Is there
a real rectangular room in the place where you perceive one to
exist? Not according to people who view it from different places.
They see that the room exists, but is not rectangular. The
rectangularity is not in the environment, though the room may be. A
lake you perceive in the distance may not have any water if it is a
mirage. These properties are subject to test. You can’t control your
perception of your thirst level my drinking from an illusory lake.

…

    As i write this though I'm beginning to doubt the examples of

thermostats or any other man made control system as clear
illustrations of control in living systems. This is because
machines input functions come about in a totally different way
to those of living things. They can be made to be much simpler;
a cruise control can sense velocity via a single variable
because of the way it is organised (a turning crank and some
frequency counter). There is no way a living system can sense
velocity with a single environmental variable (without some sort
of bio engineering).

I'm not clear why you say this. Velocity and direction sensing is

the job of “complex cells” in the primary visual cortex, which is
presumably before any possibility of being incorporated in a control
loop. At least the little neurophysiology that I read suggests that
their inputs are strictly bottom-up. I wouldn’t be at all surprised
if directional velocity rather than intensity were eventually found
to be at level 0 of the visual part of the control hierarchy. Bill
often used velocity below position in his models, so he wasn’t at
all dedicated to the hierarchy as we usually list it. He was usually
careful to point out that those eleven levels came from his own
introspection, and were individually unsupported by experiment. As
for intensity being at the base, relatively few (if any) sensor
cells give outputs that are functions of current intensity. Most
preferentially report changes both over time and with respect to
their neighbours, with some late resting level that might have a
relationship to intensity.

Visual velocity sensors function as I would imagine an engineered

velocity sensor would do if it had to rely on visual input alone.
Indeed, to me this seems to be true of a lot of systems. Evolution
has found a lot of solutions for problems addressed by engineers,
and there’s a lot of feedback between physiologists and engineers.
An engineer has a problem that biology seems to have solved; how
does the biological system do it? Maybe we could try that. Or, the
biological system seems to be doing something funny (such as the
frequency sweep of a hunting bat’s squeak); Why does it do that? Oh,
if we do that we can make our sonars more informative. The sensor
systems may be physically different, but there’s often a close
functional correspondence.

    I wonder if this difference makes these examples hard for me

understand. I find the two domains don’t map onto each other
that well.

    If cruise control p is really a single voltage in wire then

it would be only intensity controller?

Why? Every perceptual signal at every level in Powers's PCT is

carried on a single wire, isn’t it?

Sorry again for the length of this. All I really want to get across

is (1) that every perception and every corresponding environmental
variable (the new expansion of “CEV” agreed by Kent and me) has only
one degree of freedom, at least in the Powers version of PCT, which
means it is effectively “carried on a single wire”, and (2) that in
the absence of specific evidence in respect of a particular
perception and its CEV, all CEVs at every level have exactly the
same likelihood of being really in the environment. Either all of
them may be, or none of them can be. Philosophically I don’t think
there is a third possibility.

Martin

[rsmarken]

[From  Rick Marken (2017.02.15.1000)]

···

Martin Taylor (2017.02.13.17.57)-

[Vyv Huddy 1955.12.02.2017]

    This is great set of posts on this thread and the previous

one. Particularly helpful to see the posts by rick and bruce
describing the figure in the Powers Science paper showing
multiple V. I noticed that and missed why it was important.

MT: It isn't important. It's just a consequence of the fact that if you

have a perception p1, it’s evolutionarily pointless to have a
perceptual function that just takes only p1 as input and creates p2
as a function of p1.

RM: The symbol p refers to the perceptual signal that is the output of a perceptual function. A perceptual signal can also be the input to a higher level perceptual function, but this has nothing to do with the fact that multiple v’s are shown to be components of the input quantity, q.i, in the diagram in Bill’s Science paper. The multiple v’s are multiple physical variables in the environment, such as the amounts of sugar (v.1), acid (v.2) and oil (v.3) in water (v.4). The circle around the v’s indicates that the controlled input quantity, q.i, is a function of these environmental variables: q.i = f(v.1, v.2,v3, v.4). The arrows from the individual v’s into the “sensor function” (the perceptual function) indicates that the sensor signal (also called the perceptual signal, p) is the same function of the v’s (indeed, the sensor signal is shown in the diagram to be f(v.1, v2…v.n)). So the diagram shows that the input quantity (or controlled quantity), q.i, is equivalent to the perceptual (sensor) signal, p; q.i is p as seen from the observer’s perspective. In more familiar, experiential terms, q.i and p represent variations in the taste of the lemonade mixture from the point of view of the observer (me) and the controller (you), respectively.

MT: The important point is about degrees of freedom. In the Powers

version of Perceptual Control Theory, each perception has only one
degree of freedom, its value…Â

Â

MT: In the case of the CEV and the perception, you can specify one of

them, but then you can’t independently specify the other without
changing something else.

RM: If the CEV is equivalent to q.i, then this is true since q.i is the same function of environmental variables as p. But the truth of this has nothing to do the df.Â

MT: You talk about the diagrams of the loop in which the CEV is shown as

a circle containing components, and multiple lines lead from the
action output to the components in the CEV circle.

RM: I’ve never seen such a diagram. It’s certainly not what is seen in the Science diagram where there is only one output line connecting the effector function to the input quantity via the feedback function. In general there can be multiple inputs to a perceptual function but only a single output from the output function in any control loop. Â

MT: But each CEV, considered by itself, has exactly one degree

of freedom, no matter how many components and subcomponents it may
have.

RM: Yes, if the CEV is equivalent to q.i, then q.i is the scalar (1 df) output of the same function as the one that produces the scalar variable p.Â

Â

MT: We ask about the "reality" of the CEV represented by the heavy

circle. Since at every level, the heavy circle represents a function
of the little circles, the “reality” question has to be given the
same answer at every level.

Â

RM: Yes, q.i is as real as p, because they are the same function of environmental variables.Â

MT: Now there's always the question as to whether the CEV is actually in

the environment. It’s not a question one can ever answer assuredly
affirmatively.

RM: If the CEV is equivalent to q.i, then it is assuredly not in the environment (at least, according to PCT); it is a perception in the observer. The physical variables that are the basis of the perception of q.i and p are in the environment.Â

Â

MT: So far in this thread, we have been assuming that the perceptual

input comes only by way of the sensory apparatus, which means the
CEV is indeed in the environment.

RM: Then the CEV is a concept that is not equivalent to q.i.Â

Â

MT: No Test for the Controlled Variable could find a CEV that is partly

imagination.

RM: First, the TCV is aimed at determining the controlled input quantity, q.i, which is a perception. It is not aimed at determining a CEV, which is, at least partly conceived of as an entity in the environment. But it is an interesting question whether the TCV can determine a q.i that is partly imagination. I think it would be possible. The tester would  just have to be able to notice that the controlling involves an imaginary component and he or she would have to try to guess what that component was and how it was varying. For example, in the coin game you might notice that the controller seems to be imagining an extra coin in a position that completes the pattern under control. This could be tested by creating disturbances to the hypothetical controlled variable (call it q.i’)  that can only be corrected by adding what you guess to be the imagined coin. This seems like a nice research project there for those interested in studying the role of imagination in control.

MT: If you control for X to be at a reference level R, and vary R

appreciably, and someone else observing your environment says that
they perceive something changing that they would also identify as X,
that’s evidence that X or something very like it is in the
environment.

Â

RM: It’s consistent with the assumption that there are physical variables “out there” that are the basis of the perception of X. The fact that you and I perceive a particular combination of sugar, acid, oils and water as “lemonade” (X = the taste of lemonade) is not evidence that there is an entity in the environment that is the taste of lemonade; according to PCT, it is evidence that you and I have the same perceptual function that produces a perceptual signal that we experience as the taste of lemonade.Â

MT: Â Velocity and direction sensing is

the job of “complex cells” in the primary visual cortex, which is
presumably before any possibility of being incorporated in a control
loop.Â

RM: Actually, velocity and direction sensing is presumed to be carried out by arrays of receptors called “receptive fields”, which are the perceptual functions in a control loop. The neurons that carry the outputs of the receptive field computations are called “simple”, “complex” or “hypercomplex” depending on the complexity of the variable computed by their associated receptive field function. It’s not the cells themselves that are “simple” or “complex”; it’s the computations carried out by the receptive field “neural network” that determines whether a cell (actually the axon that carries the output of the neural net computation) is “simple” or “complex”.Â

Best

Rick

–
Richard S. MarkenÂ

"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.�
                --Antoine de Saint-Exupery

[Huddy_Vyv]

There are a lot of fascinating ideas and explanations in your post Martin. It will be some time before I can take it all in properly!

