TCV and Collective Control : A gedanken Experiment

[Martin Taylor 2017.01.21.11.53]

  I said I would answer this on Jan 20. I apologise for not have

done so. I was too shaken up after hearing a speech the like of
which I have not heard live in my life, and on film only with the
refrain “Deutcshland, Deutschland uber alles” in speeches around
the time of my birth. I have not been so scared for the future of
the world, ever, scary though the climate-change denier strength
has been for the last twenty or thirty years.

  But having had a good night's sleep, I will answer now as I

promised.

Here are the two collectively controlled social stabilities I was

thinking of that people can use in controlling other perceptions.
One is the identity of a President of a country, the other is the
means by which that person is collectively perceived to be the
President (election, military power, whatever). Whatever the means,
it takes the actions of more than that person to stabilize the
identity for long enough that people can use it for other purposes,
such as controlling for a particular law to be in force. There are
many other collectively controlled stabilities associated with this
whole process, but maybe you get the idea.
In another message, Rick could not see the analogy with the gedanken
experiment’s pebbles and Puck. Think of a vote as a pebble, the
possible identity of the President as a reference location on a
table that is domed rather than flat, so that the stronger “stream
of pebbles” moves the “Puck” off the flattish top onto a slope down
which the Puck slides to one reference location or the other (or to
one of many in an election in which many candidates participate).
Martin

···

[From Rick Marken (2016.12.31.1420)]

    RM: But maybe next year (like tomorrow;-) you could answer

your exercise on collectively controlled stability – the one
that I was unable to answer. Here it is again, in case you
forgot:

      MT: Since there is a potentially

world-altering example of a collectively controlled stability
enabling another collectively controlled stability going on in
the USA over these few weeks, I think I will use a Bill Powers
tutorial tactic here: I leave it as an exercise for the reader
to specify the two stabilities and their collective control
explanation. I will give one hint toward one of the answers,
though it shouldn’t be necessary: the word “authority”.

  RM: The hint didn't help me at all. So I look forward to hearing

the answer.

[From Rick Marken (2017.01.21.1710 PST)]

···

Martin Taylor (2017.01.21.11.53)–

  MT: I said I would answer this on Jan 20. I apologise for not have

done so. I was too shaken up after hearing a speech the like of
which I have not heard live in my life, and on film only with the
refrain “Deutcshland, Deutschland uber alles” in speeches around
the time of my birth.

RM: Well, he probably won’t invade Canada. And besides you’ve got yhat hedge running along the whole border, right;-) And there is pretty hefty worldwide opposition to this scuzbag. Although I was unable to attend the march I assuaged my guilt by calling to thank my congressmen (one’s a woman) for not attending the inauguration; I also derived some guilt relief seeing all those pussy hats at the DC march. My wife made most of them, or at least so it seems; she was knitting away like Madame Defarge;-)Â

Â

MT: Here are the two collectively controlled social stabilities I was

thinking of that people can use in controlling other perceptions.
One is the identity of a President of a country, the other is the
means by which that person is collectively perceived to be the
President (election, military power, whatever). Whatever the means,
it takes the actions of more than that person to stabilize the
identity for long enough that people can use it for other purposes,
such as controlling for a particular law to be in force. There are
many other collectively controlled stabilities associated with this
whole process, but maybe you get the idea.

RM: I agree that both the identity of the President and the means by which the President is selected are controlled variables, the state of which depends on the action of many people; so they are collectively controlled variables. I’ve never heard of a collectively controlled stability, however. Maybe it’s one of those newfangled concepts, like CEV;-)

Â

MT: In another message, Rick could not see the analogy with the gedanken

experiment’s pebbles and Puck. Think of a vote as a pebble, the
possible identity of the President as a reference location on a
table that is domed rather than flat, so that the stronger “stream
of pebbles” moves the “Puck” off the flattish top onto a slope down
which the Puck slides to one reference location or the other (or to
one of many in an election in which many candidates participate).

RM: Yes, this model would work (the domed table is a feedback function, of course; see how much better that sounds than atenfel). But then it’s not the same as your pebble situation, where there was no table. It also differs from Kent’s model, which achieves control by having a collection of controllers pushing the same variable toward different references with approximately equal gain, the result being that the variable is maintained in a virtual reference state.Â

RM: In collective control of who is President and how she is elected the controlled variable ends up, not in a virtual reference state but in a reference state that corresponds to the reference specifications of some subset of the individuals in the collective. In the recent US election the President was not someone between Trump and Clinton (though somewhat closer to Clinton since more people voted for her). What we got was just what a minority of the voters wanted: Trump. And we got it not because the means used to select the President represented some virtual state between the different means desired by all the voters; the means was the electoral college – which is preferred by a small minority of the voters who are in rural states – what we on the coasts now refer to as Dumbfuckistan (my old home state of Minnesota is, fortunately, excluded, though just barely).Â

RM: The models needed to explain the collective control of who is President and how she is selected, therefore, have to be considerably more complex than Kent’s model of many controllers acting to get the same variable to different reference states. You have to include the fact that many people are controlling for living in a particular kind of society (like a democracy) by controlling for certain principles (like cooperation) that are controlled by following certain rules (like “tit for tat”), etc. I think such models are needed to explain almost any kind of collective control. Again, I can’t think of any collectively controlled variable that can be understood in terms of Kent’s model. But maybe you can try thinking of one again.Â

Best

Rick

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

[Martin Taylor 2017.01.25.14.13]

[From Rick Marken (2017.01.21.1710 PST)]

CEV - First used on CSG-L in June 1992 (so far as I can find, in a

context that suggests it was not then a new term). And I doubt you
would be calling a “collectively controlled stability” newfangled if
you had read the material Kent keeps offering you.

Here's something I wrote in a message I composed to give you a few

more hints a week or so before the inauguration, but apparently did
not send (or maybe I did send it, because there’s another (long)
message that I can’t find that I thought I sent right after (or
right before) that, in the “Learning” thread. Both were just when
you were testing CSGnet). (The message says “ten days” but if I sent
it, it can’t have been on January 10).

-----quote----
Let me offer a few more examples of collectively controlled

stabilities that people can use in controlling other variables: The
meaning of a word, the text of a law, the authority to appoint
someone to an office, the rights of ownership, the value of a
dollar, national borders. I could go on for a long time, and Kent
has listed many others. Maybe these hints might help with your quest
for the collectively controlled variables to which January 20 are
relevant. You have another ten days to think a little more deeply
before the answer is revealed.

-----end quote----

I use the term "stability" in order to avoid any restrictive

connotations. In general it’s the stability itself that matters, not
the nature of what is stabilized, which matters for specific
controlled perceptions. In the case of the election method, if it
were not collectively controlled it would not be stable, and if it
were not stable citizens would have no means of registering their
preferences, thereby collectively controlling the identity of the
winning candidate. (To make it unstable seems to be one of Trump’s
objectives now, in preparation for 2020.)

Kent sometimes uses collectively controlled stability in the sense

of maintenance, such as controlling for a road surface to remain
smooth enough to drive on at reasonable speed without damaging a
car.

Yes, it sounds better, but that's OK. "Snowflake" also sounds

better, but likewise doesn’t mean the same as “atenfel”. The problem
is that both “feedback function” and “snowflake” are conventionally
used to refer to things that are not atenfels, so it gets a bit
confusing if you use them indiscriminately to refer to completely
different concepts. Incidentally, the domed table is not a feedback
function, either. It’s a table.

You are not talking about Kent's model, but about Kent's

demonstration 23 years ago. There’s a difference. I believe I listed
six kinds of collective control a few months ago, of which Kent’s
demo is a trivial example of one type.

I suppose it's a bit too much to ask what aspect of collective

control you are trying to contradict with this statement?

I'll stop here, because I sense too great a risk of a pissing match.

Martin
···

Martin Taylor (2017.01.21.11.53)–

            MT: Here are the two collectively

controlled social stabilities I was thinking of that
people can use in controlling other perceptions. One is
the identity of a President of a country, the other is
the means by which that person is collectively perceived
to be the President (election, military power,
whatever). Whatever the means, it takes the actions of
more than that person to stabilize the identity for long
enough that people can use it for other purposes, such
as controlling for a particular law to be in force.
There are many other collectively controlled stabilities
associated with this whole process, but maybe you get
the idea.

          RM: I agree that both the identity of the President and

the means by which the President is selected are
controlled variables, the state of which depends on the
action of many people; so they are collectively controlled
variables. I’ve never heard of a collectively controlled
stability, however. Maybe it’s one of those newfangled
concepts, like CEV;-)

            MT: In another message, Rick could

not see the analogy with the gedanken experiment’s
pebbles and Puck. Think of a vote as a pebble, the
possible identity of the President as a reference
location on a table that is domed rather than flat, so
that the stronger “stream of pebbles” moves the “Puck”
off the flattish top onto a slope down which the Puck
slides to one reference location or the other (or to one
of many in an election in which many candidates
participate).

          RM: Yes, this model would work (the domed table is a

feedback function, of course; see how much better that
sounds than atenfel).

          But then it's not the same as your pebble situation,

where there was no table. It also differs from Kent’s
model, which achieves control by having a collection of
controllers pushing the same variable toward different
references with approximately equal gain, the result being
that the variable is maintained in a virtual reference
state.

          RM: In collective control of who is President and how

she is elected the controlled variable ends up, not in a
virtual reference state but in a reference state that
corresponds to the reference specifications of some subset
of the individuals in the collective.

[From Rick Marken (2017.01.26.1500)]

···

Martin Taylor (2017.01.25.14.13)–

MT: CEV - First used on CSG-L in June 1992 (so far as I can find, in a

context that suggests it was not then a new term). And I doubt you
would be calling a “collectively controlled stability” newfangled if
you had read the material Kent keeps offering you.

RM: Well, it’s new relative to the development of and my involvement with PCT. The vocabulary of PCT was pretty well developed by 1973, when B:CP came out; and I got into PCT in 1978. So CEV was “new” to both Bill and myself. Â

RM: What puzzled me about the term “controlled stability” was that it seemed redundant since good control results in stability of the variable controlled. But I realized that the term “controlled stability” implies that there are “uncontrolled stabilities” and that these are all around us. Everything around us that is (for the time being, at least) not moving is an uncontrolled stability: the walls, floor and roof of my house, the table top, etc. These are things that remain stable without the assistance of a control system. Still, it strikes me as odd to talk about “collectively controlled stabilities” since I can’t think of any “collectively uncontrolled stabilities”. Perhaps they are just the same as “uncontrolled stabilities” but just with a lot of people around not controlling them.

Â

MT: In the case of the election method, if it

were not collectively controlled it would not be stable, and if it
were not stable citizens would have no means of registering their
preferences, thereby collectively controlling the identity of the
winning candidate.

RM: I agree that the election method is a collectively controlled variable (though it would be nice if you could explain why you think it is). What I have been questioning is whether Kent’s collective control model can explain how this, or any other collectively controlled variable, is controlled.Â

MT: You are not talking about Kent's model, but about Kent's

demonstration 23 years ago. There’s a difference. I believe I listed
six kinds of collective control a few months ago, of which Kent’s
demo is a trivial example of one type.

RM: That was very generous of you. But I will settle with a demonstration of how Kent’s model (the one I’m apparently not familiar with) can account for even one of the collectively controlled stabilities that you mentioned. For example, how does Kent’s model explain how the US ended up with an electoral college rather than a popular vote total or parliamentary method for electing a president.Â

MT: I suppose it's a bit too much to ask what aspect of collective

control you are trying to contradict with this statement?

RM: I am pointing out an apparent “contradiction” between a phenomenon – the collectively controlled variable of who is president – and the explanation of the phenomenon – Kent’s model of  collective control as I understand it. Perhaps I don’t understand the model. But it seems to me that Kent’s model would say that the identity of the president should end up in a virtual reference state somewhere between the references of the people who want the identity of the president to be one person and the references of the other people who it to be a different person.Â

RM: If that’s not how Kent’s model works I’d be interested in learning how it actually does work.Â

BestÂ

Rick

I'll stop here, because I sense too great a risk of a pissing match.



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

          RM: I agree that both the identity of the President and

the means by which the President is selected are
controlled variables, the state of which depends on the
action of many people; so they are collectively controlled
variables. I’ve never heard of a collectively controlled
stability, however. Maybe it’s one of those newfangled
concepts, like CEV;-)

          RM: Yes, this model would work (the domed table is a

feedback function, of course; see how much better that
sounds than atenfel).

          RM: But then it's not the same as your pebble situation,

where there was no table. It also differs from Kent’s
model, which achieves control by having a collection of
controllers pushing the same variable toward different
references with approximately equal gain, the result being
that the variable is maintained in a virtual reference
state.

          RM: In collective control of who is President and how

she is elected the controlled variable ends up, not in a
virtual reference state but in a reference state that
corresponds to the reference specifications of some subset
of the individuals in the collective.

[Martin Taylor 2016.12.13.12.09]

(Note: "Stochastic collective control" is a name I use for the kind

of collective control described throughout this message. I first
started to use it in private interchanges with Kent McClelland a
couple of months ago. If anyone knows of a place where the term has
been publicly used, I’d appreciate being made aware of it.)

-------A game-------

Imagine the following game. A number of players stand around a very

large irregularly shaped table that has a slippery surface. On the
table is an object like a big ice-hockey puck, so let’s call it “The
Puck”. Each player has access to a stock pile of objects like small
flat pebbles that they can slide across the table to hit the Puck
and nudge it a little (a “throw”). A player can get only one
“pebble” from a stockpile at a time, so it takes a little time after
each throw before they can get another pebble to throw. There is
some device that removes a thrown pebble from the table and returns
it to the stockpile after it has hit or passed by the Puck, so the
table is always clean.

Phase 1 of the gedanken experiment -- stochastic collective control

with a common reference value

All the players belong to a team. On the table is marked a target

location, and the team’s objective is to get the Puck onto the
target as quickly as possible. The figure shows the effect of one
“throw” by the player highlighted.

![Re TCV and Collective Control2.jpg|379x219](upload://r8ckCfJYAJziPOM2dyZwEmNG4xD.jpeg)

The team collectively controls the location of the Puck, in a form

of collective control I call “stochastic collective control”.
Clearly, if they are to get the Puck to the target as fast as
possible, the players for whom the target-puck-player angle is less
than 90 degrees should not throw their pebbles. For the purposes of
the game, we assume that a player is quite likely to hit the Puck,
but cannot hit it on a particular part of its circumference. This
implies that individual players cannot direct the Puck accurately.
Instead, when hit, the Puck moves with a component away from the
thrower plus a random sideways component that averages out to zero
in the long term. The above figure shows a hit that result in a
small leftward component, seen from the player’s viewpoint.

