[From Rick Marken (931109.1100)]
There have still been no comments on the relationship between
PCT and human factors models of manual control from the human
factors manual control experts. It's pretty hard to build a bridge
between two perspectives on human nature when one side is on a
raft floating downstream. I think there is considerable interest
on this net in finding common ground between PCT and other, more
familiar, approaches to understanding human behavior. Is there
any interest on the part of the human factors experts in finding
such commonalities -- or is PCT just too far "out in space"?
cc: Chris Wickens, Gavan Lintern
Tom Bourbon (931108.1005) -- who is actually right FAR more often
than I am and is one of only a handful of people who could possibly
appreciate the appropriateness of the "space" segue above-- says:
Much of the current
round of discussion about feedback-feedforward seems to hinge on the idea
that something special must happen that (at least some of the time) allows a
person to walk successfully through a familiar but now darkened room.
Martin Taylor (931108 19:40) replies:
No. It is that the same things that allow it are happening normally. One
uses the best information that is at hand. If you have direct perception of
the current situation, that's usually the best you can get. But it isn't
One uses the "best" information at hand? How does one know which
information is the "best" at hand?. Does one have information about
what information is the best to use? How does one know that the
information about the "best" information at hand is the "best"
information about that information? I think there are some
loose ends in this model.
Is is possible that what actually happens is that one controls the
perceptions that ARE at hand? That's how the PCT hierarchy works --
the "best" perceptions being the one's that match their reference
Martin Taylor (931109 11:00) --
Rick, does this "quite some time" allow for the very large slowing factors
that you use in the spreadsheet model?
I think the main reason for the stability of the higher level perceptions
is simply the fact that (as Tom said) they are made of several components
and it takes some time for changes in these components to result in
change in the aspect of the world that corresponds to the perception
that is controlled.
If some of the input variables
don't correspond to "reality", the perceptual signal ought to be giving
the wrong answer, and the error ought to be causing the wrong output to
Yes. If I understand you correctly, this is what is happening. When, say,
a level 1 system goes open loop, it's perception no longer represents
the actual state of the physical variable that corresponds to
the perception it controls. A higher order system (say, level 3) acts
as if nothing is wrong and, in the process, generates outputs to the
lower level system that are actually inappropriate (with respect to
reality). The result is that the real world variable that corresponds
to the level 3 perception being controlled starts to drift away from
its intended value. But the level 3 system knows nothing of this. Only
an observer (me) who knows what result the level 3 system intends
to produce can see that this result is eventually no longer being
produced (although it does get produced "open loop" for some time).
For example, say that the level 3 system is controlling for A>B where
A and B are two lower level perceptual signals. Say that the system
controlling B is open loop -- so B is always constant (say, zero).
In actuality, the physical correlate of B may (due to disturbance)
gradually become greater than A (because the system is open loop).
But the system controlling A>B never experiences B>A -- remember, it's
open loop, imagining that A>B. If there are any negative consequences
of A actually becoming less than B, this MIGHT be reflected in eventual
loss of control in OTHER control systems (if they exist) that depend
on the physical correlate of A being greater than B.
This is different from the situation Tom was describing,
in which one information source can be used when another, that usually
provides better information about the same CEV, has been disrupted.
I think it is precisely what Tom was describing (Tom can correct
me on this). Tom never said anything about one information source
being used when another has been disrupted. Tom was talking about
the fact that higher level perceptual signals are the outputs of
functions that take MANY lower level perceptual signals as
inputs. So removing some of these lower level inputs (by "blinding"
the control systems that produce them) does not necessarily destroy
control immediately. The spreadsheet model shows that this intuition
of Tom's was right.