MCT <> PCT, Anticipation

[From Rick Marken (970225.0900)]--

Bill Powers (970225.0530 MST)

One thing I have noticed in trying to read textbooks about
control theory and control system design:...there is essentially no
introductory level of discussion...I simply don't see how a student
can develop an intuitive understanding of control.

Excellent point. I've noticed the same thing when reading introductory
texts on control theory. One of the (many) nice things about B:CP and
your other writings about control theory is that it gives the reader
(who is willing to learn) a very nice, intuitive _feel_ for what it
means to control and for how control systems work. B:CP does this using
the behavior of living organisms as the example of control and the
nervous systems of these organsisms as the example of control systems.

Once one has learned control theory from B:CP I think one has a very
good _feel_ for control theory. I think B:CP and several of your other
writings should be the introductory material for _any_ course on
control engineering. I think it's important to have a good intuitive
understanding of something before diving into the technical details
(Laplace transforms, Bode plots, integral and proportional control,
etc). B:CP and (I think) the BYTE articles do a particularly good
job of providing this intuitive understanding.

Bill (Kid) Powers, after some practice rounds at Rick's PCT Cafe and
Gym, pounds Bruce Abbott with:

If, as you say, you can make these estimates while standing still,
you ought to be able to specify the landing location immediately.
So a test ought to be fairly simple. Lay out a numbered string
grid in the possible landing area, and have someone throw balls,
with you specifying the grid location of landing immediately after
the ball is thrown (or batted, of course). Then record the guess
and the actual landing position.

Bruce Abbott (970225.0905 EST), reeling after his failure to
_anticipate_ even one of the devastating jabs landed by Kid
Powers, cries out:

Yes, it's a testable hypothesis, as I've been saying. And quite
possibly wrong.

The referee suggests that Abbott take his hypothesis and test it
before bringing it back into the ring again. This is at least the
second time Abbott has dragged this hypothesis into the ring against
PCT. Some of the fans out here are getting pretty tired of seeing all
these helpful suggestions about what PCT _needs_ (models,
anticipations, predictions, etc etc) which are based on (previously)
received wisdom rather than scientific test.

I think it might be useful to note that many (but surely not all)
behaviors that seem to involve "anticipation" are actually behaviors
that involve control of higher level perceptual variables that are
defined over time. A simple example is control of a circular pattern of
finger movement. Ask someone to track your finger as it moves in a
circular pattern. After the person has been tracking smoothly for a few
seconds, _stop_ your finger movement abruptly. The person will
continue to trace out a _circular_ pattern with her finger for about
1/2 second. This is an actual controlled movement -- not an failure
to be able to rapidly brake the movement. It looks like the person is
_anticipating_ what your finger was going to do (but didn't do) during
the next 1/2 second. What is actually being seen (from a PCT
perspective) is the time it takes for a higher level control system to
change the reference specification for the perceptual signal
representing the _temporal pattern_ of finger movement from "circle"
to "stationary" (zero).

Best

Rick