[Vladimir Jojic 971022.1350 MET]
Simple question:
How does the hierarchy of control systems (or certain parts of it)
percieve performance of one control system? Do we control the performance
of control systems?
Your answers, please,
Vladimir
[From Bill Powers (971021.0800 MDT)]
Vladimir Jojic 971022.1350 MET--
Simple question:
How does the hierarchy of control systems (or certain parts of it)
percieve performance of one control system? Do we control the performance
of control systems?
The word "we" is ambiguous. Do we, as conscious persons, perceive the
performance of one control systems? In a sense, yes, because we can
consciously perceive the outcome of control and if it's not what we wanted
to perceive, we see the result as "wrong" or inadequate. Sometimes, in
memory and imagination, we can even see what went wrong: at the stop light
I made a left turn instead of a right turn, which explains why I'm now out
in the country instead of downtown where I wanted to be.
However, "we" seldom perceive _why_ something went wrong -- why we turned
left instead of right. So in that sense we can't consciously monitor the
performance parameters of our control systems.
"We", on the other hand, can mean "some subsystem inside of us," which need
not work within the field of consciousness. This takes us into the area of
reorganization. If a control system is experiencing a lot of error because
its parameters are misadjusted, it is possible for a reorganizing subsystem
to monitor the general level of error signal in that control system. When
the level of error is large, this subsystem starts changing the parameters
of control while the control system goes on working. There are various
possibilities for how this might be done. The changes might be purely
trial-and-error: if a change makes the error smaller, keep changing in that
direction; otherwise pick a new direction of change at random. That's the
"reorganizing system." It's also possible that the error-monitoring system
has some systematic way of adjusting parameters that will guarantee a
reduction in the error signal (if possible), meaning improvement in the
control system's performance.
Then there are other kinds of changes. For example, if you start lifting
heavy things that are near the limit of your muscular abilities, the limb
position control systems will experience large errors. Your body, however,
will repond to repeated errors of this kind by growing new muscle cells, so
your muscles will be able to exert larger forces and keep the position
errors smaller. Many parts of the body react to demands this way: they
physically grow, or organizationally change, to meet the increased demand.
"Meeting the demand," of course, means reducing the difference between what
is wanted and what is actually accomplished.
So for this kind of change to happen, do "we" have to be able to monitor
the performance of the system? Consciously, no: we simply put demands on
our own control systems, by asking them do to more, and _something_ inside
us alters the parameters of the system to make the initially poor control
better. As folk wisdom puts it, "practice makes perfect." Or at least better.
In general, we don't monitor "performance" as an abstraction. There must
always be some index or indicator of performance. The simplest and most
direct indicator of the performance of a control system is its error
signal. Good control = small error, poor control = large error. However,
it's also possible for a control system to have side-effects that are taken
as indicators of performance. The performance of your systems for getting
food can be measured in terms of your body weight, for example, which isn't
ncecessarily under direct control but does depend on your skills at getting
food into your mouth. If you're losing weight, youy can't fix that by
increasing the gain of youyr weight-control system, if there is one: you
may have to learn better ways of tracking game without scaring it away.
Your question is similar to question about "survival" as a goal of the
organism. Organisms don't learn to survive better; they learn to control
things better, specific things on which survival depends. In the same way,
organisms don't learn to perform better. They learn better control of
specific variables on which performance depends. It isn't necessary for an
organism to think in terms of survival or performance in order to improve
its potential for survival or its performance. That's fortunate for animals
that can't think about these things in human terms.
Best,
Bill P.