[Hans Blom, 941213]
(Bill Leach 941212.16:24 EST(EDT))
Thank you for expressing your perception of PCT. There are some things
that I recognize, and some things that I experienced differently. I finish
with a correction of you remarks about "information in the perception".
The real shocks [after becoming acquainted with PCT] started coming
after I obtained somewhat of a "handle" on PCT basic and began dealing
with the implications... The range and number of "cherished" beliefs
whose "fundmental" foundations are at least "rocked" if not outright
blasted away is, I think, far greater than one can imagine. Indeed,
such continues to happen as one makes an effort to apply PCT principles
to all thinking about human behaviour.
It is strange how different people experience things differently. When I
read Bill Powers' B:CP, my introduction to PCT (I accidentally discovered
B:CP in the library of my Department, Electrical Engineering!), I got a
eureka-shock as well, but it was of a different nature. B:CP did NOT
"blast away" any of my old preconceptions; rather, it gave me an ADDI-
TIONAL perspective on how organisms function, myself included. I have felt
no need -- and do not feel any yet -- to throw away what I have learned
before; I only needed to REINTERPRET things from this new point of view. I
am therefore not, to the dismay of some, a "convert" to PCT, but -- to use
an analogy -- like a Christian who discovers the deep truths of Buddhism;
rather than relinquishing my old faith, this new discovery gives me a
better, clearer, deeper appreciation of my old faith as well. Call me a
synthesist...
The often made analogy of PCT versus conventional psychology with Newton-
ian versus pre-Newtonian celestial mechanics strikes me as a bit awkward,
especially since Einstein, who told us that it is perfectly acceptable to
talk about the sun circling the earth, the added perspective now being an
explicit recognition of the importance of the position of the observer.
"It's all perception", Einstein might have said.
I am beginning to think that the learning and understanding of the
basics of PCT is the minor step. The major "step" begins when one
starts seeing how virtually everything that one has understood about
anything (including oneself) is all affected by the principles
underlying PCT.
How very true. The new perspective needs to be integrated with all of our
pre-existing notions.
I have received similar shocks -- eureka's -- upon my discovery of other
theories as well: psychoanalysis, evolution theory, and Popper's falsifi-
cation theory, to name the few that jump to mind right now. All these
intuitively appealed to me immediately as kernels of deep understanding.
Not THE truth, mind you, but of valid perspectives on "truth", however
mutually irreconcilable they may seem superficially, and are, in details.
The following on "information in the perception" needs to be modified,
however. I don't know whether you have designed (classical) adaptive
controllers, those systems with dual functions -- to control AND to learn
(how to control better) -- but at least in that arena your remarks need to
be qualified:
Since I was pretty strongly in the middle of much of this last round of
"information in the perception"...
1) What we agreed to was that there would be information about a
disturbance in the perception to the extent that control system
gain was insufficient to cancle the effects of the disturbance
upon the environmental parameter that was the object of the
control.
Modified: A perfectly controlling system cannot learn, because there is no
information in its perception (only random noise). A control system CAN
learn as long as it does not control perfectly; in that case there will be
a systematic effect (i.e. information) in its perceptions. A very simple
example from classical control theory: in a proportional controller with a
finite gain and a constant setpoint there is a systematic non-zero offset
between setpoint and feedback signal (perception). This systematic offset
can be adapted away (made zero on average) by adding an integral component
to the controller. This is true even if the offset is much smaller than
the noise amplitude.
4) It is currently impossible to even conceive of actually measuring
any of the necessary signals to make use of this "information".
This is simply not true. See the above example where this "information"
can be used to adapt the gain of the integral component in such a way that
a minimum offset results. This "information" is used to the hilt in adapt-
ive controllers.
5) The whole matter is one of idle curiosity and not relevant to PCT.
At least those who think that LEARNING (adaptation) is important, will
disagree. Adaptation isolates and uses the "information in the perception"
in order to tune the controller so that better performance -- a smaller
average offset -- results. Note that as an effect of this tuning, the
"information in the perception" disappears; IT IS USED. So those who speak
of there being no information in the perception are right only in the
extreme case: a fully adapted/correctly tuned controller.
Some simple adaptation schemes that every control engineer will be fami-
liar with implicitly use this principle. They add a dither or a pseudo-
random noise sequence to the controller's output. This extra noise
certainly does not improve the short-term control performance, of course:
it is an extra disturbance. Yet such an imperfection in control is what is
required if learning/adaptation is to be possible. What it does is some-
thing akin to establishing a gradient that the system can "climb".
Does this link with classical control schemes make sense to you?
Greetings,
Hans