Gems from Bill Leach

[From Bill Powers (950710.0700 MDT)]

Bill Leach (950709.23:29 U.S. Eastern Time Zone) --

Some real gems today, Bill.

     An impossible to deal with problem is that the invokation of
     "behaviour is caused by present AND PAST reinforcers" one is faced
     with a totally untestable assertion.

That's the problem, all right. If present reinforcers predict behavior
different from what we observe, all you have to do is say that past
reinforcers must have made up the difference. The "past reinforcements"
clause is an unlimited supply of wild cards that makes every hand a

     I suppose that some might think of me as being philosophic and
     stretching the limits of credibility I will even maintain that the
     the model proposed by Han's is STILL ULTIMATELY a closed loop
     negative feedback model when viewed as an overall system (at least
     to the extent that it actually works -- and in some environments
     such systems really do work). An important part of the negative
     feedback is largely ignored and that is the work of the design
     engineer. The engineer's work in designing a "world model" and
     engineering a means for "self-modification to that model is a
     substantial input to the system.

I've been saying the same thing. The engineer designing such a system
has to know whether the system is working right, and this means
monitoring the state of the plant and comparing it with the state that
the _engineer_ desires to perceive. The properties of the system are
then adjusted until the actual state matches the desired state. That's
closed-loop control. That's all that makes the artificial system work.

In Hans' model, the "system identification" is build in by the engineer,
in the form of a world-model with adjustable parameters. All the
adaptive system has to do is adjust the parameters of this world model
to minimize some measure of error. If the adaptive system also had to
pick the right mathematical form for the world model, the job would be
considerably more difficult, to say the least.

     I just don't see any way to convince anyone that does not already
     understand what the phenomenon of control is all about that living
     systems must be control systems. One can get so wrapped up in
     chaos theory, systems theory, game theory and the like that the
     obvious becomes something that is totally incomprehensible.

Brilliant. Precisely. It's not so much that these other approaches are
wrong; they're simply misapplied as basic explanations of behavior. As
basic explanations, they're too complicated for the phenomenon they're
meant to explain. It's like using quantum mechanics to explain the
orbits of planets. Some people revel in complexity, and what's worse
they have the brain power to deal with vast systems of arcane equations.
This ability can be a handicap because it leads to overlooking simple



Bill P.