[From: Bruce Nevin (Wed 920601 08:59:24)]
The mirroring around level 0 reminds us to what extent apparent
structure in the environment is a reflection of structure in the control
hierarchy, projected there by the observer. We assume vice versa, but
that can only be an assumption. (Right, Wayne?)
So it is worse than perhaps Martin has said. Not only may the observer
(the investigator) identify the wrong environmental variable V as that
controlled by the observed control system, the two parties may also have
differently structured control hierarchies, and so may parse the
environment differently into environmental variables.
We have to assume some commonality as a working hypothesis, in order to
proceed at all. We must not forget that this is an assumption, which
itself must be tested so soon as some other hypotheses have been tested
and found stable in its context. Start with a temporary scaffold (I
assume you and I perceive the world alike; or I assume that there is a
like world for each of us, however we perceive it). In its context,
propose some environmental variables (as perceived by oneself) as
perceptions controlled by the other. Having attained some good results
and some less good results, go back and try to get more uniformly good
results by replacing parts of the scaffold. Question whether the world
as one perceives it is identical in a few particulars to the world as
the other perceives it.
Nagel somewhere has a famous image for science, of being in a storm at
sea on a boat that we must take apart and rebuild as we are sailing.
Learning:
Seems to me introspectively that often learning and reorganization
involve lowering the gain on control of perceptions that I had been
controlling, and "listening" for the "voice" of the next-best candidate
in pandaemonium. Control at some higher level n stays the same;
presumptions about the (lower level) means for controlling at level n
are relaxed. Alternative means are tried out. But the alternatives are
not random. They are the next-most clamorous contenders. An analogy
might be drawn to recessive traits in the gene pool.
Bruce
bn@bbn.com