[Avery Andrews 9304`2.0830]
(Rick Marken (930412.0900))
>Only under very special circumstances can o be reconstructed from
>the input, i, (or from i and the reference,r); specifically, when
>i = o + d. But even in this case (as Bruce Nevin pointed out) the
>relationship is just coincidental -- it is not actually the result
I'm assuming that the characteristics of the output function are
fixed, which seems right for short time scales, tho not for long
ones (muscle fatigue). I doubt that control would work very well
if the first derivative of the output function switched from positive
to negative too often (but this has already been talked about, I
believe).
>When i is a non-linear
>function of o there is no relationship between observed values of i
>and o.
Which is why `and r' is part of my story, as well as a fixed output
function (which should have been mentioned to begin with).
>If inputs
>are at least SOMEWHAT involved in the determination of behavioral
>outputs, then why not keep trying to improve our methods of determining
>how inputs do this? Behavioral scientists are unlikely to turn to The
>Test until the realize, very clearly, that o is not a function of i.
Because this strategy is obviously guaranteed to lead nowhere, & I don't
think that all behavioral scientists are complete idiots, tho there
is undoubtedly a lot of professional obtuseness around. Repeating stuff
that people will already know will not get through this, just
like you can't demolish a stone wall by banging at it with a crowbar--
you have to stick the bar into suitable crevices and yank in the right
direction.
>But it's not useful: it's wrong. I am convinced that this cause-effect
>"expository vocabulary" has effectively prevented behavioral scientists
>from seeing what PCT is all about. It's not about cause and effect;
>it's about control.
But PCT uses the cause-effect vocabulary as well -- given its inherently
flakey nature, it either has to be entirely avoided, or used in a
flexible manner with explicit warnings. For long time scales,
i-o cause-effect exposition is not helpful when dealing with control
systems.