Consider the combined effect of three pebbles, one thrown by each of

the three players on the right side of the table in the Figure.
Roughly speaking, the most probable direction of the Puck if all
three threw more or less together is the average of their three
directions, but it could be anywhere in a wide arc. If there were
more players, the distribution of directions would be more peaked.
The total “force” (hits per second) in the most probably direction
is a bit less than the sum of their individual hit rates (actually
the sum of their hit rates times minus the cosine of the angle
subtended by the person, the Puck, and the average direction). The
Figure suggests the probability distributions of the effects of
their single pebbles and of the combination.

![Re TCV and Collective Control3.jpg|525x240](upload://wL8v2C0m8OPZIz4Qpojt3fo98vS.jpeg)
Every time a pebble is thrown, the Puck moves a little in a new

direction, but on average it will move toward the target if the
correct team members do the throwing. (This may look a bit like an
e-coli track, but it isn’t, because e-coli keeps going in the same
direction so long as it is getting closer to the target, which is
not the case for the Puck.)

![Re TCV and Collective Control4.jpg|379x219](upload://lRybjTUcEm1QBiQ2nJzrlDRkMRm.jpeg)

If there were a lot of throwers packed tightly around the table and

throwing rapidly, the Puck’s track would be much more direct,
deviating only slightly from a direct path toward the target. The
effect, to an observer who could not see the individual pebbles,
would look as though the Puck was being slid smoothly over the table
by an invisible hand, in the same way that we see movement in a
movie as smooth though we know it consists of discrete images spaced
in time. So, whereas individual players cannot direct the Puck
accurately with any one throw, the collective can, if there are
enough team members.

When there are a lot of players, each hit has a proportionately

smaller effect on the Puck. The game would be pointless if one hit
moved the Puck a long way across the table, because if it did, then
the multiple hits by a lot of players might send it right off the
table beyond the target after just one throw each. So we must
imagine two things, (a) that one hit moves the Puck only slightly,
and (b) a player is less likely to throw if the Puck is near the
target, so that the team avoids the Puck location overshooting and
oscillating erratically back and forth around the target. The net
effect is that the nearer the Puck is to the target, the smaller the
apparent force pushing it there, just as is the case in a canonical
control system, in which the contribution to the output is
proportional to the error value.

Phase 2 of the gedanken experiment -- variation among individual

reference values

In Phase 2 of the experiment, we remove the marking for the target.

Instead, at the start of a trial, the game-manager points to a spot
and tells the players that where he points is the target location.
The players therefore all agree in general where the target is, but
unknown to each of them, their reference locations all differ
slightly. So long as the Puck is far from the target, the situation
is the same as in Phase 1, but when the Puck is in the general area
of the manager’s target location, some players will think it is to
one side of the target while others think it is to the other.
Pebbles will hit the Puck from all directions, but more frequently
from the direction away from the average of the individual players’
reference locations.
Re TCV and Collective Control5.jpg

The Puck will move on average toward the average of the individual

reference locations, but will continue to jitter around that point
even when it is exactly there. (Bill Powers noted that engineered
control systems often include a zero-point jitter to avoid static
friction when stopped).

The team controls the Puck location as before, but with a virtual

reference value that might not be the reference value for any member
of the team. We might call the virtual reference value a Platonic
Ideal for the target location.

Phase 3 of the gedanken experiment -- teams in conflict

This time, the game manager tells some of the team one target

location and others a different target location. The individual
players do not know each other’s instructions. In the figure, group
1 is marked by grey dots around the table. Their reference values
for the target location are the right-hand set of reference
locations, filled grey. The black dot players with black-white
target reference values represent group 2.

![Re TCV and Collective Control6.jpg|379x219](upload://nmbBuVdkESYCU1QaJ4hGmbvPd0q.jpeg)

The puck is being hit with pebbles from players who could move it

toward their particular target reference location, which means that
in the diagrammed position most of the hits from both groups will
move the pebble “northward”, but in the east-west direction it will
be hit from the right by pebbles from the “black” players and from
the left by “grey” players. It will tend to move in whichever
direction it receives most hits. If the Puck nears one group target
area, their contribution to the apparent force on the Puck will
diminish and the contribution of the other group will increase, so
that the Puck will jitter around some point between the average
targets of the two groups. The situation is exactly the same as in
Phase 2. Only the game manager (and us) know that there are two
separate clusters of target reference locations.

The distinction between Phase 3 and Phase 2 is only in the

distribution of the individual reference locations for the target.
Instead of all being randomly clustered around one Platonic ideal
virtual reference level, they cluster around two, but the result
that the Puck will tend toward a location that is the average of all
the individual reference locations remains unchanged. The effect on
the Puck is still as if a hand were controlling a perception of the
Puck location with a reference location around the average of all
the individual reference locations.

Phase 4 of the gedanken experiment -- Testing for the Controlled

Variable.

In Phase 1, a team uses stochastic collective control to move the

Puck onto the target, which is the reference value of their
perceived location of the Puck for all of them. In Phase 2, every
player had a different reference location for the perceived target,
but all the reference locations were clustered around an “ideal”
target. By Phase 3, the individuals not only had different reference
locations for the Puck, but their reference locations were scattered
in two different clusters. Nevertheless, the location of the Puck in
all the phases appeared to be controlled to a single reference
location, even though in Phase 3 that location was far from the
reference location of any of the players.

In Phase 4 the situation for the players is exactly as in Phase 3,

but we now introduce a privileged player, who we can call the
“Disturber” or “Dee”. Dee can see the players and their pebbles, and
can move the Puck, but the players cannot see Dee. Dee wants to
determine whether the Puck location is being controlled, and if so,
to what reference value. So Dee moves the Puck in random directions
and watches what happens. When she moves the Puck, the thrown
pebbles will move it back toward the virtual reference location. For
every player, the Puck is not where the player wants it to be, and
if the individual’s target is beyond the Puck, the individual will
throw.

From Dee's point of view, the fact that the Puck keeps moving back

to the same place no matter where Dee moves it is not enough to
establish that its location is being controlled. For all she knows,
the table might not be flat, and the Puck might just be sliding back
down to the low point. The pebbles might be having no effect on it
at all. So Dee needs to do more. She must either prevent the players
from seeing the Puck or prevent their pebbles from hitting it. Let
us say she hides away the pebble stock pile, so the players have no
pebbles to throw. Now when she moves the Puck, it stays where she
puts it. So she knows that the Puck does not simply slide back down
a slope after she moves it, and she replaces the pebble stockpile.
The pebbles are essential if it is to move back. Next she tries
putting a dry-ice fog over the table so that the players cannot see
the Puck. Once again, after she moves it, it stays put, jiggin a
little when it is hit by an occasional pebble from a random
direction, but not moving toward its former location.

Dee has completed the TCV, and her results lead her to theorize that

the Puck is being controlled by an entity (which Dee calls “Con”)
for whom the players are simply agents. The players, in her theory,
continually observe the Puck location, report their observations to
Con, and throw pebbles when Con instructs them to do so, in order to
keep the Puck near Con’s reference location for it. According to
Dee, Con controls his perception of the Puck’s location, using the
players as the equivalents of his eyes and muscles.

------Commentary---------

Dee has performed the Test for the Controlled variable, and has come

to the conclusion that an entity she calls “Con” exists and has an
internal perception of the Puck location that he controls by having
pebbles shot at the puck from different directions. But we
know that Con does not exist and does not need to exist for Dee to
observe what she observed. Only the players exist, controlling their
individual perceptions of the Puck location according to their
individual reference locations for it.

How do *we* know that within any one player there exists a

perception of the Puck location? We don’t. We are in the same
predicament in that respect as Dee is with respect to Con’s
perception, which, as we know, doesn’t exist because Con does not
exist.

At this point the quote from Bill Powers that Rick found [Bill

Powers (961224.1145 MST)] becomes relevant: “Remember that as far as
the observer is concerned, what is controlled is ONLY the CV. The
idea that this CV is represented by a perceptual signal inside the
other system is theoretical. We can observe CV, but not p. When we
apply a disturbance, we apply it to CV, not to p. The action that
opposes the effect of the disturbance acts on CV, not p. The Test
does not involve p at all. It involves only observables – i.e., the
observer’s perceptions. The observations have priority; the model
comes second, and its only reason for existence is to explain the
observations.”

(Bill follows with this: "When you fool around with

thought-experiments too much, you tend to get the priorities
reversed." I hope I have not fallen into that trap with this thought
experiment).

------Using the game analogy------

Now think about the concept of a "Neural Current", which is a

fictitious quantity that is the sum of many one-shot events – nerve
firings, not pebble throws. The idea of there being a “perceptual
signal” depends on the notion of the neural current. Indeed, a
perceptual signal is the value of the neural current computed by
summing the effects of the firings of a bundle of neurons that
respond similarly to their related inputs. It is closely analogous
to the force on the Puck of a barrage of pebbles thrown by different
players.

Consider the muscular output part of a control loop. The tension in

a muscle is related to the rate of nerve firings, and the angle of
many joints is set by the balance between the firings on opposed
muscles. To me, phase 3 of this gedanken experiment looks like that,
with one group of pebble throwers trying to move the puck to the
average of one cluster of reference locations while the other group
tries to move it to the other cluster, the result being an
apparently controlled location of the Puck (the joint angle) between
the two clusters. If there had been a “Con” directing the players,
Con could have moved the Puck in a controlled manner by telling one
group to go easy and the other to throw more rapidly. Likewise we
move our joints in a controlled manner by changing the balance
between nerve firings in the opposed muscles according to reference
values sent down from higher levels.

In our thread(s) on the reality of what we perceive, the assumption

throughout has been that the perceptual signal is real and the
reality of what we think we perceive is questionable. Perhaps we
should ask the question in the other direction. Assume that there is
something real, though we know not what it is, and that some
property of that something influences the firings by the neurons in
our perceptual apparatus (and is influenced by things we do with our
muscles). The real thing is unitary but unknown, but the perception
is virtual and known (at least some perceptions are known to
consciousness). Perceptual functions are presumably implemented by
the synapses and hormonal environment of every neuron in the
perceptual apparatus, and the unitary real world affects the inputs
to them. Those inputs and their structure and local chemical
environment determine their firing rates.

So we come to the strange position that our perceptions are the only

things that we can be sure of, even though they are virtual,
existing nowhere in the brain as a locatable value. We survive
because we can use our actions to control those virtual perceptions,
just as all our ancestors have done back to the first stirrings of
life on the planet. They work in real reality, regardless of theory.

Stochastic collective control in our bodies allows us to live, and

stochastic collective control in our environment allows us to create
artifactual entities such as ways of behaving with respect to each
other that are stable, reliable, ways of getting what we want –
atenfels that are often intangible. We live because things “out
there” usually change as they would if our perceptions were
veridical when we control them, even though the perceptions we
control are themselves only virtual.

Martin

Content-Type: image/jpeg
Content-Disposition: attachment
Content-ID: <part1.1E59FB9E.ABEFFE01@mmtaylor.net>
X-Attachment-Id: a08134c88ea114f4_0.1.1

Content-Type: image/jpeg
Content-Disposition: attachment
Content-ID: <part2.90A54FF3.C06A5E88@mmtaylor.net>
X-Attachment-Id: a08134c88ea114f4_0.1.2

Content-Type: image/jpeg
Content-Disposition: attachment
Content-ID: <part3.7804C50A.844B7E2C@mmtaylor.net>
X-Attachment-Id: a08134c88ea114f4_0.1.3

Content-Type: image/jpeg
Content-Disposition: attachment
Content-ID: <part4.6D9F3FE6.805631B7@mmtaylor.net>
X-Attachment-Id: a08134c88ea114f4_0.1.4

Content-Type: image/jpeg
Content-Disposition: attachment
Content-ID: <part5.666A089D.AB57DA78@mmtaylor.net>
X-Attachment-Id: a08134c88ea114f4_0.1.5

Re TCV and Collective Control2.jpg

Re TCV and Collective Control3.jpg

Re TCV and Collective Control4.jpg

Re TCV and Collective Control5.jpg

Re TCV and Collective Control6.jpg

[Martin Taylor 2016.12.23.00.54]

Yes it would. For reference, here's a description of the TCV. The

stages, not necessarily done in this order, are:

1.    checking to determine whether the hypothetical controller

could sense the supposed controlled environmental property,

2.    checking to determine whether the hypothetical controller

could act to influence its value,

3.    applying a disturbing influence to the hypothesized controlled

property, and

4.    applying one or more of the following tests:

   a.    whether the hypothesized property changes appreciably less

than would be expected if a corresponding perception were not being
controlled, or

   b.    measuring the influence of the putative controller and

determining that it is negatively correlated with the (time-lagged)
influence of the experimenter-induced disturbance, or

   c.    measuring the varying value of the complex property and

determining that its variation has a low correlation with variation
in the applied disturbance.

5.    If all the hypothesized properties are equally plausible

according to the first three conditions, the controlled perception
is provisionally taken to be the one that best satisfies any test of
step 4.

Which steps did Dee do improperly?

Martin
···

[From Rick Marken (2016.12.22.0920)]

            Martin Taylor

(2016.12.13.12.09)–

            MT: From Dee's point of view, the fact that the Puck

keeps moving back to the same place no matter where Dee
moves it is not enough to establish that its location is
being controlled. For all she knows, the table might not
be flat, and the Puck might just be sliding back down to
the low point. The pebbles might be having no effect on
it at all. So Dee needs to do more. She must either
prevent the players from seeing the Puck or prevent
their pebbles from hitting it. Let us say she hides away
the pebble stock pile, so the players have no pebbles to
throw. Now when she moves the Puck, it stays where she
puts it. So she knows that the Puck does not simply
slide back down a slope after she moves it, and she
replaces the pebble stockpile. The pebbles are essential
if it is to move back. Next she tries putting a dry-ice
fog over the table so that the players cannot see the
Puck. Once again, after she moves it, it stays put,
jiggin a little when it is hit by an occasional pebble
from a random direction, but not moving toward its
former location.

            MT: Dee has completed the TCV, and her results lead her

to theorize that the Puck is being controlled by an
entity (which Dee calls “Con”) for whom the players are
simply agents. The players, in her theory, continually
observe the Puck location, report their observations to
Con, and throw pebbles when Con instructs them to do so,
in order to keep the Puck near Con’s reference location
for it. According to Dee, Con controls his perception of
the Puck’s location, using the players as the
equivalents of his eyes and muscles.

------Commentary---------

            MT: Dee has performed the Test for the Controlled

variable, and has come to the conclusion that an entity
she calls “Con” exists and has an internal perception of
the Puck location that he controls by having pebbles
shot at the puck from different directions. But we
know that Con does not exist and does not need to exist
for Dee to observe what she observed. Only the players
exist, controlling their individual perceptions of the
Puck location according to their individual reference
locations for it.

          RM: It would sure be nice if someone other than me

would explain what is wrong with this description of the
Test for the Controlled Variable.

