This particular fire hydrant has
always had trouble understanding howthe control hierarchy ‘decides’ which perception to control and, in
particular, how it rapidly shifts from controlling one perception to
controlling another without the delays associated with relying on
upperlevels in the hierarchy to reset reference
levels.
[From Bill Powers (2003.06.27.1407 MDT]
Bruce Gregory (2003.0627.1217)–
I don’t think we control just one variable at a time – or am I
misreading what you mean?
When I look up
from tuning my car radio and see the brake lights of the car in front of
me, how does the perceptual hierarchy know enough to slam on my
brakes?
I don’t understand this way of asking the question. It knows everything
you know, doesn’t it? If a higher system in you has temporarily removed
visual inputs from the distance-keeping control system, that system won’t
perceive any brake lights until you restore the visual inputs. If you
look up after the distance has decreased significantly below the range
you normally maintain, there will be an abnormally large distance error
which should lead to a more energetic braking action than usual. If there
is still plenty of distance to the car ahead, you’ll just brake normally,
won’t you? That’s what I do.
When the traffic is
dicey, how does the perceptual hierarchy know thatI can’t afford to direct my attention to looking for a cassette to
play? It has always seemed from my lowly position that the hierarchy
isextremely capable but lacking in much sense of what is important at
themoment.
I liked Rick’s answer to this. Since we’re dealing with a hypothetical
case, it has exactly as much sense as you choose to give it by proposing
details for your model.
The way I would answer this question would be to look at how I know
that I should pay more attention to traffic when it looks bad. When I see
a lot of cars nearby, I watch other cars and the road more, and look only
very briefly at objects or people inside the car, if at all. I think I
tense up a little, too, which moves the motor systems toward the centers
of their operating ranges as well as making the muscles stiffer. It may
be that I also raise the gain of the control systems involved in
controlling spatial relationships. The result is that I control with less
error, counteracting smaller disturbances and continuing to control until
the errors are smaller than they usually are. Why do I do this? Because
the spaces between cars are smaller than usual and my normal control
actions allow the car to get too close to others or to obstacles. The
errors actually get bigger, so I have to raise my control gain to keep
them as small as I want them. It’s not hard to imagine a higher-order
control system that does this sort of thing, is it?
To the extent that
this assessment is not wholly without merit, I offerthe following questions. Could it be that the limbic system
functionsto tell the control hierarchy what is important ‘right now’?
]
I’m not guessing about which parts of the brain are involved. Whatever
brain structure we propose as containing the needed control systems, it
would have to be able to perceive changes in the organism’s relationship
to the environment and detect whatever errors those changes create, and
automatically alter reference signals and/or system parameters in motor
systems in the way required to make the errors smaller. If the limbic
system can do all those things, it may be where that sort of control
takes place. I wouldn’t know.
Further, could it be
that the mechanism employed by the limbic system to conveythis information consists of increasing the gain on some control
loopsand reducing the gain on others?
Certainly. I have long said that higher systems may well act by varying
the parameters of lower systems as well as their reference signals. An
early demo of this effect was offered by Tom Bourbon at my suggestion, 10
or 15 years ago. He set up a model in which a higher-level system
monitored the average absolute value of error signal in a control system,
and varied the output gain in that system to achieve minimum error.
Actually, he set this up as a reorganization task, so the gain variations
were done through a random walk. More recently, I proposed a model in
which an auxiliary control system (whether you should consider it
“higher” or not is debatable) changes the weightings in an
output function in a way that emulates the convolution theorem. It worked
pretty well when embedded in the Little Man model. I called this model
the “artificial cerebellum,” because of some resemblances of
the algorithm to processes known to happen in the cerebellum. Of could
this doesn’t mean that the amygdala could not do something similar.
However, my modeling efforts focus on what kind of control process is
done, which doesn’t depend on guessing which part of the brain does
it.
These questions
do not seem totally implausible on the basis of what little I know about
the wiring of the brain.
They seem plausible to me, too.
Best,
Bill P.