Control by proxy

CSGnet Archive CSG1998
1

[From Bruce Abbott (980915.0845 EST)]

Bill Powers (980915.0646 MDT) --

So you are asserting that the Department of Transportation (or whatever
agency stands for "the state") does not need to perceive ANYTHING about
"motorists" in order to control how fast they drive. Could you please
diagram how The State does this? I really don't see how it could.

As the risk of repeating myself, the individuals who represent the state
have designed and implemented a control system and established its reference
values (the posted speed limits). "They" are controlling the speed of
motorists in the same sense that "you" are controlling the temperature of
your house when you turn on its temperature control system and set the
reference temperature. You and they establish control over the relevant
variable by proxy -- by implementing a control system to do the job for you.
Of course, it is the proxy system that is "really" doing the controlling. I
leave up to you whether to call control-by-proxy "control" or something
else. You (or the state's representatives) get a variable under control
whether you do the controlling yourself or implement a system to do the
controlling for you.

Of course, you -- and they -- can monitor the system and take further action
if it doesn't seem to be doing its job, in which case you -- and they -- are
directly controlling (or at least attempting to control) the effectiveness
of the proxy system.

Have we said enough on this?

Regards,

Bruce

2

[From Rick Marken (980915.0730)]

Bill Powers (980915.0646 MDT) --

So you are asserting that the Department of Transportation
(or whatever agency stands for "the state") does not need to
perceive ANYTHING about "motorists" in order to control how
fast they drive. Could you please diagram how The State does
this? I really don't see how it could.

Bruce Abbott (980915.0845 EST) --

As the risk of repeating myself, the individuals who represent
the state

Who is "the state"? I think "the state" IS the individuals who
agree thet they are controlling for a perception of "the state".
I don't think indiduals _represent_ a state; this implies that
there is a real thing "out there" called "the state". I think
what's actually "out there" ia individuals controlling for a
relatively similar system concept perception; a perception that
is a complex function of inputs that are perceived aspects of
the behavior of other individuals; we call this perception
"the state".

have designed and implemented a control system and established
its reference values (the posted speed limits).

I think the reference values for speed exist only in the heads of
the individuals who enforce the speed limits; signs can't control
speed; the number written on these signs doesn't function as a
reference value, unless it is read by an individual control system
and incorporated into that individual's speed control system as
its reference for maximum speed.

"They" are controlling the speed of motorists in the same sense
that "you" are controlling the temperature of your house when
you turn on its temperature control system and set the
reference temperature.

Some individuals (the "proxies") are controlling for the speed
at which other individuals travel.

You and they establish control over the relevant variable by
proxy -- by implementing a control system to do the job for you.

I think the people who do this are not really controlling a
perception of speed; they are controlling (by proxy) some
other perception, like "road safety" (which they can perceive
as statistical data on accidents) or "law and order" (which they
can perceive as the number of tickets issued per month) or
whatever.

Of course, it is the proxy system that is "really" doing the
controlling.

In this case the individual proxies (police) control the speed
of the cars. I think the people who hire these police are also
controlling (by proxy), but they are controlling variables other
than those (such as your speed) controlled by the police.

I leave up to you whether to call control-by-proxy "control" or
something else.

I agree that it's control; I just think the people controlling
by proxy are controlling perceptions other than those controlled
by the proxies.

Have we said enough on this?

I think it's a very interesting subject. I haven't looked at
the inventory model but I think we may be getting more of this
if Bill starts implementing a PCT model of the individuals who
are the components of the inventory model.

Best

Rick

ยทยทยท

--
Richard S. Marken Phone or Fax: 310 474-0313
Life Learning Associates e-mail: rmarken@earthlink.net
http://home.earthlink.net/~rmarken

3

[From Bill Powers (980915.0858 MDT)]

Bruce Abbott (980915.0845 EST)--

So you are asserting that the Department of Transportation (or whatever
agency stands for "the state") does not need to perceive ANYTHING about
"motorists" in order to control how fast they drive. Could you please
diagram how The State does this? I really don't see how it could.