···

[Martin Taylor 2017.02.13.17.57]

[Vyv Huddy 1955.12.02.2017]

This is great set of posts on this thread and the previous one. Particularly helpful to see the posts by rick and bruce describing the figure in the Powers Science paper showing multiple V. I noticed that and missed why it was important.

It isn’t important. It’s just a consequence of the fact that if you have a perception p1, it’s evolutionarily pointless to have a perceptual function that just takes only p1 as input and creates p2 as a function of p1. Why pointless? because if you control
p1 you would also be controlling p2, and vice-versa. Every perception you have depends on several variables, without exception, unless you count as perceptions the myriads of individual sensors (retinal rods and cones, auditory hair cells, and so forth). Even
they are influenced by what their neighbours are and have been doing. So when you talk about a perception, you are always talking about “multiple V” in the environment. I learned about levels of perception that were built one on the other from a Children’s
Encyclopedia when I was about 10, so for me this idea that every perception is built from a lot of others is just a given.

The important point is about degrees of freedom. In the Powers version of Perceptual Control Theory, each perception has only one degree of freedom, its value. That creates a bottleneck in the loop, which means that no matter how many variables there are at
different parts of the loop, there’s only one degree of freedom anywhere in the loop that is related to the perception. That degree of freedom may be distributed over many “wires”, but no matter how many wires, there’s only one degree of freedom. Specifically,
for the purpose of this discussion, the CEV has only one degree of freedom. That’s the important point to keep in focus.

If you are unfamiliar with the “degrees of freedom” concept, the basic idea is very simple, though the nooks and crannies of it an get rather arcane. Basically, a construct has as many degree of freedom as it has variables that can be independently changed.
A point on a plane can be moved in x and y, but once you have specified them, you can’t move the point without changing one or other of them. If you now try to describe the location as radius and angle theta (polar coordinates), you find you can’t vary either
of them without changing one or both of x and y. You could describe the location of the point as x and radius, (which would be ambiguous) but when you specified those, then you would have fixed theta and y (apart from sign). A triangle has three degrees of
freedom if you ignore its location. You can specify the lengths of its three sides, but then you can’t independently specify any of its angle. Or you can specify two sides and an angle, or one side and two angles. Three angles won’t do in this case, because
two of them determine the value of the third, and if you try to specify only angles, you have only two degrees of freedom to work with. You need three, and the third could be, say, the distance from the triangle centroid to the nearest side.

In the case of the CEV and the perception, you can specify one of them, but then you can’t independently specify the other without changing something else. They have one degree of freedom between them, but you can locate that degree of freedom anywhere around
the loop, such as in the values of the component variables that contribute to the value of the CEV.

You talk about the diagrams of the loop in which the CEV is shown as a circle containing components, and multiple lines lead from the action output to the components in the CEV circle. Those diagrams are all valid, if sometimes a little misleading, but equally
valid is the type of diagram in which there’s just one line from output to a little circle where the output meets the disturbance to create the sensor input. The “one-line” diagram correctly shows the single degree of freedom around the loop, but fails to
show any of the complexity of the processes. It’s your choice which kind of diagram you prefer, to illustrate a particular point you want to make.

The HPCT diagram usually shows layers of control units inside the organism with lots of cross links from systems at one level to systems at the other. But you can do the same sort of hierarchic visualisation with the “dots in a circle CEV” view.

In this diagram, the little inset at the top left shows one complete control loop with an incoming disturbance signal. The rest shows only the environmental parts of the loops at different levels of the hierarchy. At the left we have the “dots in a circle”
view at a level I call “Level N”. The heavy circle and incoming line from below show respectively the single degree of freedom CEV and disturbance.

Inside the circle are shown component variables of the CEV, each being affected by separate branches of the output and of the disturbance. If you ignore the value of the CEV, these three have one degree of freedom each, making three in all. But if you vary
one of them, you have to also vary one or both of the others if you are not going to change the value of the CEV. Among the four values (CEV and its three components) there is only one degree of freedom, because the CEV is a function of its components.

In the middle part, the diagram shows the same thing one level higher in the hierarchy, except that in this case the individual internal components (the level N CEVs) each are explicitly shown to be of the type illustrated at the left. Again, each component
at that level has its individual “wire” from the output and from the disturbance, so that if level N had been the base level, the action output would now be influencing 9 independent components, and there would be nine affected sensory inputs. At the next
level there would be 27, and so forth, assuming always three component CEVs for each next-level CEV. Likewise going to lower levels, several perceptions at level N-1 become the components at level N. But each CEV, considered by itself, has exactly one degree
of freedom, no matter how many components and subcomponents it may have.

(Since not all perceptions are controlled, the components could simply be the environmental variables corresponding to uncontrolled perceptions, in which case one would omit its “action wires” in the diagram.)

We ask about the “reality” of the CEV represented by the heavy circle. Since at every level, the heavy circle represents a function of the little circles, the “reality” question has to be given the same answer at every level. Furthermore, since even at the
lowest perceptual level, there are always multiple sensors (e.g. retinal rods, auditory hair cells, touch receptors. etc) involved, the same “reality” question applies before you even get to the lowest level of perceptual control. I think logically, one has
to say either that there are NO environmental variables corresponding to ANY perceptual signal, or that ALL the CEVs corresponding to ALL the perceptions that are based purely on sensory inputs are equally real. And each CEV has exactly one degree of freedom
if the corresponding perception does, no matter how many components it can be proved to depend on.

If all the CEVs at a level are equally real – either all real or all unreal – what should we say about “virtual reality” and about the apparently real scenes in movies? Firstly, you cannot control anything you see in a movie. All the components in the diagram
are devoid of “action wires”. If that had been true all your life, you probably would never have reorganized to produce perceptual functions that produced those perceptions. But it hasn’t been true for you or for your ancestors. You have most of those perceptions
because they or something like them have been usefully controlled for long enough to establish the corresponding perceptual functions. The movie uses the perceptual functions you have developed through active control. Their usefulness in selecting their specific
functions out of the environment in which you have lived has helped you or your ancestors to “live long and prosper”. As one might say, imagination may caress, but reality bites.

Now there’s always the question as to whether the CEV is actually in the environment. It’s not a question one can ever answer assuredly affirmatively. As with a scientific hypothesis, it can be disproven, but not proven. So what about “Virtual Reality”? In
VR you can control quite a few perceptions, and the more perceptions you can control, the more real (on average) the experience. But you know consciously that you are not in a world in which virtual shopping gets real food that will allow you to control your
perception of satiety-hunger. I would be surprised if that difference were not also part of the perceptions that you control in VR – at least at high enough levels.

So far in this thread, we have been assuming that the perceptual input comes only by way of the sensory apparatus, which means the CEV is indeed in the environment. But it’s quite possible for some of the input to a perceptual function to come from imagination.
In that case, the environmental portion of the input is no longer constrained to be a single degree of freedom, because changes in one or more of the environmental variables can be compensated by changes in the part that comes from imagination.

No Test for the Controlled Variable could find a CEV that is partly imagination. The CEV corresponding to the perception is only partly in the environment, and the best a TCV could do if the imagination part changes is to find control to be poor at best. Perception
might be controlled very well, but the corresponding perception in the external environment might not be. So we have to ask for each perception the degree to which its value is influenced by imagination. This is a tough problem both for an outside observer
and for an observer in the same body as the perception under examination. So we look elsewhere, elsewhere being any other observer who can access the CEV only through the environment. Introspection won’t work to tell us whether something is in the environment
or is an illusion or mirage.