[From Rick Marken (2016.12.23.0945)]

···

Martin Taylor (2016.12.23.00.54)

MT: Yes it would. For reference, here's a description of the TCV. The

stages, not necessarily done in this order, are:

1.    checking to determine whether the hypothetical controller

could sense the supposed controlled environmental property,

2.    checking to determine whether the hypothetical controller

could act to influence its value,

3.    applying a disturbing influence to the hypothesized controlled

property, and

4.    applying one or more of the following tests:

   a.    whether the hypothesized property changes appreciably less

than would be expected if a corresponding perception were not being
controlled, or

   b.    measuring the influence of the putative controller and

determining that it is negatively correlated with the (time-lagged)
influence of the experimenter-induced disturbance, or

   c.    measuring the varying value of the complex property and

determining that its variation has a low correlation with variation
in the applied disturbance.

5.    If all the hypothesized properties are equally plausible

according to the first three conditions, the controlled perception
is provisionally taken to be the one that best satisfies any test of
step 4.

Which steps did Dee do improperly?

RM: I’d rather that you or someone other than me answer this. If someone hasn’t done this by tomorrow night I’ll put it as a present under the PCT Xmas tree.

Best

Rick

Martin


Richard S. Marken

“The childhood of the human race is far from over. We
have a long way to go before most people will understand that what they do for
others is just as important to their well-being as what they do for
themselves.” – William T. Powers

            MT: From Dee's point of view, the fact that the Puck

keeps moving back to the same place no matter where Dee
moves it is not enough to establish that its location is
being controlled…

          RM: It would sure be nice if someone other than me

would explain what is wrong with this description of the
Test for the Controlled Variable.

[Martin Taylor 2016.12.23.13.57]

[From Rick Marken (2016.12.23.0945)]

[MT] I have two guesses, but maybe someone else will have a

different guess and both will differ from what Rick has in mind, so
we will have a Christmas Tree full of errors.

Here's my main guess. Dee seems not to be searching for the

“controlled variable”, but is instead using the normal TCV to
determine whether a specific variable is (hypothesis 1) controlled
or (hypothesis 2) uncontrolled, and if controlled, whether the
pebble throwers are doing the controlling. The common use of the TCV
is to assert that some unknown variable is controlled, and then
determine (a) which of several hypothesized variables is the
“unknown”, and who is doing the controlling – “who” meaning what
output actions and what perceptual inputs are functional in the
control.

In the experiment the only variable Dee is interested in is the Puck

location, and I believe she correctly uses the TCV to determine that
it is indeed controlled. My guess is that Rick thinks the TCV cannot
be used this way, but is permitted only to distinguish among two or
more hypothesized variables when control is already known to be
happening.

Here's another out-of-left-field guess: Puck location isn't one

variable but two, and the TCV is supposed to deal with only one at a
time. But I think the rebuttal is so obvious that I doubt this is
what concerns Rick.

Any more guesses as to the nature of Rick's objection?

Martin
···

Martin Taylor (2016.12.23.00.54)

            MT: Yes it would. For reference, here's a

description of the TCV. The stages, not necessarily done
in this order, are:

            1.    checking to determine whether the hypothetical

controller could sense the supposed controlled
environmental property,

            2.    checking to determine whether the hypothetical

controller could act to influence its value,

            3.    applying a disturbing influence to the

hypothesized controlled property, and

            4.    applying one or more of the following tests:

               a.    whether the hypothesized property changes

appreciably less than would be expected if a
corresponding perception were not being controlled, or

               b.    measuring the influence of the putative

controller and determining that it is negatively
correlated with the (time-lagged) influence of the
experimenter-induced disturbance, or

               c.    measuring the varying value of the complex

property and determining that its variation has a low
correlation with variation in the applied disturbance.

            5.    If all the hypothesized properties are equally

plausible according to the first three conditions, the
controlled perception is provisionally taken to be the
one that best satisfies any test of step 4.

            Which steps did Dee do improperly?
          RM: I'd rather that you or someone other than me answer

this. If someone hasn’t done this by tomorrow night I’ll
put it as a present under the PCT Xmas tree.

                          MT:

From Dee’s point of view, the fact that
the Puck keeps moving back to the same
place no matter where Dee moves it is not
enough to establish that its location is
being controlled…

                        RM: It would sure be nice if someone

other than me would explain what is wrong
with this description of the Test for the
Controlled Variable.

[Bruce Nevin (2016.12.25.21.48 ET]

A pissing contest about correct recitation of TCV scripture interests me far less than the analogy of ‘stochastic collective control’ (in the ‘game’ “the force on the Puck of a barrage of pebbles thrown by different players”) to a perceptual signal (“the value of the neural current computed by summing the effects of the firings of a bundle of neurons that respond similarly to their related inputs”).

It seems to me that this addresses the important question, how an input function is created and maintained.

A control loop does whatever works to control its perceptual input in accord with its reference input. In addition to the immediate effect of varying its output (the standard view), “whatever works” may include the effect of the slower process of modifying one or both of those inputs.

Generally, we say that an input function is created and maintained by ‘the reorganization system’. I’m wondering if a model of ‘stochastic collective control’ might be a model of how the nerve fibers arriving at a comparator do that.

Re TCV and Collective Control4.jpg

Re TCV and Collective Control2.jpg

Re TCV and Collective Control6.jpg

Re TCV and Collective Control5.jpg

Re TCV and Collective Control3.jpg

···

On Mon, Dec 19, 2016 at 11:36 PM, Martin Taylor mmt-csg@mmtaylor.net wrote:

[Martin Taylor 2016.12.13.12.09]

(Note: "Stochastic collective control" is a name I use for the kind

of collective control described throughout this message. I first
started to use it in private interchanges with Kent McClelland a
couple of months ago. If anyone knows of a place where the term has
been publicly used, I’d appreciate being made aware of it.)

-------A game-------

Imagine the following game. A number of players stand around a very

large irregularly shaped table that has a slippery surface. On the
table is an object like a big ice-hockey puck, so let’s call it “The
Puck”. Each player has access to a stock pile of objects like small
flat pebbles that they can slide across the table to hit the Puck
and nudge it a little (a “throw”). A player can get only one
“pebble” from a stockpile at a time, so it takes a little time after
each throw before they can get another pebble to throw. There is
some device that removes a thrown pebble from the table and returns
it to the stockpile after it has hit or passed by the Puck, so the
table is always clean.

Phase 1 of the gedanken experiment -- stochastic collective control

with a common reference value

All the players belong to a team. On the table is marked a target

location, and the team’s objective is to get the Puck onto the
target as quickly as possible. The figure shows the effect of one
“throw” by the player highlighted.

The team collectively controls the location of the Puck, in a form

of collective control I call “stochastic collective control”.
Clearly, if they are to get the Puck to the target as fast as
possible, the players for whom the target-puck-player angle is less
than 90 degrees should not throw their pebbles. For the purposes of
the game, we assume that a player is quite likely to hit the Puck,
but cannot hit it on a particular part of its circumference. This
implies that individual players cannot direct the Puck accurately.
Instead, when hit, the Puck moves with a component away from the
thrower plus a random sideways component that averages out to zero
in the long term. The above figure shows a hit that result in a
small leftward component, seen from the player’s viewpoint.

Consider the combined effect of three pebbles, one thrown by each of

the three players on the right side of the table in the Figure.
Roughly speaking, the most probable direction of the Puck if all
three threw more or less together is the average of their three
directions, but it could be anywhere in a wide arc. If there were
more players, the distribution of directions would be more peaked.
The total “force” (hits per second) in the most probably direction
is a bit less than the sum of their individual hit rates (actually
the sum of their hit rates times minus the cosine of the angle
subtended by the person, the Puck, and the average direction). The
Figure suggests the probability distributions of the effects of
their single pebbles and of the combination.

Every time a pebble is thrown, the Puck moves a little in a new

direction, but on average it will move toward the target if the
correct team members do the throwing. (This may look a bit like an
e-coli track, but it isn’t, because e-coli keeps going in the same
direction so long as it is getting closer to the target, which is
not the case for the Puck.)

If there were a lot of throwers packed tightly around the table and

throwing rapidly, the Puck’s track would be much more direct,
deviating only slightly from a direct path toward the target. The
effect, to an observer who could not see the individual pebbles,
would look as though the Puck was being slid smoothly over the table
by an invisible hand, in the same way that we see movement in a
movie as smooth though we know it consists of discrete images spaced
in time. So, whereas individual players cannot direct the Puck
accurately with any one throw, the collective can, if there are
enough team members.

When there are a lot of players, each hit has a proportionately

smaller effect on the Puck. The game would be pointless if one hit
moved the Puck a long way across the table, because if it did, then
the multiple hits by a lot of players might send it right off the
table beyond the target after just one throw each. So we must
imagine two things, (a) that one hit moves the Puck only slightly,
and (b) a player is less likely to throw if the Puck is near the
target, so that the team avoids the Puck location overshooting and
oscillating erratically back and forth around the target. The net
effect is that the nearer the Puck is to the target, the smaller the
apparent force pushing it there, just as is the case in a canonical
control system, in which the contribution to the output is
proportional to the error value.

Phase 2 of the gedanken experiment -- variation among individual

reference values

In Phase 2 of the experiment, we remove the marking for the target.

Instead, at the start of a trial, the game-manager points to a spot
and tells the players that where he points is the target location.
The players therefore all agree in general where the target is, but
unknown to each of them, their reference locations all differ
slightly. So long as the Puck is far from the target, the situation
is the same as in Phase 1, but when the Puck is in the general area
of the manager’s target location, some players will think it is to
one side of the target while others think it is to the other.
Pebbles will hit the Puck from all directions, but more frequently
from the direction away from the average of the individual players’
reference locations.

The Puck will move on average toward the average of the individual

reference locations, but will continue to jitter around that point
even when it is exactly there. (Bill Powers noted that engineered
control systems often include a zero-point jitter to avoid static
friction when stopped).

The team controls the Puck location as before, but with a virtual

reference value that might not be the reference value for any member
of the team. We might call the virtual reference value a Platonic
Ideal for the target location.

Phase 3 of the gedanken experiment -- teams in conflict

This time, the game manager tells some of the team one target

location and others a different target location. The individual
players do not know each other’s instructions. In the figure, group
1 is marked by grey dots around the table. Their reference values
for the target location are the right-hand set of reference
locations, filled grey. The black dot players with black-white
target reference values represent group 2.

The puck is being hit with pebbles from players who could move it

toward their particular target reference location, which means that
in the diagrammed position most of the hits from both groups will
move the pebble “northward”, but in the east-west direction it will
be hit from the right by pebbles from the “black” players and from
the left by “grey” players. It will tend to move in whichever
direction it receives most hits. If the Puck nears one group target
area, their contribution to the apparent force on the Puck will
diminish and the contribution of the other group will increase, so
that the Puck will jitter around some point between the average
targets of the two groups. The situation is exactly the same as in
Phase 2. Only the game manager (and us) know that there are two
separate clusters of target reference locations.

The distinction between Phase 3 and Phase 2 is only in the

distribution of the individual reference locations for the target.
Instead of all being randomly clustered around one Platonic ideal
virtual reference level, they cluster around two, but the result
that the Puck will tend toward a location that is the average of all
the individual reference locations remains unchanged. The effect on
the Puck is still as if a hand were controlling a perception of the
Puck location with a reference location around the average of all
the individual reference locations.

Phase 4 of the gedanken experiment -- Testing for the Controlled

Variable.

In Phase 1, a team uses stochastic collective control to move the

Puck onto the target, which is the reference value of their
perceived location of the Puck for all of them. In Phase 2, every
player had a different reference location for the perceived target,
but all the reference locations were clustered around an “ideal”
target. By Phase 3, the individuals not only had different reference
locations for the Puck, but their reference locations were scattered
in two different clusters. Nevertheless, the location of the Puck in
all the phases appeared to be controlled to a single reference
location, even though in Phase 3 that location was far from the
reference location of any of the players.

In Phase 4 the situation for the players is exactly as in Phase 3,

but we now introduce a privileged player, who we can call the
“Disturber” or “Dee”. Dee can see the players and their pebbles, and
can move the Puck, but the players cannot see Dee. Dee wants to
determine whether the Puck location is being controlled, and if so,
to what reference value. So Dee moves the Puck in random directions
and watches what happens. When she moves the Puck, the thrown
pebbles will move it back toward the virtual reference location. For
every player, the Puck is not where the player wants it to be, and
if the individual’s target is beyond the Puck, the individual will
throw.

From Dee's point of view, the fact that the Puck keeps moving back

to the same place no matter where Dee moves it is not enough to
establish that its location is being controlled. For all she knows,
the table might not be flat, and the Puck might just be sliding back
down to the low point. The pebbles might be having no effect on it
at all. So Dee needs to do more. She must either prevent the players
from seeing the Puck or prevent their pebbles from hitting it. Let
us say she hides away the pebble stock pile, so the players have no
pebbles to throw. Now when she moves the Puck, it stays where she
puts it. So she knows that the Puck does not simply slide back down
a slope after she moves it, and she replaces the pebble stockpile.
The pebbles are essential if it is to move back. Next she tries
putting a dry-ice fog over the table so that the players cannot see
the Puck. Once again, after she moves it, it stays put, jiggin a
little when it is hit by an occasional pebble from a random
direction, but not moving toward its former location.

Dee has completed the TCV, and her results lead her to theorize that

the Puck is being controlled by an entity (which Dee calls “Con”)
for whom the players are simply agents. The players, in her theory,
continually observe the Puck location, report their observations to
Con, and throw pebbles when Con instructs them to do so, in order to
keep the Puck near Con’s reference location for it. According to
Dee, Con controls his perception of the Puck’s location, using the
players as the equivalents of his eyes and muscles.

------Commentary---------

Dee has performed the Test for the Controlled variable, and has come

to the conclusion that an entity she calls “Con” exists and has an
internal perception of the Puck location that he controls by having
pebbles shot at the puck from different directions. But we
know that Con does not exist and does not need to exist for Dee to
observe what she observed. Only the players exist, controlling their
individual perceptions of the Puck location according to their
individual reference locations for it.

How do *we* know that within any one player there exists a

perception of the Puck location? We don’t. We are in the same
predicament in that respect as Dee is with respect to Con’s
perception, which, as we know, doesn’t exist because Con does not
exist.

At this point the quote from Bill Powers that Rick found [Bill

Powers (961224.1145 MST)] becomes relevant: “Remember that as far as
the observer is concerned, what is controlled is ONLY the CV. The
idea that this CV is represented by a perceptual signal inside the
other system is theoretical. We can observe CV, but not p. When we
apply a disturbance, we apply it to CV, not to p. The action that
opposes the effect of the disturbance acts on CV, not p. The Test
does not involve p at all. It involves only observables – i.e., the
observer’s perceptions. The observations have priority; the model
comes second, and its only reason for existence is to explain the
observations.”

(Bill follows with this: "When you fool around with

thought-experiments too much, you tend to get the priorities
reversed." I hope I have not fallen into that trap with this thought
experiment).

