Randomness in PCT

[Martin Taylor 950801 15:40]

Bill Leach 950712.20:09

(I'm gradually catching up on the backlog, before leaving for another 10
days or so)

PCT holds that a random function exists in the neural system. And that
the randomness is absolutely exclusive of any external influence.
...
For example if one could hypothetically set up say 10
exactly identical organisms and subject them to exactly the same
environment, the random changes that would occur in each could not be
predicted regardless of the information available in advance and they
would not all react in exactly the same way to their identical
experience.

There was some further discussion of this, but one point may not have been
emphasized.

My understanding of PCT is that "randomness" is only in the reorganizing
function. I put "randomness" in quotes, because the only reason for it is
that the organism has no way to direct changes in favourable directions
at the moments that a "random" change happens. (Much of the time, changes
are not random, at least in the e-coli approach to reorganization--they
just continue and augment changes that have proved beneficial in improving
control). In this context, "random" means "according to some means not
obviously related to the problem being solved." And it also means "If I
try again, something unrelated will happen."

What "random" does not (necessarily) mean in this context is that two
identical organisms exposed to the same influences will change differently.
According to QED, they may, of course, but the "randomness" aspect of
PCT does not require it.

This idea is not only postulated by HPCT but is actually essential. If
such a process does not exist then HPCT can not explain even in a "loose"
way how living systems accomplish that which they have been observed to
accomplish (that is the "macro" view of behaviour of the human species).

("This idea" is that identical organisms exposed to identical influences
will behave differently).

"This idea" is actually untestable except in simulation, because no two
biological organisms are identical (other than possibly genetically), and
never are two organism subject to even macroscopically identical influences.
The differences in changes that might come from QED uncertainty are heavily
swamped by the differences that must derive from the different structures
and influences of each individual organism. The QED-derived differences
will never be distinguishable from these much larger environmental
differences. And once a single reorganizational change has affected one
organism differently from another, further changes would be likely to differ
even if identical influences could be brought to bear.

ยทยทยท

-------------------

PCT claims that experience in control attempts results in "learning". I
would go so far as to claim that both successful and unsuccessful control
experiences result is "some degree of learning".

This is a question I have refrained from introducing, for some time (it's
a new thread, I think, and to go into it would probably detract from the
most interesting EAB modelling thread).

It can be rephrased: The usual proposal for "learning" is that reorganizational
changes cease occurring when all perceptual variables are under control.
Perhaps this is not so, and further reorganization continues to increase
the gain of control systems that are actually controlling (in the sense that
the gain of their feedback loop is negative and greater than unity).

If it is the case that reorganization continues by increasing the gain of
systems that are actually controlling, does this render the hierarchy less
adaptable when the environment changes? If reorganization stops as soon as
control is established, does this not mean that the organism is susceptible
to the slightest wind of change--living on the edge of instability?

The "Bomb in the hierarchy" proposition is based on the latter assumption,
that reorganization stops when control is (re)established. If it does not
stop then, the avalanches of change induced by the Bomb will be much less
likely.

The point may be moot, though, because there will always be some conflict
among different control systems in a hierarchy that share aspects of their
output pathways (the "allocation of resources" or "alerting system" problem).
The conflict may be minimized by effective reorganization, by the existence
of dead zones, by time-multiplexing, but it cannot be done away with while
permitting all control systems to control at their best all the time. Hence
it may be the case that reorganization is always ongoing, regardless.

It does seem (subjectively) that one learns to do things better while
doing them well, as Bill L. says.

Plenty of speculation room here:-)

-----------------------

All non-living matter not acted upon by living
matter moves from a state of organized to less organized, from a level of
one potential energy to one of lessor potential energy.

This is physically true only in an isolated universe. In a strong
non-equilibrium energy flow (such as we live in), it does not hold.
Self-organization of structure is the rule--in fact our world itself
is an example, a tightly organized sphere of spherically differentiated
material derived from a slightly non-uniform gas by way of a disk of dust
around the equator of a hot spinning globe. Entropy reduction, all the way.
There are pockets of entropy increase and of entropy reduction, regardless
of the presence of life. Only overall does entropy increase in a (more or
less) guaranteed way.
------------------
On dualism:

Our "only hope" in this matter of purposeful behaviour has in the past
always been the "dualistic" view. That is of course the existance of
"spirit world", a "soul" seperate and distinct from the physical body...
something that did not have to obey the known laws of physics.

I'm afraid I did not follow the argument at all. That one may be a
deterministic structure seems to me to be quite compatible with the
perception that one is one's own master, freely able to choose one's
purposes. And that, itself, is compatible with the HPCT position that
one's purposes eventually depend on the control structures one has
developed (through reorganization) in such a way as to control one's
intrinsic variables. One does not (normally) become conscious of most
of the perceptions one controls, whether they be of high-level, low-level,
or intrinsic. One is (normally) conscious only of mid-level perceptual
variables, and perhaps mostly of program-level perceptions where the
notion of "choice" just might make sense.

So again I don't think non-determinism enters the equation.

Sorry to be negative about a most interesting posting.

Martin