[From Bill Powers (960312.0630 MST)]
(Horning in on the discussion of positive feedback)
Scott Graham (960311) --
Here's another situation (from a phsyiological perspective): A
person with a heart condition takes nitroglycerin regularly and
when they experience chest pains (negative feedback -- reduce the
error). Now, for the sake of this illustration, suppose the error-
reducing effects don't happen' the patient's electrical impuses to
the heart increase, their BP goes up, etc. They get rushed to the
hopital where some vasodilating drugs are given (another attempt to
reduce the error). Still no effect...the patient's electrical
impusles go besurk and they enter v-fib.... The MDs now introduce
positive feedback -- amplify the electrical error -- hit the
patient with some voltage...the system collapses, recalibrates --
hopefully...
You have to decide what system you're talking about before you start
figuring out what kind of feedback you have. From the doctors'
perspective, the positive feedback is the first situation: when the
patient's condition is bad, the doctors do something that actually makes
it worse (they think the treatments should make it better, but the
opposite happens). Of course we don't know for sure that this is
positive feedback: the patient might have got worse even with no attempt
at treatment. But assuming that it was the treatments that made the
patient's condition move ever farther from the doctors' goal, this is
positive feedback. The bigger the error gets, the more energetic the
treatments become, and the more energetic the treatments become, the
bigger the error gets.
When the doctors finally shock the heart, this treatment brings the
patient's condition _toward_ the goal, so now we have negative feedback.
The loop we are talking about is defined relative to the active agents,
the doctors.
In the case of worrying, again you have to decide what loop you're
talking about. Now we have to think about levels of control. Worrying
is, at one level, a specific process that can be carried out to any
degree, from zero to maximum. You consider the problem and try to work
out solutions. But worrying also implies that there is some kind of
error at a higher level; the worrying is a means chosen to help correct
the higher-level error, like not enough money in your checking account.
At the lower level, the only error involved is between the reference
signal, which specifies how much worrying should be going on, and the
perception of how much worrying is actually going on. If the amount of
worrying that is actually going on is less than the reference amount,
then there is an error which can be corrected by worrying harder.
At the higher level, the controlled variable is not how much you are
worrying, but how much money is in your checking account. The worrying
is just a lower-level process that is supposed to do something about the
higher-level error. Of course it doesn't. Worrying _per se_ has no
effects on checking accounts. If you worry effectively you might
actually reach a solution (look for a better job). But worrying, for
most people, is just an occupation in itself, involving imagining all
kinds of horrible futures but not doing anything to change them. The
worrying process is not actually connected to what you are worrying
_about_, in the sense of having any effect on it.
Controlling the worrying does not mean _reducing_ it. It means being
able to increase it or decrease it at will. If you can change the amount
of worrying as you please, then once you realize that it's not solving
the problem, you can set the reference amount of worrying to zero and
stop worrying. But if that reference signal is not available for
manipulation, the worrying-system is like a little parasite living a
life of its own in your brain. You have to become able to alter the
reference signal _somehow_. Which way you alter it is not the main
question. The first problem is to be able to alter it at all.
Once you can turn the worrying-system up and down as you wish, you will
probably stop calling its actions "worrying." You'll just say that you
solve problems.
···
-----------------------------------------------------------------------
Stefan Balke (960312.1200 MEZ) --
The itching example of positive feedback isn't completely clear. One way
to interpret it is that the initial scratch creates a skin reaction
which actually increases the itch, which increases the skin reaction,
which increases the itch ... without requiring any further input from
the hand. If this is what you mean, then you have diagramed a little
local positive feedback loop.
However, another way to interpret this is to say that the net effect of
scratching is to increase the itching perception (there is local
positive feedback, but not enough to cause runaway by itself). Now we
have to consider the larger loop in which the scratching is the output.
Normally this loop is set up with a reference level for zero itching,
and the scratching is roughly proportional to the amount of itching. If
there were no dermatitis, the scratching would reduce the itching to
zero, and the scratching would stop. However, with the local positive
feedback which reverses the effect of scratching, a small itching error
leads to scratching which leads to a larger itching error which leads to
more energetic scratching, and so on until pain limits the process. The
reversal of the sign of effects at one place in the larger loop reverses
the sign of feedback in the larger loop and makes it into a positive
feedback loop.
----------------------------------------------------------------------
Best,
Bill P.