[From Rick Marken (950508.0920)]
Bruce Abbott (950507.1925 EST) --
I hope I am not detecting sarcasm in your quote.
It was "mild" sarcasm. I think Hans deserves a "wow" for providing an
excellent program that illustrates model-based control; but I think model-
based control doesn't deserve any "wow" at all. Model-based control is, at
best, a model in search of a phenomenon to explain and at worst an attempt to
model the phenomenon of control as computed output (the "controlling" output
is computed as the output that "would" control the modelled controlled
variable).
I think Hans' model should be evaluated on its OWN merits, not on those
YOU'VE imposed.
But the only merits I care about are the ones I look for in a model. I don't
have any idea what Hans' model's OWN merits are. I consider a model of
control to have merit if it can explain control phenomena. Hans model doesn't
explain the most basic control phenomenon -- the ability to maintain a
variable in a constant or varying reference state in a variable environment.
Thus, as a model of control (my measure of the merit of a control model)
Hans' model has no merit. Hans model might have merit as a model of some
phenomenon -- such as the ability to act "as though" one is controlling when
the perception of a controlled variable is briefly obscured -- but at the
moment we know next to nothing about this phenomenon.
The model was designed to function within an environment in which the
disturbance waveform is fairly regular...and to continue to function for
short periods (based on that model) when input to the perceptual function is
lost. I don't see your ordinary error-based control system doing that.
Our ordinary control model functions just fine in an environment in which the
disturbance is regular or irregular; Hans model doesn't function in an
environment with ANY disturbance, regular or irregular. Hans' model does
function for short periods when the input to the perceptual function is lost
(as long as there is no disturbance during this period) and the "ordinary"
control model does not. But there are various ways to put this capability
into the ordinary control model and right now we have no idea what kind of
capability is required. I think it is very likely that any ability to
maintain apparent control while a perceptual input is briefly lost can
probably be handled by the existing hierarchical control model.
My suspicion is that one can maintain apparent control of a suddenly (and
relatively briefly) unperceived variable only when the loss of perception is
part of a higher order perception (like a sequence -- on, off, on, etc). If
there were no higher order variable to control, the system probably would
just stop controlling the unperceived variable -- period. For example, if the
room suddenly and unexpectedly became pitch dark as I was walking across it,
I think I would probably stop (with no more visual variables to control),
imagine what perceptions I would encounter and then start again, VERY
tentatively, controlling a whole new set of variables, real and imagined. I
don't think I would smoothly transition from controlling a visual variable
to controlling the modelled version of that variable (the kind of smooth
transition Bill Powers (950507.1900 MDT) was trying to model. I think this
kind of smooth transition only happens in special cases, like in a pursuit
tracking task with fairly regular target movement where the subject knows
that the (undisturbed) cursor will become invisible at various times.
I think smooth transitions from controlling perception to controlling
imagination can occur; but since we have little or no data on the phenomenon
it seems a little premature to guess at how this transition should be
modelled. But I have a strong suspicion that when we do get the data and try
to model this transition we will find that it can be handled by a
hierarchical control of perception model; I think there will be no need to
have the model do what Hans' Kalman filter does -- which is to compute a
predicted output. I think the "prediction" that occurs in control of
imagination can be handled on the input side of a perceptual control process
that uses the perceptual control system that has last its perception.
But I think we should have some data on control during the transition from
perception to no perception of a controlled variable before we do much
more guessing about the kind of model that accounts for this (currently
make-believe) phenomenon.
Best
Rick