[Note from Gary Cziko: I had sent some of Powers's, Marken's and
Williams's remarks about manual control to Chris Wicken and I am now
forwarding Wicken's responses to CSGnet.
Since Wickens is not on CSGnet, comments for his attention should be sent
to BOTH CSGnet and <c-wickens@uiuc.edu>. If this is not done, I will have
to forward them to Chris myself, with a consequent delay of anywhere from 1
minute to 24 hours.]
Hi Greg: I like your comments on my behalf. I see a lot of what was
communicated as based upon an evaluation of smaller excerpts of my
writing, that made inferences on what I believed/didnt believe, or
didnt know on the basis of quotes that were intended for a different
audience, and on the basis of a very selective set of quotes. (And I
realize here, that you knew they were selective when you used them to
illustrate). I wonder how many of those critical of my "ignorance"
had read the full breadth of my chapter, or my 86 chapter in Boff
Thomas and Kaufmann's handbook of perception on manual control. If
they have, and find it lacking, thats fine. But if their only
criticism is based on the quotes (and that serves as a basis for the
"flunking" coment), then I have a minor disagreement.
In any case, I find that what I "really" believe about manual
control, is not all that different from the final tenants of PCT,
outlined by Rick, at the end of his "happy hunting" message, nor do I
think these tenants are much at odds with all of the pioneering work
on models of manual control carried out by Duane McRuer, Hank Jex,
Ezra Krendel, Bill Levison Dave Kleinman and Shelly Baron, in the 50s
and 60s, all of which influenced my work so heavily. (I assume that
this is the group of people that are criticized heavily in the
e-mail? The mainstream manual controllers?).
In any case, as I understand the criticsims of (this?) work (or is
it other manual control work that is being criticized?), it is based
on the purported assumption that:
a. Error exists in the real world, not in the head
b. There is no role for intentions in the models.
I dont think any of us in the serious manual control literature
believe that error (as defined explicitly by the difference between
the state of the output..its position and derivative -- and the input
-- its position and derivatives) MUST automatically be the signal
that drives control (with an opposite sign, to reduce the error).
This IS the way that negative feedback systems typically operate
(with various weights adjusted on the various derivatives), but any
feedback system can chose to respond to any state of the world that
it choses.
WHat is important to recognize is that the well developed Optimal
Control Model (OCM) of Baron Levison and Kleinman, has explcititly
incorperated the concept of intentions. The operator is assumed to
have an internal representation of the FULL state of the system and
the world (subject to his/her limitations of perception and
prediction), and on the basis of this can CHOSE to respond to any
aspects of this that are consistend with his or her GOALS (ie.,
intentions). This latter is explicitly incorperated in the model via
the cost functional J, which is an internal value characterizing the
different things the controller wants to minimize or maximize (e.g.,m
smooth ride, lane deviations from center line, number of wheel
reversals etc.), and there is at least some manual control research
out there that shows the sensitivity of parameters of the model to
these changes in goals and values.
(Indeed it is some of these internal features of the OCM, that led to
its criticism by some engineers as being too "human driven".
IN contrast, much of the original development of the Crossover
model was more susceptible to the criticisms that the PCT group seems
to offer. But it was only meant to characterize that aspect of
behavior of pilots when such nulling of an external error signal was,
indeed critical to the task at hand (landing an aircraft you HAVE to
keep pointed toward the runway). But in their defense, this group at
Hawthorne rapidly expanded the model to incorperate a broader set of
behaviors, and the flexibility of responding to different goals
(intentions) was Explicitly incorperated into Krendal and McRuer's
"Successive Operations of Perception" (SOP) model, in which, for
example they descrbe the role of internal states as generating the
signals upon which control can be based. Can I assume that the PCT
group is aware of all this work, and finds it also wanting? I am
curious as to whether these are any of the people whom one writer
described as a "manual control guru"? and who was the audience of
"manual controllers" to whom he spoke. (I too have found, regretfully
that a lot of engineering psychologists, who should know better, are
sadly ignorant of all aspects of control theory, whether from an
engineering or a PCT persepctive).
Finally, an issue was made of measurement. How can we measure what is
the basis (the intentions) of control behavior. Jex Jewell and Allen
have looked at this in terms of trying to characterize the shared
covariannce of output with different aspects of input, in order to
ascertain the extent to whcih subjects WERE responding to error
(compensatory behavior) or other aspects of the environment (pursuit
behavior)....Sorry. Make that Allen and Jex. I have often found
theirs to be a particularly elegant approach, of using control theory
to diagnose intentions.
As a personal note, I heard Bill Powers when he gave a colloquum
some years back at Illinois, and was much impressed. I shared his
concern that psycghologists lacked an appreciation
many psycholgists lacked an appreciation of the role of feedback in
behavior. I also am aware of some of his writing, in particular the
psych review paper. I am not aware of the degree of awareness of the
PCT school to much of the manual control work that I alluded to
above, which flourished in the 60s and 70s. To echo a remark that
someone (you Greg?) made in the message, if the PCT group was not
aware of this work, then at least it is exciting that two schools
appeared to be thinking along the same tracks in parallel, and
arriving at what appears to me to be converging positions.
It is with some sadness that I found that the Annual Conference on
manual control, which really served as a forum for this work, died
out in the early 80s. I am glad to know that PCT provides a forum
where it apparently still flourishes.
Please feel free to pass any or all of this along to the group, Gary.
Also, I AM curious whether the characterizations of manual control
theorists that I layed out above ARE known to the group? ANd if so,
how they are found wanting from the three tenants that Rick
outlines.
Chris
···
--
Christopher D. Wickens
University of Illinois
Aviation Research Lab.
#1 Airport Rd.
Savoy, IL 61874
(voice) (217) 244-8617 (FAX) (217) 244-8647
Internet: cwickens@s.psych.uiuc.edu