If you control for X to be at a reference level R, and vary R appreciably, and someone else observing your environment says that they perceive something changing that they would also identify as X, that’s evidence that X or something very like it is in the
environment. If the other disturbs what they see as X and you have to vary your output to bring your perception of it to R, that’s more evidence. If lots of people equally can perceive what they think of as X and can also seem to influence it so that you have
to vary your output to bring it back near R, the weight of evidence increases. But it’s never proof. All these other people might be subject to the same illusion as you. But if you and they can control other perceptions using X as though it was real, that’s
better evidence.

Proof that something is not in the environment can often be achieved. The example of the Ames Room has been brought up. The Ames room looks from one specific viewpoint like a normal rectangular room, in which people and objects change size as they move around
the room. Is the room really in the environment? Possibly. Is there a real rectangular room in the place where you perceive one to exist? Not according to people who view it from different places. They see that the room exists, but is not rectangular. The
rectangularity is not in the environment, though the room may be. A lake you perceive in the distance may not have any water if it is a mirage. These properties are subject to test. You can’t control your perception of your thirst level my drinking from an
illusory lake.

…

As i write this though I’m beginning to doubt the examples of thermostats or any other man made control system as clear illustrations of control in living systems. This is because machines input functions come about in a totally different way to those
of living things. They can be made to be much simpler; a cruise control can sense velocity via a single variable because of the way it is organised (a turning crank and some frequency counter). There is no way a living system can sense velocity with a single
environmental variable (without some sort of bio engineering).

I’m not clear why you say this. Velocity and direction sensing is the job of “complex cells” in the primary visual cortex, which is presumably before any possibility of being incorporated in a control loop. At least the little neurophysiology that I read suggests
that their inputs are strictly bottom-up. I wouldn’t be at all surprised if directional velocity rather than intensity were eventually found to be at level 0 of the visual part of the control hierarchy. Bill often used velocity below position in his models,
so he wasn’t at all dedicated to the hierarchy as we usually list it. He was usually careful to point out that those eleven levels came from his own introspection, and were individually unsupported by experiment. As for intensity being at the base, relatively
few (if any) sensor cells give outputs that are functions of current intensity. Most preferentially report changes both over time and with respect to their neighbours, with some late resting level that might have a relationship to intensity.

Visual velocity sensors function as I would imagine an engineered velocity sensor would do if it had to rely on visual input alone. Indeed, to me this seems to be true of a lot of systems. Evolution has found a lot of solutions for problems addressed by engineers,
and there’s a lot of feedback between physiologists and engineers. An engineer has a problem that biology seems to have solved; how does the biological system do it? Maybe we could try that. Or, the biological system seems to be doing something funny (such
as the frequency sweep of a hunting bat’s squeak); Why does it do that? Oh, if we do that we can make our sonars more informative. The sensor systems may be physically different, but there’s often a close functional correspondence.

I wonder if this difference makes these examples hard for me understand. I find the two domains don’t map onto each other that well.

If cruise control p is really a single voltage in wire then it would be only intensity controller?

Why? Every perceptual signal at every level in Powers’s PCT is carried on a single wire, isn’t it?

Sorry again for the length of this. All I really want to get across is (1) that every perception and every corresponding environmental variable (the new expansion of “CEV” agreed by Kent and me) has only one degree of freedom, at least in the Powers version
of PCT, which means it is effectively “carried on a single wire”, and (2) that in the absence of specific evidence in respect of a particular perception and its CEV, all CEVs at every level have exactly the same likelihood of being really in the environment.
Either all of them may be, or none of them can be. Philosophically I don’t think there is a third possibility.

Martin

[_Boris_Hartman]

Down…

···

From: Richard Marken [mailto:rsmarken@gmail.com]
Sent: Wednesday, February 15, 2017 7:04 PM
To: csgnet@lists.illinois.edu
Subject: Re: What is Qi?

[From Rick Marken (2017.02.15.1000)]

Martin Taylor (2017.02.13.17.57)-

[Vyv Huddy 1955.12.02.2017]

This is great set of posts on this thread and the previous one. Particularly helpful to see the posts by rick and bruce describing the figure in the Powers Science paper showing multiple V. I noticed that and missed why it was important.

MT: It isn’t important. It’s just a consequence of the fact that if you have a perception p1, it’s evolutionarily pointless to have a perceptual function that just takes only p1 as input and creates p2 as a function of p1.

RM: The symbol p refers to the perceptual signal that is the output of a perceptual function. A perceptual signal can also be the input to a higher level perceptual function, but this has nothing to do with the fact that multiple v’s are shown to be components of the input quantity, q.i, in the diagram in Bill’s Science paper. The multiple v’s are multiple physical variables in the environment, such as the amounts of sugar (v.1), acid (v.2) and oil (v.3) in water (v.4). The circle around the v’s indicates that the controlled input quantity, q.i,

HB : Where did you find this term »controlled input quantity« q.i. ???

RM : …is a function of these environmental variables: q.i = f(v.1, v…2,v3, v.4). The arrows from the individual v’s into the “sensor function” (the perceptual function) indicates that the sensor signal (also called the perceptual signal, p)

HB : It’s sure not »Controlled input quantity«. Rick you are again inventing PCT. Or should I say you are improving your RCT.

HB : It can’t be a »controlled input quantity« because if it is then you should have also »Controlled perceptucal signal« ??? And there is no »Controlled Perceptual signal« in PCT… Did I missed something ?

Boris

is the same function of the v’s (indeed, the sensor signal is shown in the diagram to be f(v.1, v2…v.n)). So the diagram shows that the input quantity (or controlled quantity), q.i, is equivalent to the perceptual (sensor) signal, p;

HB : Well here you are making a mistake. You should read once again carefully what does it mean f(v1,.v2…). So read Biills’ books and find out what really means f(v1, v2,…) in the structture of perceptual signal. You don’t understand what’s the structure of perceptual signal and how its controled in comparator.

It’s not that easy. Q.I. or in Bills’ language (controlled quantity) is not the same as »perceptual signal«…

I found an interesting note from Martin. I agree with him.

MT :

RM …q.i is p as seen from thhe observer’s perspective.

HB : What an ignorance. Mostly you can’t perceive waht is q.i. in the mind of controlled. …. If you are watching the car twisting on the road, what is q.i. from the drivers aspect and tne observers aspect ??? How can you from the observation point (for example hill) conclude what is p in driver ???

And I told you to experiment with saying hello to people on the road. Did you do your homework. What is q.i. from controllers and what is q.i from obervers perspective ??? How can he conlclude what the structure of perceptual signal is ?

RM : In more familiar, experiential terms, q.i and p represent variations in the taste of the lemonade mixture from the point of view of the observer (me) and the controller (you), respectively.

HB : Think again what are you drinking and how it is perceived ???

MT: The important point is about degrees of freedom. In the Powers version of Perceptual Control Theory, each perception has only one degree of freedom, its value…

MT: In the case of the CEV and the perception, you can specify one of them, but then you can’t independently specify the other without changing something else.

RM: If the CEV is equivalent to q.i, then this is true since q.i is the same function of environmental variables as p.

HB : Once again q.i. is not the same as p.

But the truth of this has nothing to do the df.

MT: You talk about the diagrams of the loop in which the CEV is shown as a circle containing components, and multiple lines lead from the action output to the components in the CEV circle.

RM: I’ve never seen such a diagram. It’s certainly not what is seen in the Science diagram where there is only one output line connecting the effector function to the input quantity via the feedback function. In general there can be multiple inputs to a perceptual function but only a single output from the output function in any control loop.

MT: But each CEV, considered by itself, has exactly one degree of freedom, no matter how many components and subcomponents it may have.

RM: Yes, if the CEV is equivalent to q.i, then q.i is the scalar (1 df) output of the same function as the one that produces the scalar variable p.

MT: We ask about the “reality” of the CEV represented by the heavy circle. Since at every level, the heavy circle represents a function of the little circles, the “reality” question has to be given the same answer at every level.

RM: Yes, q.i is as real as p, because they are the same function of environmental variables.

MT: Now there’s always the question as to whether the CEV is actually in the environment. It’s not a question one can ever answer assuredly affirmatively.

