[Martin Taylor 940608 0950]
Rick Marken (940607.1900)
Words have many possible definitions, and for the purposes of keeping
discussions intelligible to all parties, it is useful for each to know
what the other means by the use of terms. Rick has provided a definition
of what he thinks I mean when I use the term "straw man." Herewith a
definition that more closely describes what I intend (usually and in
the specific instance on which Rick comments). Rick's definition may
apply when he uses the term, or in other circumstances.
straw man, n. 2. A structure developed out of metaphorical straw, and
hence readily demolished. Usually an attempt by one person to elaborate
an incompletely described phenomenon or theory mentioned by another, to
which the elaborator is unsympathetic without knowing what the phenomenon
or theory really is. The destruction of the straw man is taken to be
destruction of the underlying phenomenon or theory. A technique often
used by opponents of PCT, but sometimes used within the PCT discussions.
Example of straw man in sense 2: "the description of the alerting
mechanism used by Powers to explain why such a mechanism is
unnecessary in a hierarchical control model." (quote from Marken's
examples of straw man, sense 1.)
No mechanism was specifically proposed for the phenomenon of alerting that
is observed in living organisms. PCT (unelaborated) explains why living
(at least mobile) organisms should be expected to have alerting systems,
thereby (once again) providing a conceptually strong link between otherwise
unrelated observations.
In discussing alerting systems, I mentioned the probability that there are
several possibilities for what an alerting mechanism might be, without
describing any in sufficient detail for them to be tested. Bill produced
an elaborate structure outside the normal PCT mechanisms, and said that
there was no need to go to such lengths to alter the PCT structures to
account for the phenomenon of changing the degrees of freedom being
controlled from minute to minute. I agree with that, but the kicker
that makes this a straw man was the conclusion--THEREFORE, there are no
alerting systems that we need to consider within PCT.
NB. When using this term, don't, repeat DON'T, make the mistake
of providing a working model to show what you actually meant.
In the matter of alerting systems, PCT shows the necessity, not the mechanism.
Bill's baroque system might actually work. I don't know. But a much
simpler system is to note that for many purposes, complex systems of ECSs
work better if each has a dead zone within which a detectable error produces
no output. Call it a "zone of tolerance" if you like. We discussed a lot
of this, I think, in the postings on pull-only systems which emulate
mulidimensional sets of linear virtual ECSs. Recognizing that almost all
ECSs at any moment are not producing significant output (at least none that
eventually effects the muscles), is to recognize that their gain is zero
(or rather, the gain of the equivalent virtual two-way ECS is zero).
The perceptual signal is either exactly equal to the reference signal or
in the dead zone for the virtual ECS. When the perceptual signal for an
ECS goes out of the dead zone, the ECS produces output, and necessarily
will introduce conflict in the hierarchy. It pulls against something else
that is controlling. The signal with the greater Gain*Error product will
have the greatest influence on the actual output, thereby shifting which
degrees of freedom are under control.
I don't think that this is the only mechanism for alerting, but it is one
that might work for many situations, and requires no additions of which
I am aware to the accepted structure of HPCT.
The point of a "straw man" is that it is built entirely by the one who
demolishes it, perhaps on a foundation provided by another, but not using
the other's design.
Martin
