[From Bill Powers (950512.0900 MDT)]
Avery Andrews (950512.1123, Eastern Oz Time) --
The idea of "ersatz" perceptions is a good one, if we can manage to take
"ersatz" in the sense of "substitute" and not "shoddy." Controlling an
airplane is normally done by looking out the window, but when you're in
a cloud layer, you have to substitute a different set of perceptions,
the instrument readings, which you're very glad to have.
Wish I had more time to think about this.
I wish you did, too.
···
--------------------------------------------------------------------
Bill Leach (950512.00:12 U.S. Eastern Time Zone) --
Your commentaries on the world-model discussion are greatly appreciated.
I think that instead of the Heisenberg Principle, what you want is the
idea of Chaos. The Heisenberg Principle doesn't come into play in
macroscopic systems, but that's just where chaos does show up. Even in a
noise-free analog world, multiple integrations are the rule rather than
the exception. The result is that even the tiniest differences between
models and reality rapidly magnify as time goes by, requiring constant
checks against real-time perceptions. A control system can maintain
order in a chaotic world, but only because it is always monitoring
outcomes and adjusting its outputs to keep the outcomes in the desired
states. There's no way that an open-loop system can create a specific
outcome in the real world for more than a few seconds or at best,
minutes. I understand that the longest that most people are willing to
drive a car (two integrations) with their eyes shut is about two
seconds.
---------------------------------------------------------------------
From Oded Maler (950512) --
It might be that this "shampoo in the eyes" situation is much more
common than you seem to admit.
Also, it may be even less common.
Whenever you control for a sequence of actions, there is an
implicit underlying model.
Fortunately for me, I very seldom control for a sequence of actions.
Normally I control for a sequence of perceived events, varying my
actions as required to keep the perceived sequence going properly.
It might be true that internally, every step in the sequences is
controlled for by means of a "standard" PCT control, but the
perceptual variables involved are correlated more to internal
states of mind than to real-world CEVs.
Tell me something: how do you know there are any real-world CEVs? What's
your proof? Come on, Oded, convince me!
Another point: does an elementary control system has a model of the
sign of the effect its effector has on the controlled variable?
Maybe if we agree that it is this sense of "having a model" that we
are talking about (and not an explicit conscious model) there will
be no dispute.
The control system IS a model of the sign of its effector. That is, the
sign of the effector action exists as a physical organization. If it
proves to be wrong for the environment, higher-level systems can detect
the resulting runaway conditon ( this takes 400 to 500 milliseconds) and
reverse the sign of the output function to regain control. If you want
to try the experiment for yourself, ask Rick Marken for the program that
he used in doing his reversal experiments.
-----------------------------------------------------------------------
John Anderson (950511.1045 EDT) --
According to B:CP, sequences are 5th order perceptions. What gives?
A couple of years ago, Gary Cziko raised some hard questions about the
then-5th level, the result of which was splitting the "event" level into
two levels. The meaning of the event level in BC:P was simply some
familiar space-time pattern of lower-level perceptions, like a bounce of
a ball or a familiar spoken phrase. That's still what the intent is for
ther 5th level. But the idea of a sequence is more one of temporal
ordering -- whether events happen in the order A,B,C or A,C,B, for
example. That kind of sequence didn't seem to fit, and also seemed out
of order in the hierarchy, because being of lower level, it didn't allow
for such things as sequences of relationships (above, below, above) or
of categories (sentences). So that aspect became the sequence level and
was inserted just above the category level.
All of this, you must realize, is just an attempt to find some
hierarchical order in perception. We don't have any good experimental
proof that such levels exist; they're more like research proposals. The
main requirements are that a perception of a given level n be a function
of (be decomposable into) perceptions of lower level up to n-1, and that
in order to control the level-n perception, it is necessary to VARY the
perceptions of lower levels. The existing definitions seem to fit those
criteria, but maybe other definitions would also work. There's no
"official" list of levels; only a default list.
----------------------------------------------------------------------
Samuel Saunders [950511:1715 EDT] --
Seems that several of you out there are starting to think like PCTers.
"How baseball outfielders where to run to catch fly balls" by
McBeath, Shaffer, and Kaiser in 28 April 1995 Science (569-573) may
be of interest.
John Anderson (950510.1630) also came up with the same article, and
several others noticed it but didn't get around to commenting on it.
It's obviously a good PCT experiment!
----------------------------------------------------------------------
Best to all,
Bill P.