------Using the game analogy------



Now think about the concept of a "Neural Current", which is a

fictitious quantity that is the sum of many one-shot events – nerve
firings, not pebble throws. The idea of there being a “perceptual
signal” depends on the notion of the neural current. Indeed, a
perceptual signal is the value of the neural current computed by
summing the effects of the firings of a bundle of neurons that
respond similarly to their related inputs. It is closely analogous
to the force on the Puck of a barrage of pebbles thrown by different
players.

Consider the muscular output part of a control loop. The tension in

a muscle is related to the rate of nerve firings, and the angle of
many joints is set by the balance between the firings on opposed
muscles. To me, phase 3 of this gedanken experiment looks like that,
with one group of pebble throwers trying to move the puck to the
average of one cluster of reference locations while the other group
tries to move it to the other cluster, the result being an
apparently controlled location of the Puck (the joint angle) between
the two clusters. If there had been a “Con” directing the players,
Con could have moved the Puck in a controlled manner by telling one
group to go easy and the other to throw more rapidly. Likewise we
move our joints in a controlled manner by changing the balance
between nerve firings in the opposed muscles according to reference
values sent down from higher levels.

In our thread(s) on the reality of what we perceive, the assumption

throughout has been that the perceptual signal is real and the
reality of what we think we perceive is questionable. Perhaps we
should ask the question in the other direction. Assume that there is
something real, though we know not what it is, and that some
property of that something influences the firings by the neurons in
our perceptual apparatus (and is influenced by things we do with our
muscles). The real thing is unitary but unknown, but the perception
is virtual and known (at least some perceptions are known to
consciousness). Perceptual functions are presumably implemented by
the synapses and hormonal environment of every neuron in the
perceptual apparatus, and the unitary real world affects the inputs
to them. Those inputs and their structure and local chemical
environment determine their firing rates.

So we come to the strange position that our perceptions are the only

things that we can be sure of, even though they are virtual,
existing nowhere in the brain as a locatable value. We survive
because we can use our actions to control those virtual perceptions,
just as all our ancestors have done back to the first stirrings of
life on the planet. They work in real reality, regardless of theory.

Stochastic collective control in our bodies allows us to live, and

stochastic collective control in our environment allows us to create
artifactual entities such as ways of behaving with respect to each
other that are stable, reliable, ways of getting what we want –
atenfels that are often intangible. We live because things “out
there” usually change as they would if our perceptions were
veridical when we control them, even though the perceptions we
control are themselves only virtual.

Martin

[Martin Taylor 2016.12.25.23.30]

Bruce, an especially happy Christmas to you, even if the wishes are

somewhat belated!

Your note is one of my better Xmas presents. Someone who gets (at

least part of) the point. The circuit I called “Friston” in [Martin
Taylor 2016.10.16.14.33], together with (I think it was your)
suggestion about modulating the relative amount of imagination input
to the perceptual input rather than switching between imagination
and sensory imput is another part of the point.

The central point, however, is that even though we cannot know "real

reality" (or as Powers called it “boss reality”), nevertheless we
have to assume there is one, and only in that real reality external
to any hypothetical Elementary Control Unit do our control actions
do what they do to change our perceptions. That’s where the combined
effects of all these nerve firings become singular as opposed to
contributing to bundles averages called “neural currents” in the
brain. And as you say, the effectiveness of that control is at least
part of, if not all of, the construction of stable(ish) perceptual
functions (as J.G.Taylor said so long ago).

Martin

Re TCV and Collective Control2.jpg

Re TCV and Collective Control3.jpg

Re TCV and Collective Control4.jpg

Re TCV and Collective Control5.jpg

Re TCV and Collective Control6.jpg

···

On Mon, Dec 19, 2016 at 11:36 PM,
Martin Taylor mmt-csg@mmtaylor.net
wrote:

          [Martin Taylor

2016.12.13.12.09]

          (Note: "Stochastic collective control" is a name I use for

the kind of collective control described throughout this
message. I first started to use it in private interchanges
with Kent McClelland a couple of months ago. If anyone
knows of a place where the term has been publicly used,
I’d appreciate being made aware of it.)

          -------A game-------

          Imagine the following game. A number of players stand

around a very large irregularly shaped table that has a
slippery surface. On the table is an object like a big
ice-hockey puck, so let’s call it “The Puck”. Each player
has access to a stock pile of objects like small flat
pebbles that they can slide across the table to hit the
Puck and nudge it a little (a “throw”). A player can get
only one “pebble” from a stockpile at a time, so it takes
a little time after each throw before they can get another
pebble to throw. There is some device that removes a
thrown pebble from the table and returns it to the
stockpile after it has hit or passed by the Puck, so the
table is always clean.

          Phase 1 of the gedanken experiment -- stochastic

collective control with a common reference value

          All the players belong to a team. On the table is marked a

target location, and the team’s objective is to get the
Puck onto the target as quickly as possible. The figure
shows the effect of one “throw” by the player highlighted.

          The team collectively controls the location of the Puck,

in a form of collective control I call “stochastic
collective control”. Clearly, if they are to get the Puck
to the target as fast as possible, the players for whom
the target-puck-player angle is less than 90 degrees
should not throw their pebbles. For the purposes of the
game, we assume that a player is quite likely to hit the
Puck, but cannot hit it on a particular part of its
circumference. This implies that individual players cannot
direct the Puck accurately. Instead, when hit, the Puck
moves with a component away from the thrower plus a random
sideways component that averages out to zero in the long
term. The above figure shows a hit that result in a small
leftward component, seen from the player’s viewpoint.

          Consider the combined effect of three pebbles, one thrown

by each of the three players on the right side of the
table in the Figure. Roughly speaking, the most probable
direction of the Puck if all three threw more or less
together is the average of their three directions, but it
could be anywhere in a wide arc. If there were more
players, the distribution of directions would be more
peaked. The total “force” (hits per second) in the most
probably direction is a bit less than the sum of their
individual hit rates (actually the sum of their hit rates
times minus the cosine of the angle subtended by the
person, the Puck, and the average direction). The Figure
suggests the probability distributions of the effects of
their single pebbles and of the combination.

          Every time a pebble is thrown, the Puck moves a little in

a new direction, but on average it will move toward the
target if the correct team members do the throwing. (This
may look a bit like an e-coli track, but it isn’t, because
e-coli keeps going in the same direction so long as it is
getting closer to the target, which is not the case for
the Puck.)

          If there were a lot of throwers packed tightly around the

table and throwing rapidly, the Puck’s track would be much
more direct, deviating only slightly from a direct path
toward the target. The effect, to an observer who could
not see the individual pebbles, would look as though the
Puck was being slid smoothly over the table by an
invisible hand, in the same way that we see movement in a
movie as smooth though we know it consists of discrete
images spaced in time. So, whereas individual players
cannot direct the Puck accurately with any one throw, the
collective can, if there are enough team members.

          When there are a lot of players, each hit has a

proportionately smaller effect on the Puck. The game would
be pointless if one hit moved the Puck a long way across
the table, because if it did, then the multiple hits by a
lot of players might send it right off the table beyond
the target after just one throw each. So we must imagine
two things, (a) that one hit moves the Puck only slightly,
and (b) a player is less likely to throw if the Puck is
near the target, so that the team avoids the Puck location
overshooting and oscillating erratically back and forth
around the target. The net effect is that the nearer the
Puck is to the target, the smaller the apparent force
pushing it there, just as is the case in a canonical
control system, in which the contribution to the output is
proportional to the error value.

          Phase 2 of the gedanken experiment -- variation among

individual reference values

          In Phase 2 of the experiment, we remove the marking for

the target. Instead, at the start of a trial, the
game-manager points to a spot and tells the players that
where he points is the target location. The players
therefore all agree in general where the target is, but
unknown to each of them, their reference locations all
differ slightly. So long as the Puck is far from the
target, the situation is the same as in Phase 1, but when
the Puck is in the general area of the manager’s target
location, some players will think it is to one side of the
target while others think it is to the other. Pebbles will
hit the Puck from all directions, but more frequently from
the direction away from the average of the individual
players’ reference locations.

          The Puck will move on average toward the average of the

individual reference locations, but will continue to
jitter around that point even when it is exactly there.
(Bill Powers noted that engineered control systems often
include a zero-point jitter to avoid static friction when
stopped).

          The team controls the Puck location as before, but with a

virtual reference value that might not be the reference
value for any member of the team. We might call the
virtual reference value a Platonic Ideal for the target
location.

          Phase 3 of the gedanken experiment -- teams in conflict

          This time, the game manager tells some of the team one

target location and others a different target location.
The individual players do not know each other’s
instructions. In the figure, group 1 is marked by grey
dots around the table. Their reference values for the
target location are the right-hand set of reference
locations, filled grey. The black dot players with
black-white target reference values represent group 2.

          The puck is being hit with pebbles from players who could

move it toward their particular target reference location,
which means that in the diagrammed position most of the
hits from both groups will move the pebble “northward”,
but in the east-west direction it will be hit from the
right by pebbles from the “black” players and from the
left by “grey” players. It will tend to move in whichever
direction it receives most hits. If the Puck nears one
group target area, their contribution to the apparent
force on the Puck will diminish and the contribution of
the other group will increase, so that the Puck will
jitter around some point between the average targets of
the two groups. The situation is exactly the same as in
Phase 2. Only the game manager (and us) know that there
are two separate clusters of target reference locations.

          The distinction between Phase 3 and Phase 2 is only in the

distribution of the individual reference locations for the
target. Instead of all being randomly clustered around one
Platonic ideal virtual reference level, they cluster
around two, but the result that the Puck will tend toward
a location that is the average of all the individual
reference locations remains unchanged. The effect on the
Puck is still as if a hand were controlling a perception
of the Puck location with a reference location around the
average of all the individual reference locations.

          Phase 4 of the gedanken experiment -- Testing for the

Controlled Variable.

          In Phase 1, a team uses stochastic collective control to

move the Puck onto the target, which is the reference
value of their perceived location of the Puck for all of
them. In Phase 2, every player had a different reference
location for the perceived target, but all the reference
locations were clustered around an “ideal” target. By
Phase 3, the individuals not only had different reference
locations for the Puck, but their reference locations were
scattered in two different clusters. Nevertheless, the
location of the Puck in all the phases appeared to be
controlled to a single reference location, even though in
Phase 3 that location was far from the reference location
of any of the players.

          In Phase 4 the situation for the players is exactly as in

Phase 3, but we now introduce a privileged player, who we
can call the “Disturber” or “Dee”. Dee can see the players
and their pebbles, and can move the Puck, but the players
cannot see Dee. Dee wants to determine whether the Puck
location is being controlled, and if so, to what reference
value. So Dee moves the Puck in random directions and
watches what happens. When she moves the Puck, the thrown
pebbles will move it back toward the virtual reference
location. For every player, the Puck is not where the
player wants it to be, and if the individual’s target is
beyond the Puck, the individual will throw.

          From Dee's point of view, the fact that the Puck keeps

moving back to the same place no matter where Dee moves it
is not enough to establish that its location is being
controlled. For all she knows, the table might not be
flat, and the Puck might just be sliding back down to the
low point. The pebbles might be having no effect on it at
all. So Dee needs to do more. She must either prevent the
players from seeing the Puck or prevent their pebbles from
hitting it. Let us say she hides away the pebble stock
pile, so the players have no pebbles to throw. Now when
she moves the Puck, it stays where she puts it. So she
knows that the Puck does not simply slide back down a
slope after she moves it, and she replaces the pebble
stockpile. The pebbles are essential if it is to move
back. Next she tries putting a dry-ice fog over the table
so that the players cannot see the Puck. Once again, after
she moves it, it stays put, jiggin a little when it is hit
by an occasional pebble from a random direction, but not
moving toward its former location.

          Dee has completed the TCV, and her results lead her to

theorize that the Puck is being controlled by an entity
(which Dee calls “Con”) for whom the players are simply
agents. The players, in her theory, continually observe
the Puck location, report their observations to Con, and
throw pebbles when Con instructs them to do so, in order
to keep the Puck near Con’s reference location for it.
According to Dee, Con controls his perception of the
Puck’s location, using the players as the equivalents of
his eyes and muscles.

          ------Commentary---------

          Dee has performed the Test for the Controlled variable,

and has come to the conclusion that an entity she calls
“Con” exists and has an internal perception of the Puck
location that he controls by having pebbles shot at the
puck from different directions. But we know
that Con does not exist and does not need to exist for Dee
to observe what she observed. Only the players exist,
controlling their individual perceptions of the Puck
location according to their individual reference locations
for it.

          How do *we* know that within any one player there

exists a perception of the Puck location? We don’t. We are
in the same predicament in that respect as Dee is with
respect to Con’s perception, which, as we know, doesn’t
exist because Con does not exist.

          At this point the quote from Bill Powers that Rick found

[Bill Powers (961224.1145 MST)] becomes relevant:
“Remember that as far as the observer is concerned, what
is controlled is ONLY the CV. The idea that this CV is
represented by a perceptual signal inside the other system
is theoretical. We can observe CV, but not p. When we
apply a disturbance, we apply it to CV, not to p. The
action that opposes the effect of the disturbance acts on
CV, not p. The Test does not involve p at all. It involves
only observables – i.e., the observer’s perceptions. The
observations have priority; the model comes second, and
its only reason for existence is to explain the
observations.”

          (Bill follows with this: "When you fool around with

thought-experiments too much, you tend to get the
priorities reversed." I hope I have not fallen into that
trap with this thought experiment).

          ------Using the game analogy------



          Now think about the concept of a "Neural Current", which

is a fictitious quantity that is the sum of many one-shot
events – nerve firings, not pebble throws. The idea of
there being a “perceptual signal” depends on the notion of
the neural current. Indeed, a perceptual signal is the
value of the neural current computed by summing the
effects of the firings of a bundle of neurons that respond
similarly to their related inputs. It is closely analogous
to the force on the Puck of a barrage of pebbles thrown by
different players.

          Consider the muscular output part of a control loop. The

tension in a muscle is related to the rate of nerve
firings, and the angle of many joints is set by the
balance between the firings on opposed muscles. To me,
phase 3 of this gedanken experiment looks like that, with
one group of pebble throwers trying to move the puck to
the average of one cluster of reference locations while
the other group tries to move it to the other cluster, the
result being an apparently controlled location of the Puck
(the joint angle) between the two clusters. If there had
been a “Con” directing the players, Con could have moved
the Puck in a controlled manner by telling one group to go
easy and the other to throw more rapidly. Likewise we move
our joints in a controlled manner by changing the balance
between nerve firings in the opposed muscles according to
reference values sent down from higher levels.

