Two Wrongs Do Make a Wednesday

[From Rick Marken (950524.2145)]

I just wanted to get this out real quick so that I could make
Thursday a little better for all my fans out there in PCT land;-)

Not only was I wrong about control of behavior (it can be done) but
I was also wrong about Hans' model -- it does control. So that's
TWO wrongs in one Wednesday for the ol' PCT policeman.

I haven't seen Bill's post on Hans' model yet but I spoke to him today and
he said that Hans' model is doing some controlling of xt. I just checked
this out and found that it's true; there is control of xt but it is VERY
weak. Using a constant reference (10 units) I measured control in terms
of RMS deviation from the reference. The measured RMS deviation was
3.29 units; the expected deviation (assuming no compensation for the
disturbance at all) was 3.59.So there is an 8% reduction in the effect
of the disturbance; not overwhelming but better than nothing (for comparison,
in out tracking tasks a trained subject typically eliminates about
95% of the effect of the disturbance.

So Hans' model does control but it controls poorly. Most important, what
control this system does exhibis is a result of control of the perceptual
representation of xt. When the system is blinded, the RMS error goes
from 3.29 to 4.32; there is actually positive feedback without the
perception of xt.

So Hans' system does control (sort of) and to the extent that it
does control, it is control of perception.

But, then, the way today is going, I may be wrong about this;-)