As the risk of repeating myself, the individuals who represent the state
have designed and implemented a control system and established its reference
values (the posted speed limits). "They" are controlling the speed of
motorists in the same sense that "you" are controlling the temperature of
your house when you turn on its temperature control system and set the
reference temperature. You and they establish control over the relevant
variable by proxy -- by implementing a control system to do the job for you.

What sense is that? When I adjust the reference temperature on a
thermostat, I do so in order to control what I am perceiving about room
temperature, not what the thermostat is perceiving (I don't even know where
its movable contact is). And all I _can_ control in that way is how warm I
feel. If I could not sense the room temperature myself, I couldn't control
it at all by adjusting the thermostat. I could certainly VARY it, but I
couldn't control it.

Bruce, all I'm doing is sticking to the technical definition of control. In
order for a system to control some variable, it must sense that very same
variable. There is no such thing as "controlling by proxy." If you're not
sensing the supposed controlled variable, you're not controlling it,
either, no matter how many other people you tell to do it for you. You
would be totally unable to oppose a disturbance that altered the controlled
variable -- for example, your "proxy troopers" deciding to let priests off
with a warning.

Of course, it is the proxy system that is "really" doing the controlling. I
leave up to you whether to call control-by-proxy "control" or something
else.

I want you to call it something else. Unless the DOT is monitoring your
driving behavior, it has no way to know whether issuing orders, open-loop,
to the troopers is having any effect on your driving behavior or any
motorists driving behavior, even once. Whatever the relation between the
DOT and any driver is, it is not control. At best, your role, should you
happen to be cited, is to have a drop-in-the-bucket effect on the DOT's
perception of the statistics concerning speeding incidents and tickets
issued. In THAT loop no one driver plays more than a miniscule part.

You (or the state's representatives) get a variable under control
whether you do the controlling yourself or implement a system to do the
controlling for you.

The "state's representatives" are the ONLY control systems acting in this
situation.

Of course, you -- and they -- can monitor the system and take further action
if it doesn't seem to be doing its job, in which case you -- and they -- are
directly controlling (or at least attempting to control) the effectiveness
of the proxy system.

Not if you still think you're right.

Best,

Bill P.

4

[From Bruce Abbott (980915.1045 EST)]

Rick Marken (980915.0730) --

Bruce Abbott (980915.0845 EST)

As the risk of repeating myself, the individuals who represent
the state

Who is "the state"? I think "the state" IS the individuals who
agree thet they are controlling for a perception of "the state".

I am not going to get into a fruitless argument with you over definitions.
I would call the "state" a particular organization of people, and its
"representatives" those who act officially in the name of that organization.
(I rather doubt that many of them are controlling for a perception of the
state.) But either definition will work without affecting my conclusion.

have designed and implemented a control system and established
its reference values (the posted speed limits).

I think the reference values for speed exist only in the heads of
the individuals who enforce the speed limits; signs can't control
speed; the number written on these signs doesn't function as a
reference value, unless it is read by an individual control system
and incorporated into that individual's speed control system as
its reference for maximum speed.

The troopers (and if the system works, the motorists) perceive the sign and
set their own references accordingly.

"They" are controlling the speed of motorists in the same sense
that "you" are controlling the temperature of your house when
you turn on its temperature control system and set the
reference temperature.

Some individuals (the "proxies") are controlling for the speed
at which other individuals travel.

Maybe. Or maybe they are just controlling for giving tickets to speeders.

You and they establish control over the relevant variable by
proxy -- by implementing a control system to do the job for you.

I think the people who do this are not really controlling a
perception of speed; they are controlling (by proxy) some
other perception, like "road safety" (which they can perceive
as statistical data on accidents) or "law and order" (which they
can perceive as the number of tickets issued per month) or
whatever.

Fine. I addressed this possibility in my post, so we agree here.

Regards,

Bruce

5

[From Bruce Abbott (980915.1200 EST)]

Bill Powers (980915.0858 MDT) --

I think we agree about the basics and are simply arguing about what to call
something.

You and they establish control over the relevant

variable by proxy -- by implementing a control system to do the job for you.

