Open-loop measurements

[Martin Taylor 951108 13:20]

Bill Powers (951107.1330 MST)

Rick asks how you can be sure that part of a system is operating open
loop. You reply:

    The experimenter predetermines (or uses a predetermined algorithm
    to select) what "stimulus" is to be presented on trial 37. Nothing
    the subject does on trials 1-36 has any effect on this choice. ...

How do you determine that the experimenter's manipulation changes a
quantity that would ordinarily be changed by the subject's own behavior?

If the experimenter manipulates the brightness of a light or the frequency
of a tone, the results could be relevant only to behaviour in which
light brightness or tone frequency form an element of a controlled
perception. So why does that affect the value of the experiment? It
seems highly unlikely that we have sensitive perceptual functions for
perceptions that never enter into control behaviour.

As you state the situation, there is no way to tell whether the
manipulated variable is a CEV for a control system, or a disturbance of
a CEV. In either case there would be a response, but if the manipulated
variable is actually only a disturbance, the response would be very
different from what it would be if the CEV is a variable that is
normally under control.

For sure. But I'm not sure how that comment relates to the question that
I see as being at hand: whether open-loop measurements of perceptual
capabilities are valid for the analyses of closed loops in which the same
perceptual function is a link.

It's hard to measure open-loop response in a human control system,
because breaking the loop destroys control, and other systems inside the
same organism will immediately start changing reference levels,
sensitivities, whatever.

If sensitivities change, it may well be the case that the open-loop
measurements are invalid for use in the analyses of closed-loop behaviour.
But in the simplified models normally used, we discuss possible changes
of output functions, not of input functions. It would be very interesting
to deal with how detectability changes under control conditions might
affect control behaviour. They couldn't make the peripheral detection
systems very much _more_ capable than they are found to be under open-loop
conditions, since with trained subjects they are typically within 3-6dB of
the mathematical ideal observer; but the closed loop situation might well
reduce the capability of any specific PIF. However, an open-loop
psychophysical study seems often to provide results that are consistent
with fairly stable limiting capabilities across a range of conditions, and
those limits may very well apply also to closed-loop conditions. As I said
before, it could be tested.

Suppose you sat a subject down in front of a
tracking experiment to measure open-loop response to cursor position. ....

I find it hard to imagine what you might be trying to measure in such a
study. All the arguments you range against it apply, no matter what the
measurement might be (except possibly the measurement of how long it takes
the subject to notice that control has failed).

In the experiments both Rick and I have been talking about, the subject
knows before the experiment starts that the measured aspect will not be
under control (or that it will, as in some psychophysical studies, such
as von Bekesy audiometry). So I don't see the point of your example.