[From Rick Marken (930921.1035)]
Bill Powers (930921.0830 MDT)--
I think we've been sucked into an argument
that has no relevance to PCT -- that is, the argument over what
causes what. In a system diagram, there is no such thing as
causation: there are only functional dependencies that run in
parallel and in loops. Why would anyone want to know "the cause"
of a behavior? Only so as to predict or control that behavior. If
you can get hold of the cause, you can manipulate the effect. But
in PCT there is nothing you can get hold of in that way without
disrupting the system. I say that we should just ignore causal
arguments and concentrate on developing system models that
explain observations. If someone insists that inputs really cause
outputs, we should just say "Maybe so. Now can we proceed with
the analysis?"
This is a brilliant (and brilliantly expressed) point, by the way.
I agree COMPLETELY. I'm out of the causal arguments business for
good. From now on it's "Maybe so. Now, was the variance of some
variable considerably less than expected in this situation?"
Best
Rick