          In our thread(s) on the reality of what we perceive, the

assumption throughout has been that the perceptual signal
is real and the reality of what we think we perceive is
questionable. Perhaps we should ask the question in the
other direction. Assume that there is something real,
though we know not what it is, and that some property of
that something influences the firings by the neurons in
our perceptual apparatus (and is influenced by things we
do with our muscles). The real thing is unitary but
unknown, but the perception is virtual and known (at least
some perceptions are known to consciousness). Perceptual
functions are presumably implemented by the synapses and
hormonal environment of every neuron in the perceptual
apparatus, and the unitary real world affects the inputs
to them. Those inputs and their structure and local
chemical environment determine their firing rates.

          So we come to the strange position that our perceptions

are the only things that we can be sure of, even though
they are virtual, existing nowhere in the brain as a
locatable value. We survive because we can use our actions
to control those virtual perceptions, just as all our
ancestors have done back to the first stirrings of life on
the planet. They work in real reality, regardless of
theory.

          Stochastic collective control in our bodies allows us to

live, and stochastic collective control in our environment
allows us to create artifactual entities such as ways of
behaving with respect to each other that are stable,
reliable, ways of getting what we want – atenfels that
are often intangible. We live because things “out there”
usually change as they would if our perceptions were
veridical when we control them, even though the
perceptions we control are themselves only virtual.

              Martin

[From Rick Marken (2016.12.26.1200)

Re TCV and Collective Control2.jpg

Re TCV and Collective Control3.jpg

Re TCV and Collective Control5.jpg

Re TCV and Collective Control4.jpg

Re TCV and Collective Control6.jpg

···

Bruce Nevin (2016.12.25.21.48 ET)–

RM: Hi Bruce. It’s been a while. Good to see you back.

BN: A pissing contest about correct recitation of TCV scripture interests me far less than the analogy of ‘stochastic collective control’ …to a perceptual signal…

RM: The problem with Martin’s gedanken experiment was not in the way he “recited” TCV scripture. The problem was that he described a TCV being carried out incorrectly. Actually, that wouldn’t have troubled me much since Martin is an unabashed “theory first” fellow who admits that observation is not his forte. What did trouble me was his use of this gedanken experiment to show that the TCV could lead to ridiculous conclusions (particularly when then phenomenon under study involves collective control) the implication being that theory not only comes first but that observation can’t be trusted. Here’s Martin’s conclusion:

MT: Dee has performed the Test for the Controlled variable, and has come to the conclusion that an entity she calls “Con” exists and has an internal perception of the Puck location that he controls by having pebbles shot at the puck from different directions. But we know that Con does not exist and does not need to exist for Dee to observe what she observed. Only the players exist, controlling their individual perceptions of the Puck location according to their individual reference locations for it.

RM: So the TCV has led the intrepid PCT researcher, Dee, to the false conclusion that the location of the puck is controlled by a single control system. But we, presumably the wise people who ignore phenomena and deal in theory first, somehow know that the controlling is done by a group of individuals controlling the location of the puck relative to their own references for perceptual variable.

RM: I don’t mind the fact that some people want to approach PCT “theory first” as long as they also approach it “phenomena second” and loop back to the theory, if necessary. But I do mind it when people approach PCT as “theory only”, which is precisely the approach that Martin’s gedanken experiment was trying to justify. And this was particularly infuriating because Martin did this using an incorrect description of how the TCV would be carried out and what it would reveal.

RM: I realize that this theory only (also know as fact-free) approach to understanding the world officially started on November 8, 2016. But that doesn’t mean I have to accept it. Indeed, I think this fact-free approach to understanding is as bad for PCT as it will surely prove to be for the world in general.

BN: It seems to me that this addresses the important question, how an input function is created and maintained.

RM: I think the basic reorganization model of PCT can handle this in principle. But in order to really understand how input functions are created and maintained I believe we first have to know what perceptions these functions are creating and how those functions create those perceptions. The former requires observational studies based on the TCV; the latter require modeling studies which would involve building neurophysiologically plausible mechanisms for constructing these perceptions.

Best

Rick

A control loop does whatever works to control its perceptual input in accord with its reference input. In addition to the immediate effect of varying its output (the standard view), “whatever works” may include the effect of the slower process of modifying one or both of those inputs.

Generally, we say that an input function is created and maintained by ‘the reorganization system’. I’m wondering if a model of ‘stochastic collective control’ might be a model of how the nerve fibers arriving at a comparator do that.


Richard S. Marken

“The childhood of the human race is far from over. We
have a long way to go before most people will understand that what they do for
others is just as important to their well-being as what they do for
themselves.” – William T. Powers

On Mon, Dec 19, 2016 at 11:36 PM, Martin Taylor mmt-csg@mmtaylor.net wrote:

[Martin Taylor 2016.12.13.12.09]

(Note: "Stochastic collective control" is a name I use for the kind

of collective control described throughout this message. I first
started to use it in private interchanges with Kent McClelland a
couple of months ago. If anyone knows of a place where the term has
been publicly used, I’d appreciate being made aware of it.)

-------A game-------

Imagine the following game. A number of players stand around a very

large irregularly shaped table that has a slippery surface. On the
table is an object like a big ice-hockey puck, so let’s call it “The
Puck”. Each player has access to a stock pile of objects like small
flat pebbles that they can slide across the table to hit the Puck
and nudge it a little (a “throw”). A player can get only one
“pebble” from a stockpile at a time, so it takes a little time after
each throw before they can get another pebble to throw. There is
some device that removes a thrown pebble from the table and returns
it to the stockpile after it has hit or passed by the Puck, so the
table is always clean.

Phase 1 of the gedanken experiment -- stochastic collective control

with a common reference value

All the players belong to a team. On the table is marked a target

location, and the team’s objective is to get the Puck onto the
target as quickly as possible. The figure shows the effect of one
“throw” by the player highlighted.

The team collectively controls the location of the Puck, in a form

of collective control I call “stochastic collective control”.
Clearly, if they are to get the Puck to the target as fast as
possible, the players for whom the target-puck-player angle is less
than 90 degrees should not throw their pebbles. For the purposes of
the game, we assume that a player is quite likely to hit the Puck,
but cannot hit it on a particular part of its circumference. This
implies that individual players cannot direct the Puck accurately.
Instead, when hit, the Puck moves with a component away from the
thrower plus a random sideways component that averages out to zero
in the long term. The above figure shows a hit that result in a
small leftward component, seen from the player’s viewpoint.

Consider the combined effect of three pebbles, one thrown by each of

the three players on the right side of the table in the Figure.
Roughly speaking, the most probable direction of the Puck if all
three threw more or less together is the average of their three
directions, but it could be anywhere in a wide arc. If there were
more players, the distribution of directions would be more peaked.
The total “force” (hits per second) in the most probably direction
is a bit less than the sum of their individual hit rates (actually
the sum of their hit rates times minus the cosine of the angle
subtended by the person, the Puck, and the average direction). The
Figure suggests the probability distributions of the effects of
their single pebbles and of the combination.

Every time a pebble is thrown, the Puck moves a little in a new

direction, but on average it will move toward the target if the
correct team members do the throwing. (This may look a bit like an
e-coli track, but it isn’t, because e-coli keeps going in the same
direction so long as it is getting closer to the target, which is
not the case for the Puck.)

If there were a lot of throwers packed tightly around the table and

throwing rapidly, the Puck’s track would be much more direct,
deviating only slightly from a direct path toward the target. The
effect, to an observer who could not see the individual pebbles,
would look as though the Puck was being slid smoothly over the table
by an invisible hand, in the same way that we see movement in a
movie as smooth though we know it consists of discrete images spaced
in time. So, whereas individual players cannot direct the Puck
accurately with any one throw, the collective can, if there are
enough team members.

When there are a lot of players, each hit has a proportionately

smaller effect on the Puck. The game would be pointless if one hit
moved the Puck a long way across the table, because if it did, then
the multiple hits by a lot of players might send it right off the
table beyond the target after just one throw each. So we must
imagine two things, (a) that one hit moves the Puck only slightly,
and (b) a player is less likely to throw if the Puck is near the
target, so that the team avoids the Puck location overshooting and
oscillating erratically back and forth around the target. The net
effect is that the nearer the Puck is to the target, the smaller the
apparent force pushing it there, just as is the case in a canonical
control system, in which the contribution to the output is
proportional to the error value.

Phase 2 of the gedanken experiment -- variation among individual

reference values

In Phase 2 of the experiment, we remove the marking for the target.

Instead, at the start of a trial, the game-manager points to a spot
and tells the players that where he points is the target location.
The players therefore all agree in general where the target is, but
unknown to each of them, their reference locations all differ
slightly. So long as the Puck is far from the target, the situation
is the same as in Phase 1, but when the Puck is in the general area
of the manager’s target location, some players will think it is to
one side of the target while others think it is to the other.
Pebbles will hit the Puck from all directions, but more frequently
from the direction away from the average of the individual players’
reference locations.

The Puck will move on average toward the average of the individual

reference locations, but will continue to jitter around that point
even when it is exactly there. (Bill Powers noted that engineered
control systems often include a zero-point jitter to avoid static
friction when stopped).

The team controls the Puck location as before, but with a virtual

reference value that might not be the reference value for any member
of the team. We might call the virtual reference value a Platonic
Ideal for the target location.

Phase 3 of the gedanken experiment -- teams in conflict

This time, the game manager tells some of the team one target

location and others a different target location. The individual
players do not know each other’s instructions. In the figure, group
1 is marked by grey dots around the table. Their reference values
for the target location are the right-hand set of reference
locations, filled grey. The black dot players with black-white
target reference values represent group 2.

The puck is being hit with pebbles from players who could move it

toward their particular target reference location, which means that
in the diagrammed position most of the hits from both groups will
move the pebble “northward”, but in the east-west direction it will
be hit from the right by pebbles from the “black” players and from
the left by “grey” players. It will tend to move in whichever
direction it receives most hits. If the Puck nears one group target
area, their contribution to the apparent force on the Puck will
diminish and the contribution of the other group will increase, so
that the Puck will jitter around some point between the average
targets of the two groups. The situation is exactly the same as in
Phase 2. Only the game manager (and us) know that there are two
separate clusters of target reference locations.

The distinction between Phase 3 and Phase 2 is only in the

distribution of the individual reference locations for the target.
Instead of all being randomly clustered around one Platonic ideal
virtual reference level, they cluster around two, but the result
that the Puck will tend toward a location that is the average of all
the individual reference locations remains unchanged. The effect on
the Puck is still as if a hand were controlling a perception of the
Puck location with a reference location around the average of all
the individual reference locations.

Phase 4 of the gedanken experiment -- Testing for the Controlled

Variable.

In Phase 1, a team uses stochastic collective control to move the

Puck onto the target, which is the reference value of their
perceived location of the Puck for all of them. In Phase 2, every
player had a different reference location for the perceived target,
but all the reference locations were clustered around an “ideal”
target. By Phase 3, the individuals not only had different reference
locations for the Puck, but their reference locations were scattered
in two different clusters. Nevertheless, the location of the Puck in
all the phases appeared to be controlled to a single reference
location, even though in Phase 3 that location was far from the
reference location of any of the players.

In Phase 4 the situation for the players is exactly as in Phase 3,

but we now introduce a privileged player, who we can call the
“Disturber” or “Dee”. Dee can see the players and their pebbles, and
can move the Puck, but the players cannot see Dee. Dee wants to
determine whether the Puck location is being controlled, and if so,
to what reference value. So Dee moves the Puck in random directions
and watches what happens. When she moves the Puck, the thrown
pebbles will move it back toward the virtual reference location. For
every player, the Puck is not where the player wants it to be, and
if the individual’s target is beyond the Puck, the individual will
throw.

From Dee's point of view, the fact that the Puck keeps moving back

to the same place no matter where Dee moves it is not enough to
establish that its location is being controlled. For all she knows,
the table might not be flat, and the Puck might just be sliding back
down to the low point. The pebbles might be having no effect on it
at all. So Dee needs to do more. She must either prevent the players
from seeing the Puck or prevent their pebbles from hitting it. Let
us say she hides away the pebble stock pile, so the players have no
pebbles to throw. Now when she moves the Puck, it stays where she
puts it. So she knows that the Puck does not simply slide back down
a slope after she moves it, and she replaces the pebble stockpile.
The pebbles are essential if it is to move back. Next she tries
putting a dry-ice fog over the table so that the players cannot see
the Puck. Once again, after she moves it, it stays put, jiggin a
little when it is hit by an occasional pebble from a random
direction, but not moving toward its former location.

Dee has completed the TCV, and her results lead her to theorize that

the Puck is being controlled by an entity (which Dee calls “Con”)
for whom the players are simply agents. The players, in her theory,
continually observe the Puck location, report their observations to
Con, and throw pebbles when Con instructs them to do so, in order to
keep the Puck near Con’s reference location for it. According to
Dee, Con controls his perception of the Puck’s location, using the
players as the equivalents of his eyes and muscles.

------Commentary---------

Dee has performed the Test for the Controlled variable, and has come

to the conclusion that an entity she calls “Con” exists and has an
internal perception of the Puck location that he controls by having
pebbles shot at the puck from different directions. But we
know that Con does not exist and does not need to exist for Dee to
observe what she observed. Only the players exist, controlling their
individual perceptions of the Puck location according to their
individual reference locations for it.

How do *we* know that within any one player there exists a

perception of the Puck location? We don’t. We are in the same
predicament in that respect as Dee is with respect to Con’s
perception, which, as we know, doesn’t exist because Con does not
exist.

At this point the quote from Bill Powers that Rick found [Bill

Powers (961224.1145 MST)] becomes relevant: “Remember that as far as
the observer is concerned, what is controlled is ONLY the CV. The
idea that this CV is represented by a perceptual signal inside the
other system is theoretical. We can observe CV, but not p. When we
apply a disturbance, we apply it to CV, not to p. The action that
opposes the effect of the disturbance acts on CV, not p. The Test
does not involve p at all. It involves only observables – i.e., the
observer’s perceptions. The observations have priority; the model
comes second, and its only reason for existence is to explain the
observations.”

(Bill follows with this: "When you fool around with

thought-experiments too much, you tend to get the priorities
reversed." I hope I have not fallen into that trap with this thought
experiment).

------Using the game analogy------



Now think about the concept of a "Neural Current", which is a

fictitious quantity that is the sum of many one-shot events – nerve
firings, not pebble throws. The idea of there being a “perceptual
signal” depends on the notion of the neural current. Indeed, a
perceptual signal is the value of the neural current computed by
summing the effects of the firings of a bundle of neurons that
respond similarly to their related inputs. It is closely analogous
to the force on the Puck of a barrage of pebbles thrown by different
players.

Consider the muscular output part of a control loop. The tension in

a muscle is related to the rate of nerve firings, and the angle of
many joints is set by the balance between the firings on opposed
muscles. To me, phase 3 of this gedanken experiment looks like that,
with one group of pebble throwers trying to move the puck to the
average of one cluster of reference locations while the other group
tries to move it to the other cluster, the result being an
apparently controlled location of the Puck (the joint angle) between
the two clusters. If there had been a “Con” directing the players,
Con could have moved the Puck in a controlled manner by telling one
group to go easy and the other to throw more rapidly. Likewise we
move our joints in a controlled manner by changing the balance
between nerve firings in the opposed muscles according to reference
values sent down from higher levels.