RM: If the CEV is equivalent to q.i, then it is assuredly not in the environment (at least, according to PCT); it is a perception in the observer. The physical variables that are the basis of the perception of q.i and p are in the environment.

MT: So far in this thread, we have been assuming that the perceptual input comes only by way of the sensory apparatus, which means the CEV is indeed in the environment.

RM: Then the CEV is a concept that is not equivalent to q.i.

MT: No Test for the Controlled Variable could find a CEV that is partly imagination.

RM: First, the TCV is aimed at determining the controlled input quantity, q.i, which is a perception. It is not aimed at determining a CEV, which is, at least partly conceived of as an entity in the environment. But it is an interesting question whether the TCV can determine a q.i that is partly imagination. I think it would be possible. The tester would just have to be able to notice that the controlling involves an imaginary component and he or she would have to try to guess what that component was and how it was varying. For example, in the coin game you might notice that the controller seems to be imagining an extra coin in a position that completes the pattern under control. This could be tested by creating disturbances to the hypothetical controlled variable (call it q.i’) that can only be corrected by adding what you guess to be the imagined coin. This seems like a nice research project there for those interested in studying the role of imagination in control.

MT: If you control for X to be at a reference level R, and vary R appreciably, and someone else observing your environment says that they perceive something changing that they would also identify as X, that’s evidence that X or something very like it is in the environment.

RM: It’s consistent with the assumption that there are physical variables “out there” that are the basis of the perception of X. The fact that you and I perceive a particular combination of sugar, acid, oils and water as “lemonade” (X = the taste of lemonade) is not evidence that there is an entity in the environment that is the taste of lemonade; according to PCT, it is evidence that you and I have the same perceptual function that produces a perceptual signal that we experience as the taste of lemonade.

MT: Velocity and direction sensing is the job of “complex cells” in the primary visual cortex, which is presumably before any possibility of being incorporated in a control loop.

RM: Actually, velocity and direction sensing is presumed to be carried out by arrays of receptors called “receptive fields”, which are the perceptual functions in a control loop.

HB : Where did you get this theory ??? You think its’ the only one ???

The neurons that carry the outputs of the receptive field computations are called “simple”, “complex” or “hypercomplex”

HB : Who is calling it like this ???

…depending on the complexity of the variable computed by their associated receptive field function.

HB : Can you show us where did you get this ???

It’s not the cells themselves that are “simple” or “complex”; it’s the computations carried out by the receptive field “neural network”

HB : Can you be more specific what is »receptive filed neural network ???

that determines whether a cell (actually the axon that carries the output of the neural net computation) is “simple” or “complex”.

HB : And how do you determine this ?

Boris

Best

Rick

–

Richard S. Marken

"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.�
–Antoine de Saint-Exupery

[EetuP]

Chad

I took a quick look at that and interestingly the symmetrical thinking sounds familiar from many experiences but especially from the way how thinking of a new born child in Plooij’s Wonder weeks.

But have you some more concrete ideas how it could be leveraged?

Eetu
(Lähetetty kännykästä / Sent from mobile)

···

Chad T. Green Chad.Green@lcps.org kirjoitti 14.2.2017 kello 16.36:

[Chad Green (2017.02.14.0936 EST)]

EP: (Perhaps Chad was referring to something like this in his last message about controlling wholeness.)

CG: Wholeness to my mind is a reflection of unconscious logic at work, e.g., see:

https://en.wikipedia.org/wiki/Ignacio_Matte_Blanco . Why not leverage it?

Best,

Chad

From: Eetu Pikkarainen [mailto:eetu.pikkarainen@oulu.fi]
Sent: Tuesday, February 14, 2017 6:56 AM
To: csgnet@lists.illinois.edu
Subject: RE: What is Qi?

[eetu pikkarainen 2017-02-14]

Hi Vyv

VH: As i write this though I’m beginning to doubt the examples of thermostats or any other man made control system as clear illustrations of control in living systems. This is because machines input functions
come about in a totally different way to those of living things. They can be made to be much simpler; a cruise control can sense velocity via a single variable because of the way it is organised (a turning crank and some frequency counter). There is no way
a living system can sense velocity with a single environmental variable (without some sort of bio engineering). I wonder if this difference makes these examples hard for me understand. I find the two domains don’t map onto each other that well.

I feel that doubt, too, (but on the other hand I still enjoy the surprisingly close relation).

The most remarkable difference for me is of course that reference value is determined by an outside force (user, engineer) for the man-made control systems while for LCS the reference in “natural”, partly innate
but anyway suited for its living needs, and they can be further adjusted by systems themselves.

As for the perceptual input, Bruce already answered that a living system could sense velocity with a single environmental variable. But I do not think it is the essential question but rather that LCS can seldom
or never concentrate to perceive only one “thing” like velocity, but because of our needs of living we are always sensing a very large spectrum of variables which are then many ways mixed together and with memories in higher level perceptions. Those higher
level wholes than determine back to the reference levels and probably to input functions too.

(Perhaps Chad was referring to something like this in his last message about controlling wholeness.)

Eetu

[MartinT]

[Martin Taylor 2017.02.16.13.24]

I'm not going to quote or comment on much of this posting, because

it’s the take-away message with which I disagree.

[From Rick Marken (2017.02.15.1000)]

And so on. (Incidentally, I don't think anyone has ever said that

“the CEV is equivalent to qi”. Qi is a value, while CEV is a name
for the variable that has that value, just as “p” is a value for a
variable we usually call “perception”. Name and number, such as CEV
and qi, are different kinds of concept.)

The take-away I get from this message, together with several

previous ones, is that “according to [new] PCT”, the output action
of a control unit acts on nothing in the environment that relates to
the perception being controlled. My engineering mind is baffled by
how the perception of something in the environment is brought nearer
its reference value even though nothing in the environment is
influenced by the control unit’s output. I therefore do not accept
the message that “qi” and “p” are synonyms referring to the value of
an intrinsically unobservable variable inside the organism, usually
called “Perception”.

···

Martin Taylor (2017.02.13.17.57)-

            MT: The important point is about

degrees of freedom. In the Powers version of Perceptual
Control Theory, each perception has only one degree of
freedom, its value…

            MT: In the case of the CEV and the

perception, you can specify one of them, but then you
can’t independently specify the other without changing
something else.

          RM: If the CEV is equivalent to q.i, then this is true

since q.i is the same function of environmental variables
as p. But the truth of this has nothing to do the df.

            MT: You talk about the diagrams of

the loop in which the CEV is shown as a circle
containing components, and multiple lines lead from the
action output to the components in the CEV circle.

          RM: I've never seen such a diagram. It's certainly not

what is seen in the Science diagram where there is
only one output line connecting the effector function to
the input quantity via the feedback function. In general
there can be multiple inputs to a perceptual function but
only a single output from the output function in any
control loop.

            MT: But each CEV, considered by

itself, has exactly one degree of freedom, no matter how
many components and subcomponents it may have.

          RM: Yes, if the CEV is equivalent to q.i, then q.i is

the scalar (1 df) output of the same function as the one
that produces the scalar variable p.

            MT: We ask about the "reality" of

the CEV represented by the heavy circle. Since at every
level, the heavy circle represents a function of the
little circles, the “reality” question has to be given
the same answer at every level.

          RM: Yes, q.i is as real as p, because they are the same

function of environmental variables.

            MT: Now there's always the question

as to whether the CEV is actually in the environment.
It’s not a question one can ever answer assuredly
affirmatively.

          RM: If the CEV is equivalent to q.i, then it is

assuredly not in the environment (at least, according to
PCT); it is a perception in the observer. The physical
variables that are the basis of the perception of q.i and
p are in the environment.

[chadtgreen]

[Chad Green (2017.02.16.1429 EST)]

EP: But have you some more concrete ideas how it could be leveraged?

CG: I think you all are doing a wonderful job right here on CSGnet.

Best,

Chad

···

Chad T. Green, PMP
Research Office
Loudoun County Public Schools
21000 Education Court
Ashburn, VA 20148
Voice: 571-252-1486
Fax: 571-252-1575

“To the humble, courageous, ‘great’ ones among us who exemplify how leadership is a choice, not a position.” - Stephen Covey (The 8^th Habit)

From: Eetu Pikkarainen [mailto:eetu.pikkarainen@oulu.fi]
Sent: Thursday, February 16, 2017 4:03 AM
To: csgnet@lists.illinois.edu
Subject: Re: What is Qi?