What sense is that? When I adjust the reference temperature on a
thermostat, I do so in order to control what I am perceiving about room
temperature, not what the thermostat is perceiving (I don't even know where
its movable contact is). And all I _can_ control in that way is how warm I
feel. If I could not sense the room temperature myself, I couldn't control
it at all by adjusting the thermostat. I could certainly VARY it, but I
couldn't control it.

I tried to take that into account in my description by suggesting that you
might not want to call it "control." If I set the thermostat and leave the
room, I have no way of knowing whether the room temperature is staying where
I want it; I have to trust that the room-temperature control system is doing
its job.

I understand perfectly well the distinction you make here; that's why I said
in my post that only the heating system is "really" controlling the room
temperature. Yet by turning over control to another system, I get the
result I intend to get (so long as the system functions properly), without
having to actively monitor the room's temperature and take any necessary
corrective action myself. The variable I desire to be controlled _is_
controlled, even if I am not personally doing the controlling. Meanwhile, I
will probably monitor the proxy system and take action if it seems to be
failing to do its job, and that's real control, too.

So when we say that "the state" is "controlling" motorists' speeds on the
highway, we mean that officials acting in the name of the state have
designed and implemented control systems that do the actual controlling in
the state's name. I called that "control by proxy" -- the delegation of
control to another system -- but the delegating system is not controlling by
the formal definition of that term unless it is perceiving the effectiveness
of its proxy systems and taking action to correct deviations from the
desired result.

Regards,

Bruce

6

[From Bruce Gregory 9980915.1322 EDT)]

Bruce Abbott (980915.1200 EST)

I understand perfectly well the distinction you make here; that's
why I said
in my post that only the heating system is "really" controlling the room
temperature. Yet by turning over control to another system, I get the
result I intend to get (so long as the system functions properly), without
having to actively monitor the room's temperature and take any necessary
corrective action myself. The variable I desire to be controlled _is_
controlled, even if I am not personally doing the controlling.
Meanwhile, I
will probably monitor the proxy system and take action if it seems to be
failing to do its job, and that's real control, too.

This seems to me to be analogous to the situation when you engage the cruise
control of your car. You control the process of engagement but leave the
controlling of the speed to it. You monitor its performance from time to
time, but rely on it to do the controlling.

Bruce Gregory

7

[From Bruce Abbott (980915.1235 EST)]

Bruce Gregory 9980915.1322 EDT --

This seems to me to be analogous to the situation when you engage the cruise
control of your car. You control the process of engagement but leave the
controlling of the speed to it. You monitor its performance from time to
time, but rely on it to do the controlling.

Yes, nice example.

Bruce A.

8

[From Bill Powers (980915.1134 MDT)]

Bruce Gregory 9980915.1322 EDT)--

Replying to Abbott:

This seems to me to be analogous to the situation when you engage the cruise
control of your car. You control the process of engagement but leave the
controlling of the speed to it. You monitor its performance from time to
time, but rely on it to do the controlling.

Yes, and while you have delegated control to the trooper or the cruise
control, you are not controlling the controlled variable.

Look, folks, you have to make up your minds. In trying to understand
situations involving control, what concepts do you want to use? The
technical ones of PCT, or the concepts you acquired along with natural
language? If you're not going to stick with the literal technical
definitions of PCT, then there's no point in my continuing these discussions.

A system is said to be in control of a variable if

1. Its actions affect that variable.

2. It perceives that variable and compares its perception with a reference
signal.

3. The resulting error signal is the basis for the actions, in the negative
feedback sense.

If any of those conditions is not met, the system is not in control of the
variable.

If you don't want to accept that definition, and that conclusion, then
you're playing some other kind of game.

Best,

Bill P.

9

[From Bill Powers (980915.1134 MDT)]

Bruce Gregory 9980915.1322 EDT)--

Replying to Abbott:

>This seems to me to be analogous to the situation when you
engage the cruise
>control of your car. You control the process of engagement but leave the
>controlling of the speed to it. You monitor its performance from time to
>time, but rely on it to do the controlling.

Yes, and while you have delegated control to the trooper or the cruise
control, you are not controlling the controlled variable.

