[Michael Fehling 930907 10:42 AM PDT]
I've just spent a couple of days enjoying Bill Powers' book on PCT (1973). To
the chagrin of my wife (who wanted to go out) I've now completed 15 chapters
and hope to finish the remainder by next weekend.
Although I am _very_ busy finishing up some funded research tasks, I cannot
resist a few quick comments:
1. From now on I will recommend this book to my students alongside other
classics such as Ross Ashby's _Design_for_a_Brain_ and _Introduction_to_
_Cybernetics_ and Norbert Weiner's _Cybernetics_ and _God_and_Golem_.
2. It is clearly the intellectual community's loss that it has not been
exposed to this work. The so-called "cognitive revolution" in psychology
failed to address some of the most important failures of behaviorism. Had
that movement been cognizant of Bill's insights and application of cybernetic
concepts, perhaps we would have made more real progress. Based on my exposure
to Bill's book and CSG-L I am developing some hypotheses about why this
important work has been so widely ignored. Once I have convinced myself that
I understand this work, I hope to offer some friendly criticisms intended to
remedy this unfortunate disconnect with the wider research community.
3. I particularly applaud Bill's insights into the _methodological_ as well
as the substantive limitations of behaviorism and his willingness to offer an
alternative. His remarks apply equally today, in spite of the so-called
cognitive revolution. The methods of cognitive science largely replay
methodological ideas of behaviorists such as Tolman or Hull (cf., Allen
Newell's book, _Unified_Theories_of_Cognition_). E.g., most cognitive models
describe a system of variables and transforms that _intervene_ between stimuli
and responses. So, while cognitive _theories_ ascribe purposive behavior to
subjects, the empirical methods used to test these theories obviate the
any operational concept of purpose.
4. Bill's treatment of consciousness, awareness, and volition, although
brief, was one of the most fascinating surprises in the book. These topics
were absolutely taboo in scientific psychology when this book was published.
Now, the research community has grudgingly begun to open up to these issues.
(E.g., I am an editor of the new journal _Consciousness_and_Cognition_. There
has been a flurry of books and major articles on these topics published
recently by highly regarded authors (e.g., Baars, Dennett, Searle, Jackendoff,
etc.). This spring I participated in the 25th annual Carnegie Symposium on
Cognition, where the focal topic was "consciousness as a scientific topic.")
Here again, I think that the inclusion of PCT would improve and clarify the
ideas finally being explored openly by the scientific community. Let me tell
you what I am doing to remedy this. I was asked to write a commentary for the
our journal on a paper entitled "Why volition is a foundational problem for
psychology." I have told the editor that I am deferring my writing of this
commentary until the next issue so that I can reflect my best understanding of
PCT. This may turn into a full article depending on how much progress I make
with my understanding of PCT, my other duties, etc. At the very least I want
the readers to know that someone was willing to address these very difficult
issues well before they became fashionable to the mainstream.
5. I am also quite pleased to see how well my own work on consciousness and
volition (with Bernard Baars) seems to parallel PCT's treatment. We view the
intelligent agent in terms of its ability to control its perceptions of itself
and its world. We are exploring a model of the cognitive-connative-affective
system as a (roughly) hierarchical collection of "unconscious processors" that
contribute to operation of "control loops" at various levels. In a manner
similar to PCT, our model proposes operations within a "global workspace" (GW)
that _controls_ the reorganization of this hierarchy of processors. Finally,
as with Bill's treatment in his PCT book, consciousness, awareness, and
volition are all explained in terms of the interaction between the
hierarchical system and the GW as the reorganizing controller.
Surely, we shall find discrepancies between PCT and our own model of an
intelligent system as we go forward. However, my first reading of Bill's book
has already made clear the importance of modeling more explicitly the
_control_loops_ to which the unconsconscious processors contribute. (We have
not made this focus sufficiently clear up to now.) In fact, I find that PCT
offers a much more principled neurobiological model on which to build a
psychological theory than the current stock of "connectionist" models (which
are largely based on behaviorists' S-R perspective). So, PCT is having an
immediate impact on my own work.
6. Based on what I have read so far (including on CSG-L), it seems clear that
a great deal of work remains to be done to clarify the higher levels of Bill's
model, in particular the "program" and higher levels. In fact, I think I may
be able to help here. For now, let me note the following interesting
correspondence between Bill's model and the elements of a so-called symbolic
programming language, (e.g., LISP). The basic element of LISP is are data
structure called _symbols_ and _lists_. A symbol is an _index_ into possibly
unnamed structure. The associated structure can represent a pattern (i.e.,
configuration) of quantities (e.g., vectors). On first reading, this compares
favorably with the sensation-configuration-sequence levels in the PCT
hierarchy. Logical operations, programs and other higher level entities
(e.g., heuristics) can be modeled in terms of operations with and on these
basic "symbolic" entities. If this correspondence can be made to hold water
then the modeling tools we have been using might be well suited to push more
deeply into PCT's higher levels of organization. In return, I am confident
that this excercise will also clarify just what a _symbol_ really is.
7. The issues being debated regarding organizations (per se) as control
systems seem clearer to me as well. Bill's monograph gives me needed guidance
on how best to approach the debate (along with the excellent posts on this
list, of course). I'll have more to say about this later, when I respond to
the challenge to show how the organizations we are studying qualify as control
systems.
In the long run I may or may not find myself a full-fledged supporter of
PCT. It's too soon to tell. Perhaps my own stubborn commitment to my
preformed theories will impair my perceptions. Regardless, I _can_ say that
whatever direction my theoretical musings take, PCT is now thoroughly installed
as a significant influence on my work.
Thanks, Bill. I'm enjoying the hell out of this!