[From Bob Clark (930306.1930 EST)]
"Errors" ("Mistakes" etc) occur. "Correction of Errors" is very common and
is an important aspect of behavior.
Errors are of many kinds and occur for many reasons. Rather than a catalog
of types and reasons, the following is an analysis of the general relations
between "Error" and Perceptual Control Theory.
Recent discussions on the Net have considered suggestions that "Feedback
Systems" consist of "Error Correcting Systems." In criticizing these
suggestions, comments have emphasized that feedback control systems neither
"Control Errors" nor "Correct Errors." Rather, they "Oppose Disturbances"
of a "Controlled Variable," maintaining its condition as specified by their
Reference Signals. The detection and correction of errors by Feedback
Control Systems were not included in those discussion.
Both of these views define "Error" as the discrepancy between current and
intended conditions of a perceptual signal. For an Error to be perceived
as an Error, both conditions of the signal must be available somewhere at
the same time. "Control of Error" consists of reducing the size of the
discrepancy.
In the discussions noted above, both the "Error Correcting System" and the
"Feedback Control System" assume a single control system, rather than a
combination of control systems. The suggested "Error Correcting System" is
not defined in terms of the origins and nature of the variables involved.
The details are left to imagination, but limited to a single system, with
"Compensatory" elements suggested. In contrast, a PCT System includes
specification of perceptual variables, reference levels, environmental
variables, error signals and their interacting components. Since the PCT
System operates as a unit, no "Error" can be perceived until the system has
completed its action. Thus a single PCT System cannot even perceive, much
less "correct," its own "error." Therefore a single feedback control system
cannot act as an "Error Correcting System."
However, as above, Error Correction is a very common and important form of
behavior. These activities can be included by the addition of higher
levels, as in the Hierarchical Control Theory. Higher levels can provide
for both the perception and control of errors.
THE FOLLOWING VIEWPOINTS ARE SUGGESTED IN MY POST: Bob Clark (930301.1730)
The "USER'S" view.
This is the viewpoint of higher levels as they use lower levels to satisfy
their own reference levels. With at least two levels, the higher level
perceives the current condition of the variable(s) controlled by the lower
level. The higher level also perceives the memory that provides the
reference signals to the lower level. The higher level has a longer
time-scale, so that it observes the situation as the lower level's action
is completed, and compares it to the condition that was intended. The
"intended condition" would have come from a still higher level. The User
may use that level, or still higher levels, for "corrective action."
The ENGINEER'S" view.
This view lacks direct access to events inside the human system. However a
"hardware" system having conceptually similar components can be used for
analysis. Thus it is possible, in concept and in "hardware," to examine
the operation of the system. Thus events can be observed during the
reduction of the discrepancy between the current and required conditions of
the controlled variable.
These events may look like "correcting the error." But it is merely the
ordinary operation of a feedback control system. An Error does not exist
until the system has completed its control action. In addition, the
"discrepancy" cannot be perceived by the system itself -- the only
perceptual variable existing within the system is the feedback signal.
Other "signals" (e.g., the output signal) within the system are not
available outside the system. They are not "Perceptual Variables."
The Engineer can analyze the effects of changing the system's parameters.
The Engineers's view can easily include the interconnections and other
relationships between levels, as in Hierarchical Perceptual Control Theory.
Thus it can also be used to analyze the User's view.
Since these events have the attention (of the Engineer's DME), they are
being recorded as they are occurring.
The "OUTSIDER'S" or "Observer's" view.
If an error, or any "corrective action," occurs, that event may be observed
-- that is, get the Observer's attention. That is, the Observer (his DME)
pays attention to the events. The observations are interpreted in terms of
those theories available to him. These events are also recorded in the
Observer's memory because they have the attention his DME.
The "EXPERIMENTER'S" view.
The Experimenter also has the view of an Outsider, but, as an Experimenter,
he may make some changes in the situation. He may change the environmental
conditions. He may give information or instructions to the Subject. For
interpretation ("Understanding") of the subsequent events, he also uses the
theories available from his memory.
The "THEORIST'S" view.
The Theorist recognizes the validity of each of these views and relates
them to his own experiences and explanations. He examines proposed
theoretical descriptions and explanations for adequacy, relevance and
consistency with the available data. Some theories may be useful in
special situations, but are seriously limited in describing or explaining
important aspects of the situation. Other theories range from being
irrelevant to down-right wrong. Perceptual Control Theory is the only
theory that comes anywhere near accounting for all aspects of the observed
events.
I am sure that the topic of "Errors" and Perceptual Control Theory can be
treated much more effectively than I have done. I offer this as a
beginning.
Regards,
Bob Clark