In our thread(s) on the reality of what we perceive, the assumption

throughout has been that the perceptual signal is real and the
reality of what we think we perceive is questionable. Perhaps we
should ask the question in the other direction. Assume that there is
something real, though we know not what it is, and that some
property of that something influences the firings by the neurons in
our perceptual apparatus (and is influenced by things we do with our
muscles). The real thing is unitary but unknown, but the perception
is virtual and known (at least some perceptions are known to
consciousness). Perceptual functions are presumably implemented by
the synapses and hormonal environment of every neuron in the
perceptual apparatus, and the unitary real world affects the inputs
to them. Those inputs and their structure and local chemical
environment determine their firing rates.

So we come to the strange position that our perceptions are the only

things that we can be sure of, even though they are virtual,
existing nowhere in the brain as a locatable value. We survive
because we can use our actions to control those virtual perceptions,
just as all our ancestors have done back to the first stirrings of
life on the planet. They work in real reality, regardless of theory.

Stochastic collective control in our bodies allows us to live, and

stochastic collective control in our environment allows us to create
artifactual entities such as ways of behaving with respect to each
other that are stable, reliable, ways of getting what we want –
atenfels that are often intangible. We live because things “out
there” usually change as they would if our perceptions were
veridical when we control them, even though the perceptions we
control are themselves only virtual.

Martin

[Martin Taylor 2016.12.26.15.16]

[From Rick Marken (2016.12.26.1200)

Rick, you said you were going to explain what Dee did wrong in

performing the TCV. I presume that when you do so, you will also let
us know how Dee should have done it properly, and how doing it
properly would allow her to tell whether the pebble-throwers are
autonomous or act the sensors and “muscles” of a single perceptual
controller.

A rant about my style of approach to PCT does not serve that

purpose.

  I do mind it when people approach PCT as

“theory only”, which is precisely the approach that Martin’s
gedanken experiment was trying to justify. And this was
particularly infuriating because Martin did this using an
incorrect description of how the TCV would be carried out and what
it would reveal.

Personally, I thought the gedanken experiment was designed to point

out the inadequacy of a theory-only approach. But that’s just me,
the designer of the experiment.

You yourself found the Powers quote: "Remember that as far as the

observer is concerned, what is controlled is ONLY the CV. The idea
that this CV is represented by a perceptual signal inside the other
system is theoretical." The gedanken experiment simply shows that
there is another theoretical possibility for control that does not
require a unitary “p”.

To insist that Powers was wrong about there being other

possibilities seems to me to be the “Theory Only” approach: “There
is One Theory to Rule them all, One Theory to bring them all and in
the darkness bind them”. Powers knew better. He knew his basic
theory would need to be expanded with new ideas and into new
territories. He did not want to bind his successors into (an
admittedly rather bright) prison cell.

As Kent said [From Kent McClelland (2016.12.14.1600)]:
"Rick is the one whose arguments are being driven by his theory (a

narrowly constricted view of how PCT should be done), rather than
any empirical observations. His flat statement that " in real conflicts, both systems quickly go to
maximum output� is simply absurd."

As for "phenomena" versus "theory", in the gedanken experiment the

observed Puck movement is what Dee wants to test by experiment, to
see whether it is an equilibrium process like a spring, or control.
Having determined that the phenomenon is control, Dee theorizes that
there is a controller, just as we do when we run the TCV on a
control phenomenon in one of your demos. We are precisely in Dee’s
position with regard to the interaction of phenomena and theory.

You can refute the analogy by showing what "infuriated" you about

the way Dee performed the TCV, and why she did it incorrectly badly
enough to invalidate both her finding of control, and her
theoretical interpretation. That is the gift you said you would put
under the Xmas tree, is it not?

Martin

[From Rick Marken (2016.12.26.1830)]

···

Martin Taylor (2016.12.26.15.16)–

MT: Rick, you said you were going to explain what Dee did wrong in

performing the TCV.

RM: I was still waiting for someone else to do it. But I’ll go ahead. I’ll be quoting from your original gedanken experiment post in what follows. The problem starts here:

MT: from Dee’s point of view, the fact that the Puck keeps moving back to the same place no matter where Dee moves it is not enough to establish that its location is being controlled.

RM: The puck moving back to the place no matter where Dee moves it has nothing to do with determining whether or not the location of the puck is controlled. Before doing the test Dee would have to know how much the puck would be expected to move if it were not under control. And this requires knowing how the pebbles as well as her own forces would affect the location of the puck if it were not under control. Once all that was figured out Dee could measure the actual amount of variation in puck location that is produced by her attempts to move it and compare that with the expected amount of variation that would be produced. If the actual change in location is much smaller than the expected change, that is evidence that the location of the puck is under control. You go on:

MT: For all she knows, the table might not be flat, and the Puck might just be sliding back down to the low point. The pebbles might be having no effect on it at all.

RM: If she were doing the test properly she would have taken into account all the forces acting on the puck – surface resistance, slant, pebble impacts, etc – that would determine how much the puck would be expected to change location if she tried to move it and it were not under control. The first thing you do when testing for control is to hypothesize what the controlled variable is; in this case Dee was hypothesizing that the location of the puck was under control. The next step is to determine the expected effect of disturbances to this hypothesized controlled variable (such as Dee’s attempts to move the puck) is the variable were not under control.

MT: So Dee needs to do more.She must either prevent the players from seeing the Puck or prevent their pebbles from hitting it.

RM: No, neither of these things are relevant. What she needs to do is determine how the location of the puck is expected to be changed by her efforts to move it if the puck were not under control.

MT: Let us say she hides away the pebble stock pile, so the players have no pebbles to throw. Now when she moves the Puck, it stays where she puts it. So she knows that the Puck does not simply slide back down a slope after she moves it, and she replaces the pebble stockpile.

RM: Actually, she has found out one of the things she should have figured out in the first place; how much the location of the puck changes when it is disturbed and not under control. She was lucky to guess that the pebbles might have something to do with the puck staying in position. She learned that the puck moved far less than expected when the pebbles were being thrown. So puck location is very likely a controlled variable and the means of control is likely to be the pebbles thrown by the players.

MT: The pebbles are essential if it is to move back.

RM: Actually, the moving back is less important than the fact that she is able to move the puck far less than she could when the pebbles are being thrown.

MT: Next she tries putting a dry-ice fog over the table so that the players cannot see the Puck. Once again, after she moves it, it stays put, jiggin a little when it is hit by an occasional pebble from a random direction, but not moving toward its former location.

RM: This establishes that it is the players that are doing the controlling.

MT: Dee has completed the TCV, and her results lead her to theorize that the Puck is being controlled by an entity (which Dee calls “Con”) for whom the players are simply agents.

RM: Well, given how poorly Dee has conducted the research up to this point I imagine she might, indeed, come up with this hair-brained theory. But a more talented researcher would conclude only that the position of the puck is likely to be a controlled variable. And that researcher would develop a model to account for that observed phenomenon. The first version of the model I would produce would have each of the players controlling for a visual representation of the position of the puck relative to a different reference specification. This model explains the observed reference state of the puck as being “virtual”, resulting from a conflict between the players. I would then use this model to make predictions about what would happen to the observed reference state of the puck given changes in things like the gain of individual players (gain can be changed in this case by changing the size of the pebbles available to each player). If these predictions didn’t hold up I would try a different model, now informed by the observations made while testing this first model.

MT: How do we know that within any one player there exists a perception of the Puck location?

RM: We don’t “know” this; it’s part of the theory – PCT – that explains what we observe – the controlled position of the puck.

RM: Now back to your comments from the present post:

MT: I presume that when you do so, you will also let

us know how Dee should have done it properly, and how doing it
properly would allow her to tell whether the pebble-throwers are
autonomous or act the sensors and “muscles” of a single perceptual
controller.

RM: Yes, done!

MT: Personally, I thought the gedanken experiment was designed to point

out the inadequacy of a theory-only approach. But that’s just me,
the designer of the experiment.

RM: I find that hard to believe. You described a situation where a researcher, Dee, was making observations using your version of the TCV. So Dee was trying to take Bill and my advice and start with phenomena first. Then she went on to formulate a cockamamie theory to explain her poorly made observations. So this was not a theory-only approach; it was an example of a poorly done phenomena-first, theory second approach.

MT: You yourself found the Powers quote: "Remember that as far as the

observer is concerned, what is controlled is ONLY the CV. The idea
that this CV is represented by a perceptual signal inside the other
system is theoretical." The gedanken experiment simply shows that
there is another theoretical possibility for control that does not
require a unitary “p”.

RM: The observation that the location of the puck is a CV does not “require a unitary “p””. All it requires is an explanation. And it would be obvious to any competent researcher looking at this peculiar game that some or all of the players throwing pebbles at the puck are contributing to control of the puck’s location. So any theory that accounts for the reference state of this CV will have to have control organizations in each of the players, meaning that each of the players has a possibly different reference for the position of the puck.

MT: To insist that Powers was wrong about there being other

possibilities seems to me to be the “Theory Only” approach:

RM: If I was “insisting” anything it was that Powers was right. What I understood Bill to be saying was that the CV is the phenomenon that we observe; the perceptual signal, p, is part of the theory that explains this phenomenon. You were apparently interpreting Bill to be saying that there is one and only one p for any CV. I guarantee you that this is not what Bill was saying. Bill understood how conflict works and was well aware that variables could be held in virtual reference states when two or more control systems were controlling perceptual signals, p, corresponding to the same CV relative to different references. The point of Bill’s post was that the phenomena of control – CV’s – are explained by the theory of control – the variables p, r, and e.

MT: You can refute the analogy by showing what "infuriated" you about

the way Dee performed the TCV, and why she did it incorrectly badly
enough to invalidate both her finding of control, and her
theoretical interpretation. That is the gift you said you would put
under the Xmas tree, is it not?

RM: The problem with your analogy is that it shows a researcher coming up with a ridiculous theory (that there is one master controller using a set of players as his means of control) based on the observation of a phenomenon (the fact that the location of the puck is a CV). It’s hard for me to see your gedanken experiment as anything other than an effort to show that the observation of a collectively controlled CV (if not all CVs) is either misleading or unnecessary or both. If, as you say, your analogy is about the shortcomings of a theory-only approach then I think a much better way to show it would be to show how a collective control model accounts for some actual, observed collectively controlled variable And that means showing it the way Powers and Bourbon and myself have shown it in our research papers; by showing the fit of the model to the data. Here’s one of my object interception papers that shows how I think it should be done.

https://www.dropbox.com/s/e7flu4vb486k0zw/Chasin%27Choppers.pdf?dl=0

See in particular Figure 6.

Best regards

Rick


Richard S. Marken

“The childhood of the human race is far from over. We
have a long way to go before most people will understand that what they do for
others is just as important to their well-being as what they do for
themselves.” – William T. Powers

[Martin Taylor 2016.12.27.11.19]

[From Rick Marken (2016.12.26.1830)]

That's a very interesting restriction on the TCV. You imply that

there’s no way to determine that the position of a car in its lane
is being controlled, because neither the driver nor the
Analyst/Experimenter/Observer can know each wind gust, each pothole,
each change of bearing friction, each change in road slope, …
Since Dee cannot observe the vertical configuration of the table or
the varying friction of differently polished parts of its surface,
nor is she measuring her output force (the experimental description
did not credit her with any way of measuring it), therefore Dee has
no way of determining that the Puck location is controlled.

I guess you have just destroyed "The Fact of Control", and the

edifice of PCT has lost its main raison d’etre. That’s quite an
accomplishment in one paragraph. Congratulations, I think.

But let's assume that PCT survives. The validity or otherwise of the

rest of your analysis is irrelevant because the following is all
that I wanted to establish.

That fact was actually my reason for introducing the gedanken

experiment.

Since you accept that "the location of the puck is a CV does not

“require a unitary “p”” (or rather, insist on it), I need no further
support for my later discussion of the relation between Bill’s
“neural current” approximation of distributed firing patterns in the
brain and the perception of the (presumed to exist) “boss reality”.

Since you accept the possibility that nowhere in the brain is there

a non-virtual perceptual value, despite the apparent singularity of
conscious perceptions, we can proceed with a serious discussion of
collective control and virtual perceptual signals, applying it where
appropriate and useful in both individual and social PCT research.

Thank you. I predict a happier New Year (on CSGnet, if not in

politics).

Martin

PS. But I still wonder why you accept only TCV condition 4a and

dismiss conditions 4b and 4c of my list as criteria for determining
whether control is occurring and what variable is controlled, when
you use them in your own demos of the TCV.

···

Martin Taylor (2016.12.26.15.16)–

            MT: Rick, you said you were going to explain what Dee

did wrong in performing the TCV.

          RM: I was still waiting for someone else to do it. But

I’ll go ahead. I’ll be quoting from your original gedanken
experiment post in what follows. The problem starts here:

            MT: From Dee's point of view,

the fact that the Puck keeps moving back to the same
place no matter where Dee moves it is not enough to
establish that its location is being controlled.

            RM: The puck moving back

to the place no matter where Dee moves it has nothing to
do with determining whether or not the location of the
puck is controlled. Before doing the test Dee would have
to know how much the puck would be expected to
move if it were not under control. And this
requires knowing how the pebbles as well as her own
forces would affect the location of the puck if it were
not under control. Once all that was figured out Dee
could measure the actual amount of variation in puck
location that is produced by her attempts to move it and
compare that with the expected amount of variation that
would be produced. If the actual change in location is
much smaller than the expected change, that is evidence
that the location of the puck is under control.

            MT: You yourself found the Powers

quote: “Remember that as far as the observer is
concerned, what is controlled is ONLY the CV. The idea
that this CV is represented by a perceptual signal
inside the other system is theoretical.” The gedanken
experiment simply shows that there is another
theoretical possibility for control that does not
require a unitary “p”.

          RM: The observation that the location of the puck is a

CV does not “require a unitary “p””.

[From Rick Marken (2016.12.28.1145)]

···

Martin Taylor (2016.12.27.11.19)–

MT: That's a very interesting restriction on the TCV. You imply that

there’s no way to determine that the position of a car in its lane
is being controlled, because neither the driver nor the
Analyst/Experimenter/Observer can know each wind gust, each pothole,
each change of bearing friction, each change in road slope, …

RM: All you have to know is how much the car’s position is expected to vary if it were not under control. We know how unexpected wind gusts, potholes and changes in bearing friction will affect the car’s position; we also know how much the car’s position will vary (due to these and other factors) if the car proceeds down the road sans driver. But if we wanted to do experiments to see what optical variable the driver is actually controlling we would do this under controlled conditions where we are able to vary one specific variable, such as the force of a side-wind, and monitor the effect of this disturbance on the hypothesized controlled variable. This kind of research would probably have to be done in a driving simulator.

