KR and reorganization

CSGnet Archive CSG1996
1

[From Rick Marken (960926.0810)]

Jeff Vancouver:

there is no input into the control system that turns the reorganizing
process on or off with knowledge of results

Me:

I think Jeff is saying that the input itself doesn't turn reorganization
on or off; it's the input relative to intended input.

Jeff Vancouver (960925.15:35) --

Here's what I think I meant. I meant "without" not "with."

So what you meant to say was: "there is no input into the control system that
turns the reorganizing process on or off WITHOUT knowledge of results".

Now I have to disagree with that. What you say now suggests that knowledge of
results (KR) has some special significance with respect to reorganization. In
fact, KR is just a poor description of what is simply an aspect of a
controlled percpetual variable. KR has precisely the role in control (and
reorganization) as the target in a pursuit tracking task; it is the
contribution to the controlled perceptual variable that is independent of the
subject's actions. Taking away KR would be like taking away the target in the
tracking task; if this happened there would be no perception to control. This
might start a reorganization process, but that reorganizatoin may be no more
complex than another system noticing that the world has changed so that it is
no longer possible to control a particular perception and removing the goal
of controlling that perception.

So the question that remains (or what distinguishes me from Rick maybe?)
is whether it is true that without KOR the reorganization control unit
does nothing

In my view, KOR (KR) is just one kind of disturbance variable. Reorganization
can happen whether there is KR or not. KR, per se, is no more important to
reorganization than any other kind of disturbance variable. The rate at which
reorganization occurs depends (in theory) on how well one is controlling.
When control is good, the rate of reorganization is low; when control is
poor, the rate of reorganization is high. Disturbances may indirectly
influence the rate of reorganization by influencing how well variables are
being kept under control; but they don't directly start or stop
reorganization.

Best

Rick

2

[from Jeff Vancouver 960926.12:40 EST]

I will tell you one reorganizating unit that has failed, it is mine. I
will never learn to keep out of these discussions. So back in I go.

[From Rick Marken (960926.0810)]

So what you meant to say was: "there is no input into the control system that
turns the reorganizing process on or off WITHOUT knowledge of results".

Yes, that is what I meant to say.

Now I have to disagree with that. What you say now suggests that knowledge of
results (KR) has some special significance with respect to reorganization.

Not necessarily, but in the case that Bruce Abbott brought up, yes I am
very much saying that.

In
fact, KR is just a poor description of what is simply an aspect of a
controlled percpetual variable. KR has precisely the role in control (and
reorganization) as the target in a pursuit tracking task; it is the
contribution to the controlled perceptual variable that is independent of the
subject's actions. Taking away KR would be like taking away the target in the
tracking task; if this happened there would be no perception to control.

Yes, hence KR can be very important for its availability will help
determine whether control is possible. That is a significant issue,
which I have been trying to communicate to the people on this net (and
why I got involved in this discussion).

This
might start a reorganization process, but that reorganizatoin may be no more
complex than another system noticing that the world has changed so that it is
no longer possible to control a particular perception and removing the goal
of controlling that perception.

Well this may be true (although it seems unlikely given Bruce's scenerio).
What is more at issue is that KR opens a channel of communication, the
signal in that channel becomes the stimuli ("correct"/"incorrect") that
some input function translates to a perception (say 1 or 0 respectively).
Now the reference signal for this system is presumable 1 (if our learner
was a teenager and the experimenter was a parent, we might presume the
reference signal is 0 - that is, "I will stick with whatever my parent
does not want me to do."). Anyway, the output is to reorganize the input
function for the system whose goal it is to please the experimenter,
which in this case means distinguishing the signals.

In my view, KOR (KR) is just one kind of disturbance variable. Reorganization
can happen whether there is KR or not. KR, per se, is no more important to
reorganization than any other kind of disturbance variable. The rate at which
reorganization occurs depends (in theory) on how well one is controlling.
When control is good, the rate of reorganization is low; when control is
poor, the rate of reorganization is high.

I have no problem with the last half of this paragraph. BUT, in the
scenerio Bruce described, how does the system know whether control is good
or not? Answer, the value of 1 or 0 which comes from the
"correct"/"incorrect" signal which is only available when there is KR!

To reiterate, the system must have some information to know "how well one
is controlling." That information comes from observing one's results.
We (in psychology) call that KNOWLEDGE OF RESULTS.

Disturbances may indirectly
influence the rate of reorganization by influencing how well variables are
being kept under control; but they don't directly start or stop
reorganization.

Here, is the crux of the confusion, I think. In this case KR should not
be thought of as a disturbance. It is a channel so that the state of q
can be observed. If it is a disturbance, it would be to the arrow from q
to s(), not to q directly.

By the way, I think Bruce Gregory's observations were absolutely correct.
Bill P.'s response about substituting "error" for KR obscurred the issue.
Although technically correct, it is only the signal from the experimenter
that will be translated into a perception and compared to a reference
signal. Thus error may only eminate from the experimenter's message,
unless we are to assume that the system uses signals from other sources to
create a perception. IN THE CASE, that does not seem likely

Later

Jeff