[Martin Taylor 970424 11:30]
Hans Blom, 970422b--
Then what would "adaptation" mean to you? There always seems to be a
highest (lowest?) level with a fixed organization, or at least with
fixed "building blocks" that lower (higher?) levels can use to
construct their own thing. Atoms come to mind as pretty well fixed
building blocks. Molecules. Cells. Organs. Etc.
In the standard model of PCT, reorganisation is what 'adaptation" means.
Or perhaps I should say that reorganization is the mechanism whereby
adaptation happens. There are at least 12 forms of reorganization consistent
with the HPCT structure, but most of them probably are relatively
ineffective. The ones that seem most likely to be effective are:
--variations in the Perceptual Input Functions of existing Elementary
Control Units
--variations in the connection weights of perceptual signals to sensory inputs
of higher level ECUs
--variations in the weights the determine the distribution of output signals
to lower level ECUs
--construction of new ECUs.
All versions of reorganization seem to wind up with the notion that low-level
ECUs form early and are very stable. The control of muscle tensions is much
the same whether you are typing, ball-catching, wrestling, or whatever. Those
control systems either are evolutionarily fixed or are adapted quickly, and
this is possible because the environment in which they operate is more or
less constant (it depends mostly on physics and biophysics). They form
"atoms" as building blocks for the next level of control, which cannot really
be stabilized until the lower level is stable. Once the next level is
approaching stability--in the sense that further adaptation no longer
offers much improvement (whether it's a local or a global optimum)-- they
can be used as molecular building blocks for higher levels...and so on.
From the other end, too, there is stability, but this is a stability of
requirement, rather than the provision of atomic building blocks. If the
control systems that are developed do not serve to keep the organism's
intrinsic variables within limits, the organism dies. It is the deviation
of the intrinsic variables from their (independent?) optimal values that
drives the reorganization of the perceptual ECUs. Changing the perceptual
control structure is the output action of the intrinsic variable control
system. Stabilizing the low-level ECU structure ensures that variable
means for controlling the higher-level ECUs will be available as possibilities
for ongoing reorganization, both in learning to control in the existing
environment and as adptation to new environments.
Does this help?
Martin