Chad

I took a quick look at that and interestingly the symmetrical thinking sounds familiar from many experiences but especially from the way how thinking of a new born child in Plooij’s Wonder weeks.

But have you some more concrete ideas how it could be leveraged?

Eetu

(Lähetetty kännykästä / Sent from mobile)

Chad T. Green Chad.Green@lcps.org kirjoitti 14.2.2017 kello 16.36:

[Chad Green (2017.02.14.0936 EST)]

EP: (Perhaps Chad was referring to something like this in his last message about controlling wholeness.)

CG: Wholeness to my mind is a reflection of unconscious logic at work, e.g., see:

https://en.wikipedia.org/wiki/Ignacio_Matte_Blanco . Why not leverage it?

Best,

Chad

[rsmarken]

[From Rick Marken (2017.02.17.1255)]

···

Martin Taylor (2017.02.16.13.24)–

MT: (Incidentally, I don't think anyone has ever said that

“the CEV is equivalent to qi”. Qi is a value, while CEV is a name
for the variable that has that value, just as “p” is a value for a
variable we usually call “perception”. Name and number, such as CEV
and qi, are different kinds of concept.)

RM: Qi is a variable; the value of Qi is called “the value of Qi” or “the state of the input quantity”. p is also a variable; the value of p is called “the value of p” or “the state of the perceptual signal”.Â

Â

MT: The take-away I get from this message, together with several

previous ones, is that “according to [new] PCT”, the output action
of a control unit acts on nothing in the environment that relates to
the perception being controlled.

RM: This is the wrong take-away. I am well aware of the fact that a control system keeps the controlled quantity (Qi) in a reference state by acting on the environment (the v variables in the Science article) so as to move, p,the perceptual correlate of Qi, toward the state specified by the reference signal in the control system and keep it there, protected from disturbances. I’ve built many control system models that can do this controlling of input quantities, Qi, that are quite simple (like the position of a cursor relative to that of the target in a tracking task) or more complex (like the area of the rectangle in the “What is Size” demo). Â

Â

MT: My engineering mind is baffled by

how the perception of something in the environment is brought nearer
its reference value even though nothing in the environment is
influenced by the control unit’s output.

RM: As well it should.Â

Â

MT: I therefore do not accept

the message that “qi” and “p” are synonyms referring to the value of
an intrinsically unobservable variable inside the organism, usually
called “Perception”.

RM: As well you shouldn’t. The message you should accept is this: PCT is designed to account for the fact that organisms can be observed to keep variable aspects of the environment in reference states. For example, we observe that a person can keep a car in its lane despite variations in in the road that would be expected to quickly lead to the car going way out of its lane. The position of the car is a controlled quantity, Qi, and the reference state of  this variable is “in the lane”. PCT explains this observation (of the fact that Qi is controlled) by positing the existence of a control system that controls a perceptual signal, p, that is equivalent to Qi; p is a theoretical (and, thus, unobservable) variable that is a component of the explanation of the behavior of the observable variable, Qi.Â

RM: An excellent explanation of this same message can be found on pp. 171 - 176 of LCS I in the section entitled “The Phenomenon of Control”.Â

Â

---

MT: As an aside, I should note that this idea from newPCT, that qi is

inside the organism rather than in the environment where it has
always been in old PCT, is very recent.

RM: Actually, it has always been there. For example, in a post I found from Bill Powers back in 1998 ([From Bill Powers (981016.0952 MDT)]) Bill explains that, when doing the test for the controlled variable, the controlled variable is a perception from the point of view of both the tester (observer) and the controller. The controlled variable from the point of view of the observer is Qi – the controlled quantity. The controlled variable from the point of view of the controller is p – the (theoretical) perceptual signal.Â

RM: I think that’s enough for now. I am still looking forward to hearing your caveats about how to do the TCV. And I would especially like to know how you came up with them. This would be great material for a textbook on research methodology for studying living control systems.Â

BestÂ

Rick

Â

So far as I remember (and my

memory is notoriously faulty these days), it appeared in the revised
canon only after I introduced my gedanken experiment that
illustrated stochastic collective control. Maybe it was shortly
after I asked why giving a name to a hitherto un-named environmental
variable that had the value qi was an issue at all. Either way, my
use of the acronym (which we now expand as “Corresponding
Environmental Variable”) seems to have been the initial disturbance
to some perception Rick controls. The output action to counter that
disturbance apparently was the invention of newPCT. On such things
does the advancement of science depend.

But newPCT, with its entanglement of observer and controller and all

the other new ideas, is sufficiently recent as a conceptual
structure, that I still don’t understand even its logical or
mathematical foundations. It needs further development and a clear
exposition of its differences from the Powers version of PCT before
it can be properly explained to the world at large. If it is truly
useful, that would be a great asset to the world, going well beyond
Powers’s ideas.

Martin

–
Richard S. MarkenÂ

"Perfection is achieved not when you have nothing more to add, but when you
have nothing left to take away.�
                --Antoine de Saint-Exupery

[AngusJ]

[From Angus Jenkinson. 2017.9.1].

I came across this recently. In it there is the remark by BA

BA: I don’t see any essential difference between human-engineered control systems and biological ones; in fact it is that correspondence
that allowed Bill Powers (and others) to apply control-system principles to biological systems. What is true is that biological systems are products of evolution, a process that involves gradual modification of what is already there. Unlike a human designer,
evolution does not begin with a fresh sheet of paper. What results is more like a hodge-podge of alterations and fixes than a clean design, and this can make it extremely difficult to tease out what the parts of a given system are and how they interconnect.
Bill Powers notes, for example, that alpha motor neurons in the spinal cord combine the functions of comparator and output function (and even that is probably an oversimplification). So it may be more difficult to figure out the system diagrams in biological
systems than in human-engineered systems, but that does not imply that the underlying system principles of the two are necessarily different. PCT is based on the well-supported assumption that behavior can be explained by control-system models.

I take a rather different view, and it is important for my research. Leaving aside the questionable comments about the process of evolution (but bearing
in mind Einstein’s comment that anything seems complex/complicated until it is understood), my concern is with the assumption of identity between biological systems and mechanical systems (leaving aside human engineering in the genetic/biological space).

It seems to me to be precisely the case that PCT actually demonstrates the difference between the two domains. While Bill Powers started off as a control
engineer, he also went the path of cybernetics, which led to new understanding, particularly in the biological, psychological, and sociological spaces. PCT leads to an understanding of autonomous purposeful behaviour via the control agency. Any mechanical
device that has a control system (governor) that enables regulation of behaviour according to sensory/signal input (like the steam engine) is capable of PCT-type behaviour on the dimensions controlled by the governor. These are normally fairly simple. And
to my knowledge, most/all of them were created by human beings so that the mechanical device makes use of biological principles.

What examples are there in the nonbiological, non-human-designed, material world of PCT behaviour?

···

………â¦â€¦â€¦………………………………………………………………….

Angus Jenkinson

On 12/02/2017, 23:19, “Bruce Abbott” bbabbott@frontier.com wrote:

[From Bruce Abbott (2017.02.12.1820 EST)]

[Vyv Huddy 1955.12.02.2017]

VH: This is great set of posts on this thread and the previous one. Particularly helpful to see the posts by rick and bruce describing the figure in the Powers Science paper showing multiple V. I noticed that and missed why it was important.

VH: I don’t have my books with me now but i recall Bill Powers (i think) wrote somewhere that the controlled variable of a thermostat is actually the amount of coil of the strip not the temperature of the room itself. This is partly shown
by putting a flame directly under the thermostat - the furnace goes off without the room temperature changing.

BA: The controlled variable is the temperature of the air. The position of the coil (actually, the contact on the free end of the coil) is the thermostat’s perception of that temperature. The position of the other contact sets the thermostat’s
reference level. When the contacts meet, this generates a nonzero error signal indicating that the room temperature has fallen below the reference level. This causes the furnace to switch on (output function), heating the air and thus bringing the room temperature
up.

