[From Rick Marken (960725.1500)]
Martin Taylor (960725 14:10) --
When one controls a perception, one acts on the world, and some of the
effects of that action affect the perception in question. That's "feedback".
Right. A variable (perception in this case) has an effect on itself via the
closed loop of causality.
When someone says "Allow me to give you some feedback," they normally do
so because one effect of your action has been to disturb a controlled
perception of theirs. That effect may have been a pure side-effect of your
own perceptual control, but nevertheless their "giving" you feedback is
part of the environmental feedback path between your action and your
controlled perception.
Wrong. Think about it in terms of a tracking experiment (I find that this is
always a good way to think about control). While controlling the cursor I may
trace out a profanity with my hand; this profanity is a side-effect of my
control actions but prehaps it is a disturbance to a variable you are
controlling. So you say "Allow me to give you some feedback about your
tracking behavior; you just traced out "f**k you" and I am offended". But
this "feedback" (conventional sense) of yours is NOT part of the
environmental feedback (PCT sense) loop relating my actions (hand movements)
to the controlled perception (cursor position). Your "feedback"
(conventional sense) _might_ be a disturbance to a different variable I am
controlling (such as the loudness of the environment) but it is not part of
the cursor control (negative feedback in the PCT sense) loop.
Bruce Gregory (960725.1610 EDT) --
Stepping on a banana peel is an example of positive feedback, no? 
No, I don't think so. It would be positive feedback if the sensed
acceleration of the leg caused by the banana peel led to muscle tensions
that further _increased_ the leg's acceleration. I don't believe this is
what actually happens. I think that the increased leg acceleration due to the
banana peel leads to muscle tensions that tend to decrease the acceleration.
Unfortunately, the leg acceleration caused by the banana peel can be more
than the muscle system can handle, so you "slip". But the slip, I think,
represents good ol' negative feedback control that _fails_ because the
disturbance -- the increase in acceleration -- is too large for the system to
handle).
Best
Rick
[From Bruce Gregory (960726.1030 EDT)]
Rick Marken (960725.1500)]
Martin Taylor (960725 14:10) --
When someone says "Allow me to give you some feedback," they normally do
so because one effect of your action has been to disturb a controlled
perception of theirs. That effect may have been a pure side-effect of your
own perceptual control, but nevertheless their "giving" you feedback is
part of the environmental feedback path between your action and your
controlled perception.
Wrong. Think about it in terms of a tracking experiment (I find that this is
always a good way to think about control). While controlling the cursor I may
trace out a profanity with my hand; this profanity is a side-effect of my
control actions but perhaps it is a disturbance to a variable you are
controlling. So you say "Allow me to give you some feedback about your
tracking behavior; you just traced out "f**k you" and I am offended". But
this "feedback" (conventional sense) of yours is NOT part of the
environmental feedback (PCT sense) loop relating my actions (hand movements)
to the controlled perception (cursor position). Your "feedback"
(conventional sense) _might_ be a disturbance to a different variable I am
controlling (such as the loudness of the environment) but it is not part of
the cursor control (negative feedback in the PCT sense) loop.
I think the issue here is whether one is attempting to control
the onlooker's report of his or her perceptions (presumably the
only access we have to those perceptions, which is what we are
ultimately trying to control). If so, "feedback" seems, as Martin says,
to be a part of the feedback path. If not, it is a disturbance to a
different variable as Rick argues.
I'll keep an eye out for banana peels.
Bruce
[Martin Taylor 960726 11:00]
Rick Marken (960725.1500)
When someone says "Allow me to give you some feedback," they normally do
so because one effect of your action has been to disturb a controlled
perception of theirs. That effect may have been a pure side-effect of your
own perceptual control, but nevertheless their "giving" you feedback is
part of the environmental feedback path between your action and your
controlled perception.
Wrong. Think about it in terms of a tracking experiment (I find that this is
always a good way to think about control).
OK. In the tracking software, the experimenter has included code that
takes the mouse movements, and some few seconds later adds them to the
influence on the cursor (as well as using the current mouse movements
immediately). That's more like the condition I described, where the effect
of one's actions returns later to affect the _same_ perceptual variable.
In the "real world" situation, the "extra" side-effect goes through a much
less easily traced route before it returns through the person "giving
feedback" to affect the perceptual variable, but I think the analogy is
not bad, given the extreme simplification required.
While controlling the cursor I may
trace out a profanity with my hand; this profanity is a side-effect of my
control actions but prehaps it is a disturbance to a variable you are
controlling. So you say "Allow me to give you some feedback about your
tracking behavior; you just traced out "f**k you" and I am offended". But
this "feedback" (conventional sense) of yours is NOT part of the
environmental feedback (PCT sense) loop relating my actions (hand movements)
to the controlled perception (cursor position).
No. As far as the controlled perceptual variable is concerned, it's still
a side effect. But what if the person, in "giving you feedback" swings
you forcibly around in your seat, or joggles your hand? Or if the
implications disturb you and ruin your concentration? These "disturbances"
are not directly added to the CEV, but they still affect the behaviour
of the loop, and are influenced by your prior actions--the side effects of
your control behaviour. Other consequences of your actions might affect
the CEV directly. Does the fact that you didn't anticipate them make the
difference between considering them part of the environmental feedback
function versus considering them as autonomously generated disturbances?
If so, or if not, how does that relate to the ongoing discussion about
"models of the world?"
Your "feedback"
(conventional sense) _might_ be a disturbance to a different variable I am
controlling (such as the loudness of the environment) but it is not part of
the cursor control (negative feedback in the PCT sense) loop.
That's often the case. Now, how about considering the situation I proposed?
I think the implications are quite interesting.
Martin
