The _Relevance_ of Predictability

[Martin Taylor 960326 11:10]

Bruce Gregory 960325.1742

I am also aware that Martin feels that neither of us is under-
standing the points he wants to make.

What follows illustrates the accuracy of that statement:

I know from my own experience driving that I rarely attempt to predict
what action I will have to take to offset a disturbance, although I
have learned to avoid certain disturbances or to slow down to give
myself more room to respond before a forseeable disturbance (such as
windblown snow on the road) makes its presence known.

If there were any question as to the predictability of the influence of
your action--for example, which way the car would turn if you move the
steering wheel clockwise--you might begin to attempt to predict it. You
might start looking for contextual perceptions that would help you to
determine whether on this occasion you should turn it clockwise or
counterclockwise in order to make your left turn. It is because the
influence of your action is highly predictable that you don't have to
concern yourself with it. But in your examples, you do exactly that,
when you see the likelihood that the influence of your steering action
may soon become less predictable unless you do something to maintain its
predictability.

Remember that we are talking about what happens _within_ one control
system when we are talking about the predictability of disturbances or
of environmental feedback functions. When you talk about "avoiding
certain disturbances" you are dealing with something quite different.
It is predictable that if you drive on a snow-covered icy surface, then
the enviromental feedback functions relating to steering and braking
actions become unpredictable. When you slow down or avoid driving on
snow that may cover ice, you are using the predictability of control
at one level to enhance the predictability of control at a different level.
You have a perception of, and a reference for, the predictability of your
steering control; you want to maintain at a high value the predictability
of the effects of steering-wheel movements.

In the examples you give, you are using the fact that it is predictable
that some other control may well become unpredictable. You are controlling
predictably a perception of predictability.

I think you make my point better than I have been doing. Thank you.

Martin

[From Bruce Gregory 960326.1400]

Martin Taylor 960326.1110

  ...in your examples, you do exactly that, when you see the likelihood
  that the influence of your steering action may soon become less
  predictable unless you do something to maintain its predictability.

Could I express my concern, not in terms of how predictable I want
my steering action to be, but in terms of the control I want to be able
to exercise? I foresee (predict) a potential loss of control over the
direction in which my car is moving. I reset my reference level for
speed from 60 mph to 30 mph, and I control my speed in the usual way?

   It is predictable that if you drive on a snow-covered icy surface, then
   the enviromental feedback functions relating to steering and braking
   actions become unpredictable. When you slow down or avoid driving on
   snow that may cover ice, you are using the predictability of control
   at one level to enhance the predictability of control at a different level.
   You have a perception of, and a reference for, the predictability of your
   steering control; you want to maintain at a high value the predictability
   of the effects of steering-wheel movements.

I think, not in terms of the predictability of my steering control, but in
terms of controlling the direction in which my car is moving (my reference
is to keep the car on the highway, no matter what I have to do with the
steering wheel. I am sure I am missing something here, but I don't know
what.

Regards,

Bruce G.

[Martin Taylor 960326 15:00]

Bruce Gregory 960326.1400

Martin Taylor 960326.1110

...in your examples, you do exactly that, when you see the likelihood
that the influence of your steering action may soon become less
predictable unless you do something to maintain its predictability.

Could I express my concern, not in terms of how predictable I want
my steering action to be, but in terms of the control I want to be able
to exercise?

Yes, you would be saying the same thing in different words.

I foresee (predict) a potential loss of control over the
direction in which my car is moving.

Yes, that's it.

I am sure I am missing something here, but I don't know what.

The next step from what you foresaw...the reason why you might be about to
lose control: that the connection between your turning of the steering
wheel and the change of direction of the car might become less predictable
or even be lost altogether unless...

I reset my reference level for
speed from 60 mph to 30 mph, and I control my speed in the usual way.

You imagine (predict, foresee) that if your speed is 30 mph, the perceptual
effect of turning the steering wheel will be more predictable than at
60 mph, or that if it is not, you will have time to perform other actions
that might prevent some higher perception from going out of control (e.g
tensing or relaxing some muscles, putting your hands up to save your face,
or whatever). If you continued to believe that a 10 degree turn of the
steering wheel would have the same effect on the direction of the car as
it usually does, you wouldn't bother to slow down.

I think, not in terms of the predictability of my steering control, but in
terms of controlling the direction in which my car is moving (my reference
is to keep the car on the highway, no matter what I have to do with the
steering wheel.

Of course you do. Or rather, I would put it even further from "thinking about
predictability" by saying that you don't think about those perceptions you
have learned to control, such as the normal connection between moving the
steering wheel and changing the car direction. If you have a good car (as
I did once) you "think" the car where you want it to be, you don't "drive"
it there. In a well buit car, the connections are so consistent that all
you perceive is the "tactical" traffic situation, not the car at all. That
does not mean you cease to control your perception of the car's position
on the road or cease acting by turning the steering wheel. You just don't
notice those controls any more than you notice the controlled perceptions
involved in normal walking.

When you are a skilled driver, accustomed to winter conditions, I don't
think you usually perceive (consciously) the snowy and possibly icy patch
of road, though you control using that perception, using actions that might
involve slowing down or deking around the patch. You just do it, without
noticing you've done it unless something draws your attention to it.

···

---------------
What I'm talking about is the same kind of thing as trusting that a metal
rod you might be using for some purpose will not suddenly melt. It's
about the reliability of the environmental feedback path. That's not
something you think about when it is normally reliable and the influence
of your actions on your perceptions is predictable at least over the time
frame of your control loop delay.

So the fact that you don't _think_ about predictability doesn't mean that
you don't control the perception of predictability (i.e. the perception of
an imagined future ability to control).

I hope this helps.

Martin