Re: Time, Causality and Demand-Pull
You've certainly got me pegged wrong, Chris. I was not trying to trip you
up verbally but to suggest on your own terms, which do not happen to be my
own, that in the exchange with Chris Partridge you are falling into a
temporal version of Zeno's paradox, by neglecting the purposive principle,
usually called an entelechy, that provides the thread of continuity and
intelligibility which enables us to recognize a series of purposive acts
as belonging to an activity. There can be specifiable advantages in
prescinding purpose from activity in a computational artifact, as PSL
does, but there are no advantages in not being aware that in doing so
elements of meaning will be lost and will need to be resupplied in some
other way.
Your message did persuade me that I needed to get up to speed on PSL.
What I found was an admirably worked out manufacturing process interlingua
built on primitives and postulates consistent with a Lesniewskian logic,
reinforced by the choice not to ground the system in set theory. All
primitive elements, including activities, are postulated as 'things', and
the use of timepoints as a primitive dissolves all continuities into
discrete occurrents. The axioms and extensions carry through the
fundamental conceptual character by concatenating timepoints to cover
duration and occurrents to cover all states of activities. The basic
expressivity is not changed in the extensions, but non-defintional
elements are permitted. I had a few questions about the status of
commutativity in the duration relation, but on the whole I found PSL as
clear and consistent as one could hope. The translatability to other
process ontologies is sketched sufficiently to suggest a high degree of
inclusiveness and interoperability, which was, after all, the goal.
What seemed to me especially artful about PSL was that the activity
primitive is left vague enough to accommodate all kinds of notions of
process, but that once an activity enters the PSL system it is cast into
a discretized form consistent with all other system elements.
For reasons of personal history this gave me considerable satisfaction. I
spent a good bit of time a decade and more ago inquiring into the logical
foundations of Polish praxiology (a term that covers, among other things,
all expert and decision support systems). [An historical note: In the
interwar period Warsaw was one of three international centers-- Vienna and
Cambridge, England, being the others-- for formal logic; after the war one
of the leading logicians, Kotarbinski, attempted a praxiology grounded in
his earlier logical contributions, which insisted that all reasoning had
to be grounded in concrete things.] I began as a severe critic, with the
view that Kotarbinski's extreme nominalism (which he variously called
reism, concretism, pansomatism), and the still more rigorously formulated
one of Lesniewski, was incompatible with a praxiology. This criticism was
shared by all but a few Polish logicians, who thought the stumbling block
of the older generation lay in the rejection of set theory. But I
concluded that a praxiology founded on minimalist grounds was tenable and
might have certain virtues complementary to those of a rival view that I
found, and still find, more cogent. Now here in PSL is a computational
example of the tenability of a praxiological vocabulary grounded in
a minimalist logic that does not depend on set theory. Contrary to some
earlier postings, both the strengths and limitations of PSL seem to have
been exactly anticipated in previous work in formal logic.
Of course, for any translation, there is always the question of what is
lost. John Sowa's suggestion seems hard to rebut: that a view of process
in which a space-time continuum is a primitive or even a prominent feature
will "lose" something crucial in translation into a language like PSL
grounded in timepoints and discrete occurrents. What is lost cannot be
recovered at the superficial level of extensions to the system. This
translational loss-- which Michael Gruninger compared with inadvertant
prescience to the Tarski-Hilbert views of geometry (Tarski being the pupil
of Lesniewski and Kotarbinski)-- goes well beyond mere computational
interoperability in the bare, non-interpreted sense, since, if one is
impervious to meaning, interoperability can always be achieved by one or
another kind of hack.
For the effort to create an upper level standard, all this poses further
issues. Let me pull out and address in a further message four threads
that seem worthy of attention:
1. relating loose upper level terms to axiomatized ontologies
2. distinguishing 'practical' time and process in decision-support
and agent-friendly ontologies
3. the gradations of theoretical reasoning in creating a standard
4. the stipulative aspirations of the current ieee effort
Lee
> >1. Menzel to Partridge:
> > > Intervals and timepoints *both* seem to be part of a commonsense
> > > ontology of activities, and there are ways of defining each in terms of
> > > the other, so it doesn't really matter which we start with -- though
> > > perhaps then both should be part of a basic formalization of the
> > > commonsense ontology.
2. Menzel to Auspitz:
Well, obviously, a process ontology in which time is explicitly discrete
would not be interoperable with one in which it was continuous.