MT: Since you accept the possibility that nowhere in the brain is there

a non-virtual perceptual value,

RM: I do? What I “accept” is a theory, called PCT, that explains the existence of observed reference states for controlled variables in terms of the existence of neural reference signals that specify the state of perceptual signals that correspond to variations in those controlled variable. I also accept that there are situations where a variable is observed to be maintained in what turns out to be a virtual reference state, one that results from a conflict between control systems trying to get that variable to match reference signals that specify different states of that variable.

MT: despite the apparent singularity of

conscious perceptions, we can proceed with a serious discussion of
collective control and virtual perceptual signals, applying it where
appropriate and useful in both individual and social PCT research.

RM: I have no idea what a “virtual perceptual signal” is. But perhaps what it is – and why it’s needed – will become clear like in a serious discussion of collective control, a discussion I would very much like to have. I think we should start with you (or Kent) describing an example of a stable social variable and then showing how this variable is maintained in a stable state (a virtual reference state) by collective control.

MT: PS. But I still wonder why you accept only TCV condition 4a and

dismiss conditions 4b and 4c of my list as criteria for determining
whether control is occurring and what variable is controlled, when
you use them in your own demos of the TCV.

RM: I don’t dismiss them. Here are your three conditions (criteria) for determining that a variable is under control:

a. whether the hypothesized property changes appreciably less than would be expected if a corresponding perception were not being controlled, or
b. measuring the influence of the putative controller and determining that it is negatively correlated with the (time-lagged) influence of the experimenter-induced disturbance, or
c. measuring the varying value of the complex property and determining that its variation has a low correlation with variation in the applied disturbance.

RM: All these criteria are fine with me. Criteria a and c are basically the same: both involve looking to see whether variations in a disturbance to a hypothesized controlled variable produces less variance in that variable than expected. Criterion b is certainly a good one but often difficult to use in practice since the researcher’s is often not the only disturbance affecting a hypothesized controlled variable. This is especially true when doing the TCV in real-world situations. I have, indeed, used it in some of my demos because in these situations it is possible to control (in the methodological sense of “hold constant”) all influences on the hypothetical controlled variable other than the experimentally produced disturbance.

RM: I look forward to learning how collective control explains the stability of a social variable.

Best

Rick


Richard S. Marken

“The childhood of the human race is far from over. We
have a long way to go before most people will understand that what they do for
others is just as important to their well-being as what they do for
themselves.” – William T. Powers

            RM: The puck moving back

to the place no matter where Dee moves it has nothing to
do with determining whether or not the location of the
puck is controlled. Before doing the test Dee would have
to know how much the puck would be expected to
move if it were not under control.

Hi Rick, I think to critique Martin’s thought experiment, don’t you need to assume from the start that this game is an analogy of a single individual who, we know, will have multiple control systems at lower levels of their hierarchy operating through various physical feedback functions? So we are assuming that a TCV experimenter analysing control without anatomy and physics is like Dee doing the TCV without awareness of the many people doing the control?

I think Martin has demonstrated that when we test for a CV we are still aware that there is a cascade of control systems implementing the perceptual control of a higher level system. I think in tandem it represents that it can look as though there is a higher level system simply because every lower level control unit has the same or a similar reference point. But I think this might be overstepping the analogy. In this game, there IS a master controller - the leader of the game who the players have chosen to follow and take on board his or her reference point. This higher level control system would be needed within each individual in order to have any ability to update the target point to a new value - it would need to be sent down to these lower level control systems.

So my view is that Martin’s thought experiment does remind us that a hypothesised CV does not tell us about the full working, hierarchical architecture of how a perceived aspect of the environment is controlled. However, it doesn’t to me explain how a virtual higher level controller seems to exist when it doesn’t - it seems to me that for the people in this thought experiment to even be willing to take part in it and update their targets, they need a higher level control system in addition to the one that Martin proposed to exist to get the puck on the target.

A rethink here : Maybe it is fairer to say it doesn’t rule out the possibility of a higher level controller - this scenario could work if there was a system operating at a broader timescale that led people in a shared environment to adopt similar reference values (I am not sure whether ‘evolution’ is in fact a master controller in functional terms though - probably is?) - an intrinsic tendency to prefer familiarity, or an intrinsic, evolved reference value for the target in question (such as a bee’s in built preference for petal markings that converge on a point in the centre of a flower - where the nectar is). Group living insects certainly appear to have a master controller, and an experimenter who only had data on the nectar levels in flowers and nectar levels in a hive might think a massive creature was shifting the stuff into the hive.

I think this does remind us that we need not to ignore the value we get from the bleeding obvious when doing the TCV if we want a fully working model - the anatomy, physiology, dynamics, etc…

I feel as if I am sitting in between Martin and Rick here, which is probably my preferred reference point anyway and might have been the ‘higher level reference point’ that led me to the above conclusions!

Hope everyone had a great Christmas!

I would be interested in Martin and Kent’s responses to the above.

Warren

···

On Wed, Dec 28, 2016 at 7:47 PM, Richard Marken rsmarken@gmail.com wrote:

[From Rick Marken (2016.12.28.1145)]

Martin Taylor (2016.12.27.11.19)–

MT: That's a very interesting restriction on the TCV. You imply that

there’s no way to determine that the position of a car in its lane
is being controlled, because neither the driver nor the
Analyst/Experimenter/Observer can know each wind gust, each pothole,
each change of bearing friction, each change in road slope, …

RM: All you have to know is how much the car’s position is expected to vary if it were not under control. We know how unexpected wind gusts, potholes and changes in bearing friction will affect the car’s position; we also know how much the car’s position will vary (due to these and other factors) if the car proceeds down the road sans driver. But if we wanted to do experiments to see what optical variable the driver is actually controlling we would do this under controlled conditions where we are able to vary one specific variable, such as the force of a side-wind, and monitor the effect of this disturbance on the hypothesized controlled variable. This kind of research would probably have to be done in a driving simulator.

MT: Since you accept the possibility that nowhere in the brain is there

a non-virtual perceptual value,

RM: I do? What I “accept” is a theory, called PCT, that explains the existence of observed reference states for controlled variables in terms of the existence of neural reference signals that specify the state of perceptual signals that correspond to variations in those controlled variable. I also accept that there are situations where a variable is observed to be maintained in what turns out to be a virtual reference state, one that results from a conflict between control systems trying to get that variable to match reference signals that specify different states of that variable.

MT: despite the apparent singularity of

conscious perceptions, we can proceed with a serious discussion of
collective control and virtual perceptual signals, applying it where
appropriate and useful in both individual and social PCT research.

RM: I have no idea what a “virtual perceptual signal” is. But perhaps what it is – and why it’s needed – will become clear like in a serious discussion of collective control, a discussion I would very much like to have. I think we should start with you (or Kent) describing an example of a stable social variable and then showing how this variable is maintained in a stable state (a virtual reference state) by collective control.

MT: PS. But I still wonder why you accept only TCV condition 4a and

dismiss conditions 4b and 4c of my list as criteria for determining
whether control is occurring and what variable is controlled, when
you use them in your own demos of the TCV.

RM: I don’t dismiss them. Here are your three conditions (criteria) for determining that a variable is under control:

a. whether the hypothesized property changes appreciably less than would be expected if a corresponding perception were not being controlled, or
b. measuring the influence of the putative controller and determining that it is negatively correlated with the (time-lagged) influence of the experimenter-induced disturbance, or
c. measuring the varying value of the complex property and determining that its variation has a low correlation with variation in the applied disturbance.

RM: All these criteria are fine with me. Criteria a and c are basically the same: both involve looking to see whether variations in a disturbance to a hypothesized controlled variable produces less variance in that variable than expected. Criterion b is certainly a good one but often difficult to use in practice since the researcher’s is often not the only disturbance affecting a hypothesized controlled variable. This is especially true when doing the TCV in real-world situations. I have, indeed, used it in some of my demos because in these situations it is possible to control (in the methodological sense of “hold constant”) all influences on the hypothetical controlled variable other than the experimentally produced disturbance.

RM: I look forward to learning how collective control explains the stability of a social variable.

Best

Rick


Richard S. Marken

“The childhood of the human race is far from over. We
have a long way to go before most people will understand that what they do for
others is just as important to their well-being as what they do for
themselves.” – William T. Powers

            RM: The puck moving back

to the place no matter where Dee moves it has nothing to
do with determining whether or not the location of the
puck is controlled. Before doing the test Dee would have
to know how much the puck would be expected to
move if it were not under control.

Dr Warren Mansell
Reader in Clinical Psychology

School of Health Sciences
2nd Floor Zochonis Building
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

Advanced notice of a new transdiagnostic therapy manual, authored by Carey, Mansell & Tai - Principles-Based Counselling and Psychotherapy: A Method of Levels Approach

Available Now

Check www.pctweb.org for further information on Perceptual Control Theory

[Martin Taylor 2016.12.29.11.02]

Warren, have you given up using a date-time stamp, or did I never notice that you refrained from using one? Neither of your messages today have one, and I have not looked back to check.

Hi Rick, I think to critique Martin's thought experiment, don't you need to assume from the start that this game is an analogy of a single individual who, we know, will have multiple control systems at lower levels of their hierarchy operating through various physical feedback functions? So we are assuming that a TCV experimenter analysing control without anatomy and physics is like Dee doing the TCV without awareness of the many people doing the control?

I think Martin has demonstrated that when we test for a CV we are still aware that there is a cascade of control systems implementing the perceptual control of a higher level system. I think in tandem it represents that it can look as though there is a higher level system simply because every lower level control unit has the same or a similar reference point. But I think this might be overstepping the analogy. In this game, there IS a master controller - the leader of the game who the players have chosen to follow and take on board his or her reference point. This higher level control system would be needed within each individual in order to have any ability to update the target point to a new value - it would need to be sent down to these lower level control systems.

I guess I should have included a "Phase 2a" in which the players individually choose their reference values for the puck location with no "game manager" to guide them. The diagram would be exactly the same as for Phase 2 as originally described, and so would be the result. The analogy would be to a set of conflicting control units in two dimensions, just like Kent's original one-dimensional conflicted control demonstration -- apart from the stochastic nature of their control in my gedanken experiment.

So my view is that Martin's thought experiment does remind us that a hypothesised CV does not tell us about the full working, hierarchical architecture of how a perceived aspect of the environment is controlled. However, it doesn't to me explain how a virtual higher level controller seems to exist when it doesn't - it seems to me that for the people in this thought experiment to even be willing to take part in it and update their targets, they need a higher level control system in addition to the one that Martin proposed to exist to get the puck on the target.

Why? If they were real people rather than simple pebble-throwing mechanisms with the characteristics I defined (imperfect throwing accuracy, reduced throwing rate as the Puck nears the target), all that would be necessary would be a willingness to sign up for an experiment, for which their instruction would be something like "There's a thick disk on this big table we call the Puck. Please choose a point on the table and slide pebbles from this stockpile one at a time to hit the Puck so that it finally comes to sit on your chosen point."

The Puck would move in exactly the same way after every time Dee put is in a random place on the table as it would if the players inaccurately took their target location from a master game-manager (higher-level setter of reference values).

A rethink here : Maybe it is fairer to say it doesn't rule out the possibility of a higher level controller

Absolutely. The reason for introducing the experiment was to show that you can't tell from one such experiment whether the observations are the result of individual independent controllers or individual controllers that get their reference values from a centralized higher-level controller, whether the individuals are in separate bodies (sociology) or in one body (psychology).

I have a short taxonomy of types of collective control. In the experiment, the CCEV (Collective Complex Environmental Variable) is the Puck's location, which is also the CEV for each individual controller in the first three types:

1. Conflicted Control: The participants have individually derived reference values for perceptions whose CEVs are closely related to the CCEV. The CCEV remains as if it corresponds to a controlled perception, but the outputs of the individual controllers tend to increase as in any conflict. Several people throw at the Puck, all wanting it in a different place
2. Collaborative Control: The participants control a higher level set of perceptions of belonging and being seen to belong to “the group�?, bringing their reference values for their perceptions of the CEVs that combine to form the CCEV toward a common value, eliminating the conflict while maintaining strong control. Several people throw at the Puck trying to move it to a place on which they agree.
3. Coordinated Control: All members controlling for perceiving themselves and being perceived as belonging to the group accept reference values provided by an agreed leader. Several people throw at the Puck trying to get it to a place designated by one of them.

In addition, there are at least three forms of Collective Control in which the participants act on different aspects of the environment in order to achieve a purpose all have in common, rather than all trying to influence the common CCEV in the same way. These are irrelevant to the gedanken experiment, but I include them anyway for later reference if required.

4. Guided Control: A plan, with or without a specific planner, determines who does what ("I’ll hold the pole while you hammer it into the ground", when we both have a controlled perception of the pole, with a reference for it to be fixed vertically).
5. Social Control Unit: Different people or groups of people use protocols in ways that make some play the roles of the different units of a control unit (Sensors, Perceptual Function, Reference � the CEEO or the Commander �, Comparator, Output function, Effectors), so that tthe whole social structure (organisation, company, military unit ...) acts as a controller (See <http://www.mmtaylor.net/PCT/CSG2005/CSG2005cSocialControl.ppt&gt;\)
6. Hierarchy of Social Control Units: Same as 5, with different levels of controller interacting as in the Powers hierarchy for control units within an organism.

- this scenario could work if there was a system operating at a broader timescale that led people in a shared environment to adopt similar reference values (I am not sure whether 'evolution' is in fact a master controller in functional terms though - probably is?) - an intrinsic tendency to prefer familiarity, or an intrinsic, evolved reference value for the target in question (such as a bee's in built preference for petal markings that converge on a point in the centre of a flower - where the nectar is). Group living insects certainly appear to have a master controller, and an experimenter who only had data on the nectar levels in flowers and nectar levels in a hive might think a massive creature was shifting the stuff into the hive.

An evolved version of coordinated control. This is what I imagine to happen in the evolution of protocols, language, and culture generally.

I think this does remind us that we need not to ignore the value we get from the bleeding obvious when doing the TCV if we want a fully working model - the anatomy, physiology, dynamics, etc....

I feel as if I am sitting in between Martin and Rick here, which is probably my preferred reference point anyway and might have been the 'higher level reference point' that led me to the above conclusions!

If that is so, since insofar as I understand your comments above I totally agree with you, therefore Rick agrees with both of us, and a Happy New Year on CSGnet will follow!

Hope everyone had a great Christmas!

I would be interested in Martin and Kent's responses to the above.

Warren

I hope you do find mine interesting.

