Loss of feedback

[Martin Taylor 950605 11:00]

Bill Powers (950603.1550 MDT)

Martin Taylor (950603.1415) --

Rather than starting with the premise that loss of feedback has to leave
a control system flailing around, why not start from the other end?

What premise? Are you referring to:

But if the PIF works only on the current values of the input signal, then
when the input values change dramatically (such as by vanishing), the
perceptual signal will change dramatically, creating probably large error
values, and ineffective but large changes in output.

Is that a premise or a consequence of talking about the "standard model,"
the form of which is shown in the CSG logo?

Or are you talking about:

The "use only the current
value of the inputs" kind of PIF is tuned for all possible unpredictability.
It results in immediate loss of control when the input is cut off.

Is THAT a premise? Or a fact about the "standard model," a fact that you
have been at pains to emphasize in your interchanges with Hans Blom?

In the case of visual tracking, for example, we tend to assume that the
sense of cursor position is derived entirely from visual information.
But we also know that while we are tracking, we can feel our arms and
hands moving. What kind of input function could use both the visual and
the kinesthetic feedback information to produce a cursor position
signal?

This refers NOT to a case in which the input is cut off, but to a case in
which the input comes from a variety of lower-level perceptions, some of
which no longer have access to the real world data (no telling whether they
use prediction, but assume they don't), and some of which continue to have
access.

Controlling this rough indication would result in errors in
the actual cursor position -- but that is exactly what _does_ happen.
The main thing is that the system would not go open-loop; it would
simply change its apparent characteristics and precision.

Of course I agree with the statements in your message, about the ability to
use multiple sources of input. But they are irrelevant to the question that
was at issue: what happens when the input is cut off, with a "predicting"
PIF or a "use current values only" PIF.

My point was that there is a conceptual distinction between "having a value
of zero" and "having an unspecified value," and that the "standard model"
control loop has no way for that distinction to be represented, either
in the PIF or elsewhere. Such a distinction has to come either from
outside, such as by variation in the reference signal from higher levels
or in the gain (also affected by higher levels), or else it must appear
inside a more complex, non-standard, model. A one degree-of-freedom system
in which the single degree of freedom is dedicated to the value of a
variable cannot simultaneously represent the validity of that variable.

Martin

<[Bill Leach 950606.19:34 U.S. Eastern Time Zone]

[Martin Taylor 950605 11:00]

My point was that there is a conceptual distinction between "having a
value of zero" and "having an unspecified value," and that the "standard
model" control loop has no way for that distinction to be represented,
either in the PIF or elsewhere. Such a distinction has to come either
from outside, such as by variation in the reference signal from higher
levels or in the gain (also affected by higher levels), or else it must
appear inside a more complex, non-standard, model. A one
degree-of-freedom system in which the single degree of freedom is
dedicated to the value of a variable cannot simultaneously represent the
validity of that variable.

The issue was essentially "What do or even can, complex control systems
do upon loss of the actual perception for a "high level" controlled
perception?"

In the simple model there is not much of a problem deciding and the only
even slightly complex part would be deciding which of the two full output
extremes would the output go to? Obviously requiring a little though
upon what exactly are the signal polarities for that system.

In the more complex model of the tracking task, it was actually noted
that in some manner the original goal of tracking was "transformed" into
lower level references related to the original variable but, importantly
still available as current perceptions. This action _could_ be some form
of "model based" control. The early experimental results were not too
encouraging as to accuracy on the part of the subjects!

Finally, what was also considered were "less tangible" perceptual control
efforts. As always, most of us, considered these to be a control of
current perception based upon some "perceptual model" of "boss reality".
In some of the cited examples, control is very good.

Your "incoming" remarks about what a simple control loop would do, only
called attention to the rather obvious fact that you had not been
following that particular thread (not an accusation or "slam" or any
other such things as most all of us have done the same a time or two).

-bill