BA: The thermostat is a sensor of room temperature only to the extent that the coil temperature matches the room temperature. All sensors work this way, sensing a variable through its effect on the sensor. In this example changing temperatures
of the air are conveyed to the coil, producing different expansions on the two sides of the bimetallic strip, thus moving the contact on the free end. In the case of vision, photons striking molecules of photochemical in the photoreceptors cause light-sensitive
chemicals within the photoreceptors to break down, which through a series of chemical events causes the “generator potential� of the photoreceptor to change. This in turn alters the rate of firing of associated neurons.

VH: If so the controlled variable (qi) of a cruise control would be the frequency of the counting wheel turns in the speedometer mechanism? If so from that perspective there is no “actual” speed; cruise controls don’t work effectively on
a slippery surfaces and drivers are encouraged not to use them then.

BA: One could design cruise controls to sense the car’s speed in a variety of ways. The typical example is that speed is sensed indirectly by counting the rotational frequency of the car’s drive shaft. Under normal conditions this is
proportional to the speed of the car. This relationship breaks down if the wheels slip, as you note. This problem could be avoided by using an optical sensor to read the “optic flow� of the road beneath the car. However measured, there is still an actual
speed that is being measured, whether directly or indirectly, so long as the expected relationships hold. In the absence of wheel slippage, driveshaft rotations map directly onto the car’s speed. Biological systems experience the same difficulties; sensor
reports do not always match reality. Our sensors remain useful, however, because most of the time they provide readings that are good enough to go by.

VH: As i write this though I’m beginning to doubt the examples of thermostats or any other man made control system as clear illustrations of control in living systems. This is because machines input functions come about in a totally different
way to those of living things. They can be made to be much simpler; a cruise control can sense velocity via a single variable because of the way it is organised (a turning crank and some frequency counter). There is no way a living system can sense velocity
with a single environmental variable (without some sort of bio engineering).

BA: Why not? Optic flow will do it.

VH: I wonder if this difference makes these examples hard for me understand. I find the two domains don’t map onto each other that well.

BA: I don’t see any essential difference between human-engineered control systems and biological ones; in fact it is that correspondence that allowed Bill Powers (and others) to apply control-system principles to biological systems. What
is true is that biological systems are products of evolution, a process that involves gradual modification of what is already there. Unlike a human designer, evolution does not begin with a fresh sheet of paper. What results is more like a hodge-podge of
alterations and fixes than a clean design, and this can make it extremely difficult to tease out what the parts of a given system are and how they interconnect. Bill Powers notes, for example, that alpha motor neurons in the spinal cord combine the functions
of comparator and output function (and even that is probably an oversimplification). So it may be more difficult to figure out the system diagrams in biological systems than in human-engineered systems, but that does not imply that the underlying system principles
of the two are necessarily different. PCT is based on the well-supported assumption that behavior can be explained by control-system models.

VH: If cruise control p is really a single voltage in wire then it would be only intensity controller? For me Intensity control is best explained in the context of the muscle tone example, as in b:cp.

BA: In HPCT, perceptual signals at all levels are equivalent to single voltages in a wire (i.e., average neural current in a nerve). Thus, the fact that p in cruise control is embodied as a single voltage in a wire does not make the cruise
control an intensity controller.

In fact, all controllers in HPCT, at whatever level, are intensity controllers in the sense that they act to control the intensity of their neural signal p’s.

Bruce

[bnhpct]

[From Bruce Nevin (2017.09.03.14:20)]

Angus Jenkinson. 2017.9.1 –

···

On Fri, Sep 1, 2017 at 11:49 AM, Angus Jenkinson angus@angusjenkinson.com wrote:

[From Angus Jenkinson. 2017.9.1].

Â

I came across this recently. In it there is the remark by BA

BA: I don’t see any essential difference between human-engineered control systems and biological ones; in fact it is that correspondence
that allowed Bill Powers (and others) to apply control-system principles to biological systems. What is true is that biological systems are products of evolution, a process that involves gradual modification of what is already there. Unlike a human designer,
evolution does not begin with a fresh sheet of paper. What results is more like a hodge-podge of alterations and fixes than a clean design, and this can make it extremely difficult to tease out what the parts of a given system are and how they interconnect.Â
Bill Powers notes, for example, that alpha motor neurons in the spinal cord combine the functions of comparator and output function (and even that is probably an oversimplification). So it may be more difficult to figure out the system diagrams in biological
systems than in human-engineered systems, but that does not imply that the underlying system principles of the two are necessarily different. PCT is based on the well-supported assumption that behavior can be explained by control-system models.

Â

I take a rather different view, and it is important for my research. Leaving aside the questionable comments about the process of evolution (but bearing
in mind Einstein’s comment that anything seems complex/complicated until it is understood), my concern is with the assumption of identity between biological systems and mechanical systems (leaving aside human engineering in the genetic/biological space).

Â

It seems to me to be precisely the case that PCT actually demonstrates the difference between the two domains. While Bill Powers started off as a control
engineer, he also went the path of cybernetics, which led to new understanding, particularly in the biological, psychological, and sociological spaces. PCT leads to an understanding of autonomous purposeful behaviour via the control agency. Any mechanical
device that has a control system (governor) that enables regulation of behaviour according to sensory/signal input (like the steam engine) is capable of PCT-type behaviour on the dimensions controlled by the governor. These are normally fairly simple. And
to my knowledge, most/all of them were created by human beings so that the mechanical device makes use of biological principles.

Â

What examples are there in the nonbiological, non-human-designed, material world of PCT behaviour?

Â

………€¦â€¦……………………………………………………… ¦â€¦â€¦â€¦â€¦.

Angus Jenkinson

Â

Â

Â

On 12/02/2017, 23:19, “Bruce Abbott” bbabbott@frontier.com wrote:

Â

[From Bruce Abbott (2017.02.12.1820 EST)]

Â

[Vyv Huddy 1955.12.02.2017]

Â

VH: This is great set of posts on this thread and the previous one. Particularly helpful to see the posts by rick and bruce describing the figure in the Powers Science paper showing multiple V. I noticed that and missed why it was important.Â

Â

VH: I don’t have my books with me now but i recall Bill Powers (i think) wrote somewhere that the controlled variable of a thermostat is actually the amount of coil of the strip not the temperature of the room itself. This is partly shown
by putting a flame directly under the thermostat - the furnace goes off without the room temperature changing.

Â

BA: The controlled variable is the temperature of the air. The position of the coil (actually, the contact on the free end of the coil) is the thermostat’s perception of that temperature. The position of the other contact sets the thermostat’s
reference level. When the contacts meet, this generates a nonzero error signal indicating that the room temperature has fallen below the reference level. This causes the furnace to switch on (output function), heating the air and thus bringing the room temperature
up.

Â

BA: The thermostat is a sensor of room temperature only to the extent that the coil temperature matches the room temperature. All sensors work this way, sensing a variable through its effect on the sensor. In this example changing temperatures
of the air are conveyed to the coil, producing different expansions on the two sides of the bimetallic strip, thus moving the contact on the free end. In the case of vision, photons striking molecules of photochemical in the photoreceptors cause light-sensitive
chemicals within the photoreceptors to break down, which through a series of chemical events causes the “generator potential� of the photoreceptor to change. This in turn alters the rate of firing of associated neurons.

Â

VH: If so the controlled variable (qi) of a cruise control would be the frequency of the counting wheel turns in the speedometer mechanism? If so from that perspective there is no “actual” speed; cruise controls don’t work effectively on
a slippery surfaces and drivers are encouraged not to use them then.

Â

BA:  One could design cruise controls to sense the car’s speed in a variety of ways. The typical example is that speed is sensed indirectly by counting the rotational frequency of the car’s drive shaft. Under normal conditions this is
proportional to the speed of the car. This relationship breaks down if the wheels slip, as you note. This problem could be avoided by using an optical sensor to read the “optic flow� of the road beneath the car. However measured, there is still an actual
speed that is being measured, whether directly or indirectly, so long as the expected relationships hold. In the absence of wheel slippage, driveshaft rotations map directly onto the car’s speed. Biological systems experience the same difficulties; sensor
reports do not always match reality. Our sensors remain useful, however, because most of the time they provide readings that are good enough to go by.

