Tom Bourbon [950110.1723]
[From Rick Marken (950110.1030)]
Martin Taylor (950109 17:30) --
Nothing in this world is "a cause of itself" if you include the time
Then, by the same argument, nothing in the world is a cause of anything else.
That is the *appearance*, if not the intent, of what Martin said. Martin?
It [the perceptual signal] is part of the cause ofthe perceptual signal at
time t+tau, where tau is definitively non-zero, and at later times, possibly
extending beyond limit.
But this is true of all variables in the loop:
p --> e --> o -->q -->p
The error signal, e, is caused by p (and the reference signal,r) but at some
(very small) time delay; output, o is caused by p at some (longer) time
delay; same for the environmental variable, q and, finally, p itself. So, if
you don't want to say that p is a cause of itself in this loop because of the
time delay between p as cause and p as effect of that cause then you would
also have to say that p is not a cause of q, o and e either, since there is a
tine delay between cause and effect with all these variables.
Rick, you are taking the words out of my mouth. Is that sanitary?
Of course, even Rick's graphic presentation depends on laying all of the
variables out as though they worked in a lineal fashion. In a closed loop,
causality does work in one direction, but it does not work in a lineal
manner, something Rick makes clearer below.
So the "opposite" of a straight line should be seen as a helix, not a circle.
If you want to capture the time dimension, I suppose this is true. But a
helix can also be a way of preserving the (incorrect) intuitions of a
sequential analysis of a control loop; p THEN e THEN o THEN q THEN p again.
In fact, all variables in the loop (except the disturbance and reference
variables, which are not actually in the loop) are always (and
simultaneously) a cause and an effect of the other variables in the loop. A
helix gives the impression that the value of p as CAUSE differs from the
value of p as EFFECT. This provides a way to sneak a cause-effect view of
control back into the picture; p causes output and then waits until the
output changes p so that a new output can be caused. This is NOT the way a
control loop works. At any instant, the value of p is BOTH a cause (of future
values of p) AND an effect (of previous values of p).
Now you are taking the words right out of my keyboard. Definitely
Only slightly more seriously, Martin, when I read your post I was
immediately struck by the idea that you were trying to preserve lineal
causality and hide the fact by invoking the image of a helix. Event would
follow event, in lockstep, along the straight line that has been bent into
a helix. I doubt that was your intent, Martin, but that was the appearance.
All functions in the loop act all of the time. All signals are present all
of the time. Except for the few milliseconds after a big impulse
disturbance hits an environmental variable that is controlled by the system,
there are no times when you can trace a big impluse-like "event" through
the system, the way you can follow a small rodent through a big snake.
Are Rick and I missing something in your analogy, Martin?