Sorry, I never meant to imply that you were controlling the controlled
variable "speed". In fact, I said "you control the process of engagement but
leave controlling of the speed to it." I thought that was clear. I was
obviously wrong. It's a little disconcerting to be so frequently
misunderstood. I'm not sure what to do to be any clearer. Perhaps silence is
the only alternative.

Bruce

10

[From Bruce Abbott (980915.1325 EST)]

Bill Powers (980915.1134 MDT) --

Bruce Gregory 9980915.1322 EDT)

This seems to me to be analogous to the situation when you engage the cruise
control of your car. You control the process of engagement but leave the
controlling of the speed to it. You monitor its performance from time to
time, but rely on it to do the controlling.

Yes, and while you have delegated control to the trooper or the cruise
control, you are not controlling the controlled variable.

Nobody is disputing that, Bill.

Look, folks, you have to make up your minds. In trying to understand
situations involving control, what concepts do you want to use? The
technical ones of PCT, or the concepts you acquired along with natural
language? If you're not going to stick with the literal technical
definitions of PCT, then there's no point in my continuing these discussions.

I didn't see that Bruce Gregory's description violated the technical
definition of control. As you seem to think that it did, perhaps you should
point out where.

With respect to my own statements, I took pains to clarify my position in
the post to which Bruce G. was replying. There, I agreed that control by
proxy done open loop does not meet the formal definition of control. In
control by proxy, the proxy does the controlling (literally, control _by_
proxy). By creating a properly functioning proxy control system, I (through
my proxy) cause a variable to be brought under control, and to the reference
level I choose, but I do not control the variable (the proxy does).

If you're going to rely on your imagination rather than what people actually
say in their posts, then I must regretfully agree: there's not much point
in your continuing these discussions.

Regards,

Bruce

11

[From Bill Powers (980916.0651 MDT)]

From Bruce Abbott (980915.1325 EST)]

Yes, and while you have delegated control to the trooper or the cruise
control, you are not controlling the controlled variable.

Nobody is disputing that, Bill.

I didn't see that Bruce Gregory's description violated the technical
definition of control. As you seem to think that it did, perhaps you should
point out where.

Bruce said:

You control the process of engagement but leave the
controlling of the speed to it. You monitor its performance from time to
time, but rely on it to do the controlling.

And I said, replying to both Bruces at once:

Yes, and while you have delegated control to the trooper or the cruise
control, you are not controlling the controlled variable.

With respect to my own statements, I took pains to clarify my position in
the post to which Bruce G. was replying. There, I agreed that control by
proxy done open loop does not meet the formal definition of control. In
control by proxy, the proxy does the controlling (literally, control _by_
proxy). By creating a properly functioning proxy control system, I (through
my proxy) cause a variable to be brought under control, and to the reference
level I choose, but I do not control the variable (the proxy does).

What other definition of control than the formal one are you proposing that
we allow into this discussion?

How do you (the DOT) know your "proxy" (the trooper) brings the variable to
the reference level you choose without your observing the variable and
comparing it with your reference level? In Bruce G's example, how do you
know that the cruise control is bringing the speed of the car to the
reference speed you have chosen, without your perceiving the speed and
comparing it with your reference level?

In the case of the cruise control, the answer is easy: you DO perceive the
speed, and when you havemade it match your reference level you press the
"Set" button to store the cruise control's speed perception as the cruise
control's reference level. You control the speed at first, while the cruise
control is not controlling it. Then you cease to control the speed, and the
cruise control takes over. Most people, I suspect, then check the
speedometer to see that the cruise control caught at the right speed.

In the case of the DOT controlling your speed, some person in the DOT
issues a directive to the state police defining the maximum speed. That is
an open-loop action, not a control process. The troopers "give 'em five,
let 'em take ten, and nail 'em on fifteen" (actual quote from a state
trooper to me). So the actual speed at which arrests take place is
different from the official speed limit. If the loop is ever closed, it
would be closed (perceptually) in the form of statistics about arrests and
the reported speeds of "traffic," not individual cars. Whoever is charged
with monitoring the effects of the arrest policies would compare those
statistics with target values, and convert any notable errors into new
directives, or recommendations to other people who can issue directives, to
the state police. By this means, people in the DOT can control the average
speed of traffic, although not very well and not at the speed they think
they've specified. They can't control the speed of an individual car at all.