Martin

···

On 2016/12/29 5:34 AM, Warren Mansell wrote:

Thanks Martin, that elaboration does help! I think the time stamp happens automatically when you reply?
Warren

···

On 29 Dec 2016, at 16:52, Martin Taylor <mmt-csg@mmtaylor.net> wrote:

[Martin Taylor 2016.12.29.11.02]

Warren, have you given up using a date-time stamp, or did I never notice that you refrained from using one? Neither of your messages today have one, and I have not looked back to check.

On 2016/12/29 5:34 AM, Warren Mansell wrote:
Hi Rick, I think to critique Martin's thought experiment, don't you need to assume from the start that this game is an analogy of a single individual who, we know, will have multiple control systems at lower levels of their hierarchy operating through various physical feedback functions? So we are assuming that a TCV experimenter analysing control without anatomy and physics is like Dee doing the TCV without awareness of the many people doing the control?

I think Martin has demonstrated that when we test for a CV we are still aware that there is a cascade of control systems implementing the perceptual control of a higher level system. I think in tandem it represents that it can look as though there is a higher level system simply because every lower level control unit has the same or a similar reference point. But I think this might be overstepping the analogy. In this game, there IS a master controller - the leader of the game who the players have chosen to follow and take on board his or her reference point. This higher level control system would be needed within each individual in order to have any ability to update the target point to a new value - it would need to be sent down to these lower level control systems.

I guess I should have included a "Phase 2a" in which the players individually choose their reference values for the puck location with no "game manager" to guide them. The diagram would be exactly the same as for Phase 2 as originally described, and so would be the result. The analogy would be to a set of conflicting control units in two dimensions, just like Kent's original one-dimensional conflicted control demonstration -- apart from the stochastic nature of their control in my gedanken experiment.

So my view is that Martin's thought experiment does remind us that a hypothesised CV does not tell us about the full working, hierarchical architecture of how a perceived aspect of the environment is controlled. However, it doesn't to me explain how a virtual higher level controller seems to exist when it doesn't - it seems to me that for the people in this thought experiment to even be willing to take part in it and update their targets, they need a higher level control system in addition to the one that Martin proposed to exist to get the puck on the target.

Why? If they were real people rather than simple pebble-throwing mechanisms with the characteristics I defined (imperfect throwing accuracy, reduced throwing rate as the Puck nears the target), all that would be necessary would be a willingness to sign up for an experiment, for which their instruction would be something like "There's a thick disk on this big table we call the Puck. Please choose a point on the table and slide pebbles from this stockpile one at a time to hit the Puck so that it finally comes to sit on your chosen point."

The Puck would move in exactly the same way after every time Dee put is in a random place on the table as it would if the players inaccurately took their target location from a master game-manager (higher-level setter of reference values).

A rethink here : Maybe it is fairer to say it doesn't rule out the possibility of a higher level controller

Absolutely. The reason for introducing the experiment was to show that you can't tell from one such experiment whether the observations are the result of individual independent controllers or individual controllers that get their reference values from a centralized higher-level controller, whether the individuals are in separate bodies (sociology) or in one body (psychology).

I have a short taxonomy of types of collective control. In the experiment, the CCEV (Collective Complex Environmental Variable) is the Puck's location, which is also the CEV for each individual controller in the first three types:

1. Conflicted Control: The participants have individually derived reference values for perceptions whose CEVs are closely related to the CCEV. The CCEV remains as if it corresponds to a controlled perception, but the outputs of the individual controllers tend to increase as in any conflict. Several people throw at the Puck, all wanting it in a different place
2. Collaborative Control: The participants control a higher level set of perceptions of belonging and being seen to belong to “the group�?, bringing their reference values for their perceptions of the CEVs that combine to form the CCEV toward a common value, eliminating the conflict while maintaining strong control. Several people throw at the Puck trying to move it to a place on which they agree.
3. Coordinated Control: All members controlling for perceiving themselves and being perceived as belonging to the group accept reference values provided by an agreed leader. Several people throw at the Puck trying to get it to a place designated by one of them.

In addition, there are at least three forms of Collective Control in which the participants act on different aspects of the environment in order to achieve a purpose all have in common, rather than all trying to influence the common CCEV in the same way. These are irrelevant to the gedanken experiment, but I include them anyway for later reference if required.

4. Guided Control: A plan, with or without a specific planner, determines who does what ("I’ll hold the pole while you hammer it into the ground", when we both have a controlled perception of the pole, with a reference for it to be fixed vertically).
5. Social Control Unit: Different people or groups of people use protocols in ways that make some play the roles of the different units of a control unit (Sensors, Perceptual Function, Reference � the CEO or the Commander �, Comparator, Output function, Effectors), so that the wwhole social structure (organisation, company, military unit ...) acts as a controller (See <http://www.mmtaylor.net/PCT/CSG2005/CSG2005cSocialControl.ppt&gt;\)
6. Hierarchy of Social Control Units: Same as 5, with different levels of controller interacting as in the Powers hierarchy for control units within an organism.

- this scenario could work if there was a system operating at a broader timescale that led people in a shared environment to adopt similar reference values (I am not sure whether 'evolution' is in fact a master controller in functional terms though - probably is?) - an intrinsic tendency to prefer familiarity, or an intrinsic, evolved reference value for the target in question (such as a bee's in built preference for petal markings that converge on a point in the centre of a flower - where the nectar is). Group living insects certainly appear to have a master controller, and an experimenter who only had data on the nectar levels in flowers and nectar levels in a hive might think a massive creature was shifting the stuff into the hive.

An evolved version of coordinated control. This is what I imagine to happen in the evolution of protocols, language, and culture generally.

I think this does remind us that we need not to ignore the value we get from the bleeding obvious when doing the TCV if we want a fully working model - the anatomy, physiology, dynamics, etc....

I feel as if I am sitting in between Martin and Rick here, which is probably my preferred reference point anyway and might have been the 'higher level reference point' that led me to the above conclusions!

If that is so, since insofar as I understand your comments above I totally agree with you, therefore Rick agrees with both of us, and a Happy New Year on CSGnet will follow!

Hope everyone had a great Christmas!

I would be interested in Martin and Kent's responses to the above.

Warren

I hope you do find mine interesting.

Martin

[From Rick Marken (2016.12.29.1045)]

···

On Thu, Dec 29, 2016 at 2:34 AM, Warren Mansell wmansell@gmail.com wrote:

WM: Hi Rick, I think to critique Martin’s thought experiment, don’t you need to assume from the start that this game is an analogy of a single individual

RM: No. It was pretty clear that Martin was describing a situation involving several people affecting the position of the puck. And all these people were visible to Dee, the person testing to see if the position of the puck was under control. This is not my imagination; it’s what Martin said:

MT: we now introduce a privileged player, who we can call the “Disturber” or “Dee”. Dee can see the players and their pebbles, and can move the Puck, but the players cannot see Dee. Dee wants to determine whether the Puck location is being controlled, and if so, to what reference value.

RM: Since Dee can see the players, the pebbles and the puck I presume Dee can see that the pebbles are hitting and thus likely affecting the position of the puck.

WM: who, we know, will have multiple control systems at lower levels of their hierarchy operating through various physical feedback functions?

RM: We don’t know that. That’s what we do research to find out. All Dee knows is that there are a bunch of people throwing pebbles at a puck (not the most common social situation) and for some reason she suspects that the position of the puck is being controlled. At least, her starting hypothesis if that it is the position of the puck that is being controlled; it might be something else about the puck that is being controlled, such as its size or the sounds made as the pebbles hit it. And she must suspect, given what she can see, that if something about the puck is being controlled, it is very likely that it is the players throwing pebbles who are doing the controlling.

WM: So we are assuming that a TCV experimenter analysing control without anatomy and physics is like Dee doing the TCV without awareness of the many people doing the control?

RM: But Dee was not unaware of the possibility that it was the people doing the controlling, assuming that something about the puck was indeed being controlled.

WM: I think Martin has demonstrated that when we test for a CV we are still aware that there is a cascade of control systems implementing the perceptual control of a higher level system.

RM: Then he demonstrated something that is not true. We suspect, based on theory, that the controlling done by organisms is hierarchically organized; but we aren’t “aware” that this is the case. We are aware only of the possibility that this is the case. And this awareness is irrelevant to the question of whether anything about the puck is actually being controlled. If we find that some aspect of the puck is being controlled then we can start looking to see how it is controlled and why.

WM: I think in tandem it represents that it can look as though there is a higher level system simply because every lower level control unit has the same or a similar reference point.

RM: Sure it can. And we can test to see if this appearance is the case – that the puck is being controlled for some higher level reason – once we’ve established what (if anything) about the puck is being controlled

WM: But I think this might be overstepping the analogy. In this game, there IS a master controller - the leader of the game who the players have chosen to follow and take on board his or her reference point. This higher level control system would be needed within each individual in order to have any ability to update the target point to a new value - it would need to be sent down to these lower level control systems.

RM: Sure, but this can all be found out after it has been established that something about the puck is, indeed, under control.

WM: So my view is that Martin’s thought experiment does remind us that a hypothesised CV does not tell us about the full working, hierarchical architecture of how a perceived aspect of the environment is controlled.

RM: That doesn’t seem like something a competent researcher needs to be reminded of. The TCV has never been presented as something other than a method of determining whether some variable aspect of the environment is under control.

RM: I think what we need are fewer gedanken experiments and more real experiments. And, if not experiments, at least some observations of real phenomena. I’m still waiting to see an example of a real social stability that can be explained by multiple control systems controlling the same perceptual variable relative to the same or different references.

Best

Rick

However, it doesn’t to me explain how a virtual higher level controller seems to exist when it doesn’t - it seems to me that for the people in this thought experiment to even be willing to take part in it and update their targets, they need a higher level control system in addition to the one that Martin proposed to exist to get the puck on the target.

A rethink here : Maybe it is fairer to say it doesn’t rule out the possibility of a higher level controller - this scenario could work if there was a system operating at a broader timescale that led people in a shared environment to adopt similar reference values (I am not sure whether ‘evolution’ is in fact a master controller in functional terms though - probably is?) - an intrinsic tendency to prefer familiarity, or an intrinsic, evolved reference value for the target in question (such as a bee’s in built preference for petal markings that converge on a point in the centre of a flower - where the nectar is). Group living insects certainly appear to have a master controller, and an experimenter who only had data on the nectar levels in flowers and nectar levels in a hive might think a massive creature was shifting the stuff into the hive.

I think this does remind us that we need not to ignore the value we get from the bleeding obvious when doing the TCV if we want a fully working model - the anatomy, physiology, dynamics, etc…

I feel as if I am sitting in between Martin and Rick here, which is probably my preferred reference point anyway and might have been the ‘higher level reference point’ that led me to the above conclusions!

Hope everyone had a great Christmas!

I would be interested in Martin and Kent’s responses to the above.

Warren


Richard S. Marken

“The childhood of the human race is far from over. We
have a long way to go before most people will understand that what they do for
others is just as important to their well-being as what they do for
themselves.” – William T. Powers

On Wed, Dec 28, 2016 at 7:47 PM, Richard Marken rsmarken@gmail.com wrote:

[From Rick Marken (2016.12.28.1145)]


Dr Warren Mansell
Reader in Clinical Psychology

School of Health Sciences
2nd Floor Zochonis Building
University of Manchester
Oxford Road
Manchester M13 9PL
Email: warren.mansell@manchester.ac.uk

Tel: +44 (0) 161 275 8589

Website: http://www.psych-sci.manchester.ac.uk/staff/131406

Advanced notice of a new transdiagnostic therapy manual, authored by Carey, Mansell & Tai - Principles-Based Counselling and Psychotherapy: A Method of Levels Approach

Available Now

Check www.pctweb.org for further information on Perceptual Control Theory

Martin Taylor (2016.12.27.11.19)–

MT: That's a very interesting restriction on the TCV. You imply that

there’s no way to determine that the position of a car in its lane
is being controlled, because neither the driver nor the
Analyst/Experimenter/Observer can know each wind gust, each pothole,
each change of bearing friction, each change in road slope, …

RM: All you have to know is how much the car’s position is expected to vary if it were not under control. We know how unexpected wind gusts, potholes and changes in bearing friction will affect the car’s position; we also know how much the car’s position will vary (due to these and other factors) if the car proceeds down the road sans driver. But if we wanted to do experiments to see what optical variable the driver is actually controlling we would do this under controlled conditions where we are able to vary one specific variable, such as the force of a side-wind, and monitor the effect of this disturbance on the hypothesized controlled variable. This kind of research would probably have to be done in a driving simulator.

MT: Since you accept the possibility that nowhere in the brain is there

a non-virtual perceptual value,

RM: I do? What I “accept” is a theory, called PCT, that explains the existence of observed reference states for controlled variables in terms of the existence of neural reference signals that specify the state of perceptual signals that correspond to variations in those controlled variable. I also accept that there are situations where a variable is observed to be maintained in what turns out to be a virtual reference state, one that results from a conflict between control systems trying to get that variable to match reference signals that specify different states of that variable.

MT: despite the apparent singularity of

conscious perceptions, we can proceed with a serious discussion of
collective control and virtual perceptual signals, applying it where
appropriate and useful in both individual and social PCT research.

RM: I have no idea what a “virtual perceptual signal” is. But perhaps what it is – and why it’s needed – will become clear like in a serious discussion of collective control, a discussion I would very much like to have. I think we should start with you (or Kent) describing an example of a stable social variable and then showing how this variable is maintained in a stable state (a virtual reference state) by collective control.

MT: PS. But I still wonder why you accept only TCV condition 4a and

dismiss conditions 4b and 4c of my list as criteria for determining
whether control is occurring and what variable is controlled, when
you use them in your own demos of the TCV.

RM: I don’t dismiss them. Here are your three conditions (criteria) for determining that a variable is under control:

a. whether the hypothesized property changes appreciably less than would be expected if a corresponding perception were not being controlled, or
b. measuring the influence of the putative controller and determining that it is negatively correlated with the (time-lagged) influence of the experimenter-induced disturbance, or
c. measuring the varying value of the complex property and determining that its variation has a low correlation with variation in the applied disturbance.

RM: All these criteria are fine with me. Criteria a and c are basically the same: both involve looking to see whether variations in a disturbance to a hypothesized controlled variable produces less variance in that variable than expected. Criterion b is certainly a good one but often difficult to use in practice since the researcher’s is often not the only disturbance affecting a hypothesized controlled variable. This is especially true when doing the TCV in real-world situations. I have, indeed, used it in some of my demos because in these situations it is possible to control (in the methodological sense of “hold constant”) all influences on the hypothetical controlled variable other than the experimentally produced disturbance.

RM: I look forward to learning how collective control explains the stability of a social variable.

Best

Rick


Richard S. Marken

“The childhood of the human race is far from over. We
have a long way to go before most people will understand that what they do for
others is just as important to their well-being as what they do for
themselves.” – William T. Powers

            RM: The puck moving back

to the place no matter where Dee moves it has nothing to
do with determining whether or not the location of the
puck is controlled. Before doing the test Dee would have
to know how much the puck would be expected to
move if it were not under control.