[From Bill Powers (940604.0820 MDT)]
Martin Taylor (940603.1400) --
The experiment I see as based on these ideas is a simple
compensatory tracking experiment, in which the subject has to
keep a cursor in its reference position. Quite ordinary,
except that the subject doesn't see the target directly.
People can control only what they can perceive. If the subject
doesn't see the cursor's relation to the target, that relationship
can't be controlled.
Instead, what happens is that the subject sees Rick's display,
and the proportion of A1 events is some function of the
placement of the cursor relative to the target.
The subject is seeing a periodic repetition of a1/a2 events
following presentation of a symbol S, and per instructions is
perceiving the fraction of the time that S is followed by a1. That
fraction is the CEV with respect to this control system. The
connection between the subject's handle movement and this
presentation is
subject's output Perceptual input to subject
>
Handle --> cursor position --- - ------> S -> a1 OR S -> a2
^
>
target position ---
The fact that the display is a function of an invisible cursor and
target is unknown to the subject. So this is equivalent to what Rick
proposed, the situation in which the handle position plus a
disturbance affects the probability that a1 or a2 will appear after
S.
This experiment is worth doing, simply to show that the relative
frequency of occurrance of two mutually-exclusive events can become
a controlled variable.
The uncertainty involved here, however, is not the uncertainty with
which the cursor-target displacement is represented in the subject's
perceptions. The CEV is not the cursor-target displacement, but a
function of it derived by an environmental process (not by a
perceptual process in the person).
To achieve what was originally described, we must arrange for the
cursor-target displacement to be represented in perception in an
uncertain manner, and also to provide the subject with a way to
perceive the cursor-target displacement exactly. By using the
uncertain perception and the certain perception, the subject can
then calculate the uncertainty in the perception of cursor-target
displacement in the uncertain system. This would provide the basis
for altering the uncertainty in the uncertain perception of cursor-
target displacement, given a means to do so. Finding that means is
somewhat of a problem.
And I repeat my question: if there is a way for the subject to
perceive the actual state of the cursor-target relationship, why
would that perception not be used directly for control of the cursor
relative to the target?
ยทยทยท
---------------------------------------------------------------
Best,
Bill P.