The basic rule of control tbeory is that a control system controls what it
perceives and _nothing else._ It can affect, influence, alter, perturb, or
disturb other variables, but it can't control them unless it can perceive
them. If the driver of the car couldn't perceive the speedometer or some
other indication of the car's speed, he or she couldn't set the cruise
control's reference speed to any particular value, and know that that speed
was being maintained. Even if the driver could type a speed in on a
keyboard, the driver couldn't control the speed actually being maintained
by the cruise control. If the speed typed in was "65", the actual speed
might be 63 or 67, and the driver would never know of the error unless
there was a way to perceive the same variable that the cruise control is
perceiving.

If you're going to rely on your imagination rather than what people actually
say in their posts, then I must regretfully agree: there's not much point
in your continuing these discussions.

I'm not objecting to what I imagine they say, but to what they do say. You
said that traffic statistics and perceptions by the DOT had nothing to do
with the process you were describing, control by proxy as you came to call
it. You were wrong -- even control by proxy is impossible if you can't
perceive what the proxy is supposed to be controlling. The DOT, if it is a
control system, can control ONLY what it can perceive -- by proxy or by any
other means. There is no way to get around that.

Best,

Bill P.

12

[From Bruce Abbott (980916.1015 EST)]

Bill Powers (980916.0651 MDT) --

If you're going to rely on your imagination rather than what people actually
say in their posts, then I must regretfully agree: there's not much point
in your continuing these discussions.

I'm not objecting to what I imagine they say, but to what they do say. You
said that traffic statistics and perceptions by the DOT had nothing to do
with the process you were describing, control by proxy as you came to call
it. You were wrong -- even control by proxy is impossible if you can't
perceive what the proxy is supposed to be controlling. The DOT, if it is a
control system, can control ONLY what it can perceive -- by proxy or by any
other means. There is no way to get around that.

Bill, as far as I can tell, we agree completely on how what I have called
"control-by-proxy" works. I have acknowledged -- from the beginning -- that
the only real control in the open-loop version is done by the proxy. I have
also described a closed-loop version of the process, in which real
controlling is done by both the delegating system and the proxy. I don't
know what else to say that will convince you that this is and has been my
position. Our models of the process appear to be identical, so as far as I
am concerned, we are in agreement.

You seem to be far more interested in "proving" me wrong (and youself right)
than in thinking about the implications of the system I described. How boring.

Regards,

Bruce

13

[From Bill Powers (980916.1151 MDT)]

Bruce Abbott (980916.1015 EST)]

Bill, as far as I can tell, we agree completely on how what I have called
"control-by-proxy" works. I have acknowledged -- from the beginning -- that
the only real control in the open-loop version is done by the proxy. I have
also described a closed-loop version of the process, in which real
controlling is done by both the delegating system and the proxy. I don't
know what else to say that will convince you that this is and has been my
position. Our models of the process appear to be identical, so as far as I
am concerned, we are in agreement.

If you will agree that the two systems are controlling _different
variables_, I think we can put this to rest. So far you haven't come right
out and said that the "proxy" is controlling one variable (a motorist's
speed) and the DOT is controlling, by sending reference signals to the
proxy, the average reported speed of motorists or some other such variable
that is possible for a person in the DOT to perceive. Do you agree now that
it is possible for the trooper, but not the DOT, to be controlling the
motorist's speed?

Best,

Bill P.

14

[From Bruce Abbott (980916.1455 EST)]

Bill Powers (980916.1151 MDT)

If you will agree that the two systems are controlling _different
variables_, I think we can put this to rest. So far you haven't come right
out and said that the "proxy" is controlling one variable (a motorist's
speed) and the DOT is controlling, by sending reference signals to the
proxy, the average reported speed of motorists or some other such variable
that is possible for a person in the DOT to perceive. Do you agree now that
it is possible for the trooper, but not the DOT, to be controlling the
motorist's speed?

