Alerting, Control

[From Rick Marken (940717.1400)]

Bill Leach (940717.12:12) --

Ithink you make some excellent points; my comments are meant as
(probably superfluous) expansions on your points; not as disagreement.

The whole concept of "Alerting" probably stems from two considerations:

The first is that we "experience" such things.
The second is that we "observe" such behaviour in others.

Yes. We experience it, observe it in others and call it "alerting". But it's
just disturbance resistance because there must be a reference for a certain
state of a perceptual variable or there is nothing "alerting" about any
particular value of that variable.

I rather vividly remember my reaction upon hearing "flooding in the
engine room" come over the announcing system. This would seem to me
to have fit the description of an "Alerting" phenomenon.

I agree. "Flooding in the engine room" is "alerting", however, only because
those words describe a perception of the engine room that deviates from
_your_ reference for the state of the engine room. The exact same words
would _not_ be alerting if the your reference specified that the state of the
engine room be "flooded" (an unlikely reference to have, I admit, but
still possible; perhaps when the engine is on fire and you want to flood the
room to put it out?).

My argument with Martin about "alerting" is motivated (on my side) by
the following:

Saying that some perception is "alerting" implies that some states of
perceptual variables are always, _absolutely_ alerting. It is easy to show
that this is not true. All I am saying is that people, by virtue of the
setting of their reference signals, determine what states of perceptual
variables will be "alerting". If that's what Martin means by "alerting",
then I have no problem; but then there is no need for any special concept
of "alerting" in PCT; "alerting" is just disturbance resistence.

The sensation of "Alerting" in this example, I think, is due to the fact
that a perception (really many) that was controlled (possibly without any
actual physical action necessary) suddenly perceived to be in a state
that represented a large error.

Yes. I don't think we are aware of many of the variables we are controlling at
any time; and when perceptions are matching their reference states there is
not much to be aware of; things just seem alright. When one of these
variables is suddenly pushed from its reference state, it seems like something
_bad_ happened; the perception itself seems bad or "alerting". If my monitor
suddenly blew up it would seem like a bad thing happened; but all that
happened was that a monitor blew up; it seems bad to me because it pushes a
lot of perceptual variables away from my references for them.

We have a tough time noticing that we (via our references) determine that
any particualr perception is _bad_ or "alerting". PCT puts people, US,
right smack dab in the center of things; the world (and what we see people
doing in it-- human behavior) just ARE; they are just perceptions. Each
individual determines the "value" (good or bad, alerting or non-alerting) of
any perception. In PCT, "Man is the measure of all things" with a vengence;
the "things" are perceptions and the "measure" is the setting of our own
reference signals.

The "Alerting function" then could be "real" in the sense that it is a
lable for a control system operation under certain conditions - that of
sudden large error.

No problem. A sudden error signal implies that a perception suddenly
deviated from a reference signal. I think the term "alerting" (if it
must be used -- I think it's misleading) would be limited, also, to the
case where the sudden error was created by a change in the state of the
perception (due to disturbance) rather than by a change in the state of
the reference signal.

Bill Leach (940717.11:27 EST)

Rick; I think that this statement could lead to some confusion. What
you are saying is (I believe absolutely correct) but it is predicated
upon the simple idea that "postion of door" is all that is being
controlled and I suspect such is rarely the case.

Agreed. My point was just that the the only variable that is EVER controlled
in a control loop is the perceptual variable. When we control the position of
the door by controlling the force exerted, the force is a perceptual variable
controlled in another, lower level loop. In a hierarchy of control loops, it
is always the case that in any loop and in the hierarchy in general, all that
is (or can be) controlled is perceptual variables (signals). In the door
operning example, the system controlling position of the door achieves this
_perceptual_ goal by setting the reference for lower order force
_perceptions_; these force perceptions are achieved by setting the
reference for even lower order muscle fiber length and tension _perceptions_.
At every level of control, what is controlled is a perception; not an output.

Mary Powers (940717) --

There really are no baby steps to take between behavior as
outcome, consequence or result, and behavior as the control of
perception. One of the big difficulties PCT has in making its way
in the world is that you can't shape the understanding of it
incrementally. Either you force PCT data to fit your world view,
or you make the jump, and the world never looks quite the same
again. Nothing in between.

Unfortunately, truer words were never spoken.

Nice post Mary.

Best

Rick

[Martin Taylor 940718 18:50]

Rick Marken (940717.1400)

Bill Leach (940717.12:12)

The whole concept of "Alerting" probably stems from two considerations:

The first is that we "experience" such things.
The second is that we "observe" such behaviour in others.

Neither of them is where MY concept of Alerting comes from. It comes
from the requirement that Bill P. noted as "multiplexing." You simply
don't have as many output degrees of freedom by a long shot as you have
perceptual degrees of freedom. Uncontrolled perceptions are simply
uncontrolled perceptions. That is to say that when they change value,
NOTHING HAPPENS. Why, then, should we even have perceptual functions
to generate those perceptions? They must sometimes be controlled.
When? And what in the hierarchy effects the change of control from
those perceptions for which something happens when they change value
to those for which nothing was happening, but which have (perhaps) reached
values that might prove fatal for the orgnaism if they remained uncontrolled?

The question is: What is the control of multiplexing? Is "multiplexing"
a perception :wink: I seriously doubt it.

Saying that some perception is "alerting" implies that some states of
perceptual variables are always, _absolutely_ alerting.

Implication not intended. It seems probable that evolution has set us up
so that there are some such states, but that isn't a necessary implication.

All I am saying is that people, by virtue of the
setting of their reference signals, determine what states of perceptual
variables will be "alerting".