Â

VH: As i write this though I’m beginning to doubt the examples of thermostats or any other man made control system as clear illustrations of control in living systems. This is because machines input functions come about in a totally different
way to those of living things. They can be made to be much simpler; a cruise control can sense velocity via a single variable because of the way it is organised (a turning crank and some frequency counter). There is no way a living system can sense velocity
with a single environmental variable (without some sort of bio engineering).

Â

BA: Why not? Optic flow will do it.

Â

VH: I wonder if this difference makes these examples hard for me understand. I find the two domains don’t map onto each other that well.Â

Â

BA: I don’t see any essential difference between human-engineered control systems and biological ones; in fact it is that correspondence that allowed Bill Powers (and others) to apply control-system principles to biological systems. What
is true is that biological systems are products of evolution, a process that involves gradual modification of what is already there. Unlike a human designer, evolution does not begin with a fresh sheet of paper. What results is more like a hodge-podge of
alterations and fixes than a clean design, and this can make it extremely difficult to tease out what the parts of a given system are and how they interconnect. Bill Powers notes, for example, that alpha motor neurons in the spinal cord combine the functions
of comparator and output function (and even that is probably an oversimplification). So it may be more difficult to figure out the system diagrams in biological systems than in human-engineered systems, but that does not imply that the underlying system principles
of the two are necessarily different. PCT is based on the well-supported assumption that behavior can be explained by control-system models.

Â

VH: If cruise control p is really a single voltage in wire then it would be only intensity controller? For me Intensity control is best explained in the context of the muscle tone example, as in b:cp.Â

Â

BA: In HPCT, perceptual signals at all levels are equivalent to single voltages in a wire (i.e., average neural current in a nerve). Thus, the fact that p in cruise control is embodied as a single voltage in a wire does not make the cruise
control an intensity controller.

Â

In fact, all controllers in HPCT, at whatever level, are intensity controllers in the sense that they act to control the intensity of their neural signal p’s.

Â

 Bruce

[bnhpct]

[From Bruce Nevin (217.09.03.14:55)]

Angus Jenkinson. 2017.9.1 –

···

On Fri, Sep 1, 2017 at 11:49 AM, Angus Jenkinson angus@angusjenkinson.com wrote:

[From Angus Jenkinson. 2017.9.1].

Â

I came across this recently. In it there is the remark by BA

BA: I don’t see any essential difference between human-engineered control systems and biological ones; in fact it is that correspondence
that allowed Bill Powers (and others) to apply control-system principles to biological systems. What is true is that biological systems are products of evolution, a process that involves gradual modification of what is already there. Unlike a human designer,
evolution does not begin with a fresh sheet of paper. What results is more like a hodge-podge of alterations and fixes than a clean design, and this can make it extremely difficult to tease out what the parts of a given system are and how they interconnect.Â
Bill Powers notes, for example, that alpha motor neurons in the spinal cord combine the functions of comparator and output function (and even that is probably an oversimplification). So it may be more difficult to figure out the system diagrams in biological
systems than in human-engineered systems, but that does not imply that the underlying system principles of the two are necessarily different. PCT is based on the well-supported assumption that behavior can be explained by control-system models.

Â

I take a rather different view, and it is important for my research. Leaving aside the questionable comments about the process of evolution (but bearing
in mind Einstein’s comment that anything seems complex/complicated until it is understood), my concern is with the assumption of identity between biological systems and mechanical systems (leaving aside human engineering in the genetic/biological space).

Â

It seems to me to be precisely the case that PCT actually demonstrates the difference between the two domains. While Bill Powers started off as a control
engineer, he also went the path of cybernetics, which led to new understanding, particularly in the biological, psychological, and sociological spaces. PCT leads to an understanding of autonomous purposeful behaviour via the control agency. Any mechanical
device that has a control system (governor) that enables regulation of behaviour according to sensory/signal input (like the steam engine) is capable of PCT-type behaviour on the dimensions controlled by the governor. These are normally fairly simple. And
to my knowledge, most/all of them were created by human beings so that the mechanical device makes use of biological principles.

Â

What examples are there in the nonbiological, non-human-designed, material world of PCT behaviour?

Â

………â€â€¦â€¦……………………………………………………….………….

Angus Jenkinson

Â

Â

Â

On 12/02/2017, 23:19, “Bruce Abbott” bbabbott@frontier.com wrote:

Â

[From Bruce Abbott (2017.02.12.1820 EST)]

Â

[Vyv Huddy 1955.12.02.2017]

Â

VH: This is great set of posts on this thread and the previous one. Particularly helpful to see the posts by rick and bruce describing the figure in the Powers Science paper showing multiple V. I noticed that and missed why it was important.Â

Â

VH: I don’t have my books with me now but i recall Bill Powers (i think) wrote somewhere that the controlled variable of a thermostat is actually the amount of coil of the strip not the temperature of the room itself. This is partly shown
by putting a flame directly under the thermostat - the furnace goes off without the room temperature changing.

Â

BA: The controlled variable is the temperature of the air. The position of the coil (actually, the contact on the free end of the coil) is the thermostat’s perception of that temperature. The position of the other contact sets the thermostat’s
reference level. When the contacts meet, this generates a nonzero error signal indicating that the room temperature has fallen below the reference level. This causes the furnace to switch on (output function), heating the air and thus bringing the room temperature
up.

Â

BA: The thermostat is a sensor of room temperature only to the extent that the coil temperature matches the room temperature. All sensors work this way, sensing a variable through its effect on the sensor. In this example changing temperatures
of the air are conveyed to the coil, producing different expansions on the two sides of the bimetallic strip, thus moving the contact on the free end. In the case of vision, photons striking molecules of photochemical in the photoreceptors cause light-sensitive
chemicals within the photoreceptors to break down, which through a series of chemical events causes the “generator potential� of the photoreceptor to change. This in turn alters the rate of firing of associated neurons.

Â

VH: If so the controlled variable (qi) of a cruise control would be the frequency of the counting wheel turns in the speedometer mechanism? If so from that perspective there is no “actual” speed; cruise controls don’t work effectively on
a slippery surfaces and drivers are encouraged not to use them then.

Â

BA:  One could design cruise controls to sense the car’s speed in a variety of ways. The typical example is that speed is sensed indirectly by counting the rotational frequency of the car’s drive shaft. Under normal conditions this is
proportional to the speed of the car. This relationship breaks down if the wheels slip, as you note. This problem could be avoided by using an optical sensor to read the “optic flow� of the road beneath the car. However measured, there is still an actual
speed that is being measured, whether directly or indirectly, so long as the expected relationships hold. In the absence of wheel slippage, driveshaft rotations map directly onto the car’s speed. Biological systems experience the same difficulties; sensor
reports do not always match reality. Our sensors remain useful, however, because most of the time they provide readings that are good enough to go by.

Â

VH: As i write this though I’m beginning to doubt the examples of thermostats or any other man made control system as clear illustrations of control in living systems. This is because machines input functions come about in a totally different
way to those of living things. They can be made to be much simpler; a cruise control can sense velocity via a single variable because of the way it is organised (a turning crank and some frequency counter). There is no way a living system can sense velocity
with a single environmental variable (without some sort of bio engineering).

Â

BA: Why not? Optic flow will do it.

Â

VH: I wonder if this difference makes these examples hard for me understand. I find the two domains don’t map onto each other that well.Â

Â

BA: I don’t see any essential difference between human-engineered control systems and biological ones; in fact it is that correspondence that allowed Bill Powers (and others) to apply control-system principles to biological systems. What
is true is that biological systems are products of evolution, a process that involves gradual modification of what is already there. Unlike a human designer, evolution does not begin with a fresh sheet of paper. What results is more like a hodge-podge of
alterations and fixes than a clean design, and this can make it extremely difficult to tease out what the parts of a given system are and how they interconnect. Bill Powers notes, for example, that alpha motor neurons in the spinal cord combine the functions
of comparator and output function (and even that is probably an oversimplification). So it may be more difficult to figure out the system diagrams in biological systems than in human-engineered systems, but that does not imply that the underlying system principles
of the two are necessarily different. PCT is based on the well-supported assumption that behavior can be explained by control-system models.