Oh, man, not the trooper again. (;->

Yes, of course. The DOT is establishing a mechanism to control the
motorist's speed, by implementing the control system involving the trooper,
but the trooper is doing the controlling of the motorist's speed, not the DOT.

Now, what's happening with the Eberlein thread? (:->

Regards,

Bruce

15

[From Bruce Gregory (980916.1600 EDT)]

Bill Powers (980916.1151 MDT)

If you will agree that the two systems are controlling _different
variables_, I think we can put this to rest. So far you haven't come right
out and said that the "proxy" is controlling one variable (a motorist's
speed) and the DOT is controlling, by sending reference signals to the
proxy, the average reported speed of motorists or some other such variable
that is possible for a person in the DOT to perceive. Do you
agree now that
it is possible for the trooper, but not the DOT, to be controlling the
motorist's speed?

At the risk of being excommunicated, I suggest that society works by
arrangements in which the citizens do _not_ directly control their
perceptions, but instead control a "proxy" perception. As a citizen, I do
not want to encounter drunken drivers on the road. But rather than
controlling my perception that the drivers I encounter are sober (obviously
drunken drivers are disturbances to this controlled perception), I delegate
control of drunken drivers to the state government. I carry out similar
delegations when it comes to honest advertising, water purity, etc. If I
perceive that these proxies are not doing their job, I vote them out of
office and try with a new set.

In a modern society, the citizenry cannot control all the perceptions that
are important to it, so they set up proxies who do control these
perceptions.

Bruce Gregory

16

[From Bill Powers (980916.1916 MDT)]

Bruce Abbott (980916.1455 EST)--

If you will agree that the two systems are controlling _different
variables_, I think we can put this to rest. So far you haven't come right
out and said that the "proxy" is controlling one variable (a motorist's
speed) and the DOT is controlling, by sending reference signals to the
proxy, the average reported speed of motorists or some other such variable
that is possible for a person in the DOT to perceive. Do you agree now that
it is possible for the trooper, but not the DOT, to be controlling the
motorist's speed?

Oh, man, not the trooper again. (;->

Yes, of course. The DOT is establishing a mechanism to control the
motorist's speed, by implementing the control system involving the trooper,
but the trooper is doing the controlling of the motorist's speed, not the
DOT.

I still hesitate to give that a pass. The DOT's action consists of issuing
orders to the state police through some channels; it can vary those orders,
but varying those orders is not the same as controlling. The DOT (or more
properly the legislature) doesn't exactly establish the mechanisms of law
enforcement; those are already there. It takes advantage of them by
providing reference signals.

I want some way to describe this situation that doesn't make it seem that
the legislature is controlling at the same level as the state troopers. In
Bruce G's example of the cruise control, the driver initially controls
speed at the same level as the cruise control, but this is not how the DOT
and the troopers are related. The DOT does not go out arresting people,
and then push a button that says to the troopers, "There, keep it that
way." The picture of the traffic situation that the legislators have is
very general, averaged over many cases and reflecting no individual
motorist's behavior. So there is simply no way that the legislature can
control at the level of individual behavior, or know enough to design a
control system that would work at that level. The dealings with individual
motorists have to be handled at the level where the state police interact
with them one at a time, every case involving special circumstances and
judgement calls. All the legislators can tell the troopers is what the
legal limit is. But that is not the same thing as designing the control
system that is embodied in the troopers. It is not the same thing as
ensuring that every motorist will be forced to comply with the speed laws.

Well, we've obviously gone asymptotic on this subject, so your last remark
is apropos: What about Bob Eberlein's project? I'm thinking about it, but
not very hard at the moment. A couple of ideas, not ready for the light of
day yet. I have to go cut down more dead trees, and since it takes me about
three or four days to dispose of a tree (with suitable intervals for
panting and lying down) I am in no danger of having nothing to do for quite
a while. At some point I expect to be able to walk up the hill behind my
house without having to stop for breath. Getting over the effects of
smoking 2+ packs a day for 57 years is taking a while.

Best,

Bill P.