I'm happy with that. But what you seem to go on to is that the perception
to which control is switched is the alerting perception. Normally, it
isn't. The newly controlled perception is something else entirely, in
most cases. Of course, the alerting perception and the newly controlled
perception might coincidentally be the same, but that doesn't happen often.

What I do agree with, from you and Bill P., is that the alerting stimulus
may often (and perhaps always) be an uncontrolled component perception of
a controlled higher-level perception, and that disturbances to that higher
level perception cause error at that level, and from that error generate
output that causes the switch in control. The mechanism is unclear, but
the linkage seems plausible (only plausible, not necessary).

The sensation of "Alerting" in this example, I think, is due to the fact
that a perception (really many) that was controlled (possibly without any
actual physical action necessary) suddenly perceived to be in a state
that represented a large error.

If it was being controlled, it wouldn't be in a state of large error,
would it? Do you mean that a large, sudden, disturbance occurred, beyond
the range of control? I don't see how that relates to the concept of
alerting.

We have a tough time noticing that we (via our references) determine that
any particualr perception is _bad_ or "alerting".

Why do you associate "alerting" with "bad?" A tinkling bell might well be
an alerting signal that I can soon bring my perception of eating an ice cream
to its reference level, when for a long time that has not been a controllable
perception (at least not without conflict).

Martin

<[Bill Leach 940719.01:45 EST(EDT)]

NET

Yike! I believe that I said something like: "... why a perception can
not be a part of a control loop that is not generating an error
signal..." What I should have said is "an error signal other than zero".

As far as I understand, there is no reason why a control loop can not
have an actual absolute error value of zero. The loop then would be
"inactive" as far as any behaviour controlling the perception UNLESS the
error goes from zero to some positive value.

I am quite sure that there are many control loop that are normally in
this situation though not necessarily ones that we would think of as
having possibly causing "alerting" sensations. I believe almost all
motor control loops have "complementary" muscle groups and other than in
the case of isometric exercise only one control loop would have an error
signal that would attempt to exceed zero, the other would be zero (and
would be negative if such were possible).

-bill

<[Bill Leach 940719.01:04 EST(EDT)]

[Martin Taylor 940718 18:50]

Bill Leach (940717.12:12)

The whole concept of "Alerting" probably stems from two considerations:
The first is that we "experience" such things.
The second is that we "observe" such behaviour in others.

Neither of them is where MY concept of Alerting comes from. ...

OK, my mistake.

I am still not sure that I understand why a perception that is not
generating an error signal must necessarily not be a part of an active
control loop. In other words, is it really unreasonable to posit that
such perceptions that we refer to as being uncontrolled might instead be
controlled with a reference value that in essence "sets limits" on
variations for the perception?

If this were then the case, there would be no need for "multiplexing" in
the sense that the term is usually applied (some sort of on/off
switching). The "uncontrolled" perceptions then would "automagically"
begin controlling if they deviated enough to exceed their reference
value. I envision that the error signal for such perceptions that we
might typically consider as being important (alerting) then would be
rather high priority signals and have the capability of over-riding other
"more mudane" control loops. I think that this also satifies your next
consideration.

I'm happy with that. But what you seem to go on to is that the
perception to which control is switched is the alerting perception.
Normally, it isn't. The newly controlled perception is something else
entirely, in most cases. ...

I am having a little trouble with this comment. What do you mean by,
"... switch is the alerting perception. Normally, it isn't. The new ...

I think that where I am having difficulty is that I do not see how when a
change in the value of a perceptual signal has "triggered" some action
(or is just being controlled) can possibly not activate control action
that is related to the same perception.

I believe that you gave the example of movement in the peripheral vision
as an example. The input signal is movement in the peripheral vision but
the Perception is that there might be danger. Of course the being would
not be controlling movement in the peripheral vision. The movement might
exceed a reference for no (or some minimal movement) and then the control
would be to reduce danger (or at least verify that there is no danger).

The mechanism is unclear, but the linkage seems plausible (only
plausible, not necessary).

I admit that I will for the moment defer to the "experts" on this but I
don't personally see that any "switching" is necessary at all.

The sensation of "Alerting" in this example, I think, is due to the
fact that a perception (really many) that was controlled (possibly
without any actual physical action necessary) suddenly perceived to be
in a state that represented a large error.

If it was being controlled, it wouldn't be in a state of large error,
would it? Do you mean that a large, sudden, disturbance occurred,
beyond the range of control? I don't see how that relates to the
concept of alerting.

Yes, I specifically mean that the error suddenly increased to a large
value. I am not trying to make any claim as to how but I do think that
there may be several different ways that this could happen. I would also
include control loops that are operating in continous close control in
this description as well as those for which the error signal is zero
without active behaviour expressly for that perception required.

I think that it might well be very closely related to the concept of
"alerting" in that errors significantly greater than zero are an
"abnormal" situation. Even if not an intrinsic, anytime and error value
exceeds the minor magnitude required by less than infinity loop gain, it
is not unreasonable to consider that such would result in a "sensation"
or "feeling" (and we may well have labeled such as alerting).

Rick

We have a tough time noticing that we (via our references) determine
that any particualr perception is _bad_ or "alerting".

Why do you associate "alerting" with "bad?" A tinkling bell might well
be an alerting signal that I can soon bring my perception of eating an
ice cream to its reference level, ...

Now here is one place where I seem to be agreeing with you as opposed to
Rick (Horrors!). I would be willing to suggest that such a mechanism
probably evolved as a survival mechanism (natural selection) but it
certainly does not have to function only in that manner. A pleasent
surprise would be just as "alerting" as many negative ones.

-bill