Â

VH: If cruise control p is really a single voltage in wire then it would be only intensity controller? For me Intensity control is best explained in the context of the muscle tone example, as in b:cp.Â

Â

BA: In HPCT, perceptual signals at all levels are equivalent to single voltages in a wire (i.e., average neural current in a nerve). Thus, the fact that p in cruise control is embodied as a single voltage in a wire does not make the cruise
control an intensity controller.

Â

In fact, all controllers in HPCT, at whatever level, are intensity controllers in the sense that they act to control the intensity of their neural signal p’s.

Â

 Bruce

[FredNickols]

[From Fred Nickols (2017.09.03.1521 ET)]

Angus:

I might have an example but I’ll let you be the judge of that.

When I joined the Navy in 1955 I was trained as a fire control technician. The “fire� in question was gunfire from the big guns found on warships. I won’t go into the intricacies of solving the fire control problem (namely, figuring out how to hit a moving target) but suffice it to say that the many calculations performed by the computer at the heart of a shipboard gunfire control system resulted in a set of orders to the gun mounts. The orders sent to the gun mounts constituted reference signals (one for horizontal positioning of the gun mount or “gun train order� and one for vertical positioning of the gun barrel or “gun elevation order.�

Servomechanisms and motor generators served to move the gun mount and its barrel and to ascertain (i.e., perceive) its current positions and compare those with the ordered positions. The “disturbances� included the pitch and roll of the ship, friction between the gun mount and the base ring upon which it sat and turned and any sudden, sharp maneuvers being made by the ship.

Is that the kind of example you are seeking?

Fred Nickols

···

From: Angus Jenkinson [mailto:angus@angusjenkinson.com]
Sent: Friday, September 1, 2017 11:50 AM
To: csgnet@lists.illinois.edu
Subject: Re: What is Qi?

[From Angus Jenkinson. 2017.9.1].

I came across this recently. In it there is the remark by BA

BA: I don’t see any essential difference between human-engineered control systems and biological ones; in fact it is that correspondence that allowed Bill Powers (and others) to apply control-system principles to biological systems. What is true is that biological systems are products of evolution, a process that involves gradual modification of what is already there. Unlike a human designer, evolution does not begin with a fresh sheet of paper. What results is more like a hodge-podge of alterations and fixes than a clean design, and this can make it extremely difficult to tease out what the parts of a given system are and how they interconnect. Bill Powers notes, for example, that alpha motor neurons in the spinal cord combine the functions of comparator and output function (and even that is probably an oversimplification). So it may be more difficult to figure out the system diagrams in biological systems than in human-engineered systems, but that does not imply that the underlying system principles of the two are necessarily different. PCT is based on the well-supported assumption that behavior can be explained by control-system models.

I take a rather different view, and it is important for my research. Leaving aside the questionable comments about the process of evolution (but bearing in mind Einstein’s comment that anything seems complex/complicated until it is understood), my concern is with the assumption of identity between biological systems and mechanical systems (leaving aside human engineering in the genetic/biological space).

It seems to me to be precisely the case that PCT actually demonstrates the difference between the two domains. While Bill Powers started off as a control engineer, he also went the path of cybernetics, which led to new understanding, particularly in the biological, psychological, and sociological spaces. PCT leads to an understanding of autonomous purposeful behaviour via the control agency. Any mechanical device that has a control system (governor) that enables regulation of behaviour according to sensory/signal input (like the steam engine) is capable of PCT-type behaviour on the dimensions controlled by the governor. These are normally fairly simple. And to my knowledge, most/all of them were created by human beings so that the mechanical device makes use of biological principles.

What examples are there in the nonbiological, non-human-designed, material world of PCT behaviour?

………………………………………………………………………………….

Angus Jenkinson

On 12/02/2017, 23:19, “Bruce Abbott” bbabbott@frontier.com wrote:

[From Bruce Abbott (2017.02.12.1820 EST)]

[Vyv Huddy 1955.12.02.2017]

VH: This is great set of posts on this thread and the previous one. Particularly helpful to see the posts by rick and bruce describing the figure in the Powers Science paper showing multiple V. I noticed that and missed why it was important.

VH: I don’t have my books with me now but i recall Bill Powers (i think) wrote somewhere that the controlled variable of a thermostat is actually the amount of coil of the strip not the temperature of the room itself. This is partly shown by putting a flame directly under the thermostat - the furnace goes off without the room temperature changing.

BA: The controlled variable is the temperature of the air. The position of the coil (actually, the contact on the free end of the coil) is the thermostat’s perception of that temperature. The position of the other contact sets the thermostat’s reference level. When the contacts meet, this generates a nonzero error signal indicating that the room temperature has fallen below the reference level. This causes the furnace to switch on (output function), heating the air and thus bringing the room temperature up.

BA: The thermostat is a sensor of room temperature only to the extent that the coil temperature matches the room temperature. All sensors work this way, sensing a variable through its effect on the sensor. In this example changing temperatures of the air are conveyed to the coil, producing different expansions on the two sides of the bimetallic strip, thus moving the contact on the free end. In the case of vision, photons striking molecules of photochemical in the photoreceptors cause light-sensitive chemicals within the photoreceptors to break down, which through a series of chemical events causes the “generator potential� of the photoreceptor to change. This in turn alters the rate of firing of associated neurons.

VH: If so the controlled variable (qi) of a cruise control would be the frequency of the counting wheel turns in the speedometer mechanism? If so from that perspective there is no “actual” speed; cruise controls don’t work effectively on a slippery surfaces and drivers are encouraged not to use them then.

BA: One could design cruise controls to sense the car’s speed in a variety of ways. The typical example is that speed is sensed indirectly by counting the rotational frequency of the car’s drive shaft. Under normal conditions this is proportional to the speed of the car. This relationship breaks down if the wheels slip, as you note. This problem could be avoided by using an optical sensor to read the “optic flow� of the road beneath the car. However measured, there is still an actual speed that is being measured, whether directly or indirectly, so long as the expected relationships hold. In the absence of wheel slippage, driveshaft rotations map directly onto the car’s speed. Biological systems experience the same difficulties; sensor reports do not always match reality. Our sensors remain useful, however, because most of the time they provide readings that are good enough to go by.

VH: As i write this though I’m beginning to doubt the examples of thermostats or any other man made control system as clear illustrations of control in living systems. This is because machines input functions come about in a totally different way to those of living things. They can be made to be much simpler; a cruise control can sense velocity via a single variable because of the way it is organised (a turning crank and some frequency counter). There is no way a living system can sense velocity with a single environmental variable (without some sort of bio engineering).

BA: Why not? Optic flow will do it.

VH: I wonder if this difference makes these examples hard for me understand. I find the two domains don’t map onto each other that well.

BA: I don’t see any essential difference between human-engineered control systems and biological ones; in fact it is that correspondence that allowed Bill Powers (and others) to apply control-system principles to biological systems. What is true is that biological systems are products of evolution, a process that involves gradual modification of what is already there. Unlike a human designer, evolution does not begin with a fresh sheet of paper. What results is more like a hodge-podge of alterations and fixes than a clean design, and this can make it extremely difficult to tease out what the parts of a given system are and how they interconnect. Bill Powers notes, for example, that alpha motor neurons in the spinal cord combine the functions of comparator and output function (and even that is probably an oversimplification). So it may be more difficult to figure out the system diagrams in biological systems than in human-engineered systems, but that does not imply that the underlying system principles of the two are necessarily different. PCT is based on the well-supported assumption that behavior can be explained by control-system models.

VH: If cruise control p is really a single voltage in wire then it would be only intensity controller? For me Intensity control is best explained in the context of the muscle tone example, as in b:cp.

BA: In HPCT, perceptual signals at all levels are equivalent to single voltages in a wire (i.e., average neural current in a nerve). Thus, the fact that p in cruise control is embodied as a single voltage in a wire does not make the cruise control an intensity controller.

In fact, all controllers in HPCT, at whatever level, are intensity controllers in the sense that they act to control the intensity of their neural signal p’s.

Bruce