Re: CCAT: TIME and STATEfritz@rodin.wustl.edu (Fritz Lehmann)
Date: Tue, 4 Oct 94 08:54:55 CDT
From: email@example.com (Fritz Lehmann)
Subject: Re: CCAT: TIME and STATE
Date: Tue, 4 Oct 1994 06:40:37 -0700
From: Fritz Lehmann <firstname.lastname@example.org>
Subject: Re: CCAT: TIME and STATE
Cc: M.J.Johnson@qmw.ac.uk, anquetil@IRO.UMontreal.CA,
email@example.com, firstname.lastname@example.org, email@example.com,
firstname.lastname@example.org, email@example.com, dick@Glue.umd.edu,
firstname.lastname@example.org, email@example.com, firstname.lastname@example.org,
email@example.com, firstname.lastname@example.org, email@example.com,
firstname.lastname@example.org, email@example.com, firstname.lastname@example.org,
Pat Hayes wrote:
Fritz>I would like CCAT/TIME to settle
>upon either a deep axiomatization, ...
Let me suggest that this is a mistake. Part of the motivation for the
comparative survey is precisely that one cannot settle on a single,
universally accepted, ontology for time. The best we can hope for is that
people who believe that time is the real line, and people who believe that
time is discrete (for example), can at least communicate with one another
and understand one another's assumptions. We arent going to get people to
agree on which is right, they have been arguing about it for thousands of
I expect similar things will happen elsewhere.
Agreed. Commonsense and other forms of reasoning
will put _constraints_ on the time ontology, but choices of the deep
level may be possible. If certain strange phenomena are consequences
of the choice, so be it. We'll document it, and "Let a hundred
flowers bloom. Let a hundred schools of thought contend."
(Unlike the originator of this phrase, Mao, we
won't then run a lawnmower through the flowers.)
fritz>...OR an "ontological
>interface specification" which lists the minimal
>predicates/relations to be expected from any time
>ontology. (The interface would be a list of
>commonsense expectations of these predicates
>in typical situations, but would not place
>demands on extreme or borderline cases.)
That would be most useful. What do people expect a time reasoner to be able
to think about? Please send me lists of concepts, if necessary with brief
intuitive explanations of what they mean, thanks.
We are in complete accord. I hope all who have demands
to make of the TIME ontologies will send Pat (and the CCAT list)
their required concepts and relations. These will at least include
the Allen-style relations in Cyre's ICCS-93 paper, and _some_
notion of (quasi at least) time-point.
I have some immediate suggestions. Let's demand at least a time-
interval (or set of time-points) equal to 1/2 of the Planck time-length,
and all integral multiples thereof. Also, the international standard
atomically-defined second. Also a true year as internationally defined
(with intercalary days and seconds). Also, Universal Time, Sidereal
Time and Greenwich Mean time. Although we don't have to know which
counties of Indiana are in which time zones (or which houses in Sau'di
Arabia), we should have a well defined notion of Standard Time and
local time, with the ontology "knowing" that these are dependent on
geographic location, the day of the year, and the controlling authority.
Much of this can be taken directly from existing standards documents
and programs, I surmise.
A complication has arisen. I had thought relativistic
time would be irrelevant to all practical and common-sense work,
but I may have been wrong. The new Global Positioning System of
satellites (which may be involved in things as mundane as
routing delivery trucks) depends on very precise timing. The
satellites are adjusted before launch to compensate for both
special and (when orbits are eccentric) general relativity.
Failure to do this adjustment reportedly causes an 18-meter range
error -- the GPS system would misplace you by 18 meters according
one source, although others give smaller values. See the book
"GPS Satellite Surveying" by Alfred Leick, Wiley-Interscience, 1990.
Regrettably, CCAT may need to provide for relativistic time, maybe as
an option. GPS gets its time from an atomic clock at Master Control
at Falcon Air Force base in Colorado; it lacks intercalary seconds.
The UTC to GPS correction is available from the US Naval Observatory
"time announcement" publications. The last "common epoch" of UTC and GPS
was January 1980. See P. S. Jorgensen, "Relativity Correction in GPS User
Equipment," Proceedings of the Position, Location and Navigation Symposium,
IEEE Inc., New York, 1986.
One can imagine the consequences of an 18 meter position error
in a legal case involving a traffic accident or burglary.
Also, we will need a notion of tolerance and imprecision in time
(as in many other things). The ARPA Rome Planning Ontology has
Such a notion. The tolerance in ordinary life ("I'll be home at
five, Hon'") is much larger than in switching electronic circuits.
The appropriate leeway will depend on the application and on the
context/microtheory in use.
As I suggested to Gerard Ellis, communicating processes
may impose other burdens on the time ontologies.
We will also need everything required by natural language,
including indexicals like "now". Bernard Moulin has done some
interesting work on the nesting of indexical time-contexts, for
the purpose of representing compound tenses in real life and
nested stories. See his ICCS-93 and ICCS-94 papers.
Pat, can your three time ontologies handle all of the above?
Can they bear the strain?
On the subject of STATE, Hayes said:
>suggests that it's a set of propositions which
>remain unchanged over an interval of time. (Did I get
>that right, Pat?)
No. I dont think a state is a set of propositions at all, I think it
is....well, a *state* of something. States are real, they cause things and
are caused by them, and things are in states from time to time. States
arent propositional or symbolic in nature, and they don't (usually) refer
to other things. Changes of a property of something can be described as
changes in the state of that thing, so maybe a state could be identified
with the collection of all properties that something has, but I dont think
there is any need to get this abstract, and it leads into all kinds of
wierd questions (is the set of all properties countable or describable?)
People might want to distinguish different kinds of state, such as states
of motion and states of rest, but they are always things that physical (and
some abstract) things have, or are in, for some period of time (which might
itself be only a single timepoint).
What I was trying to get at is that a _change_of_time_ does not
change the state. Something else, other than the passage of time, is
needed for a change in state. Maybe a very low level "occurrence" or
"event" is a difference in something other than the time, which
accompanies (is a function of?) a difference in time. It's
not too wrong of me to suggest that a change in something corresponds
to a change in the propositions true of it -- or is it, Pat?
Many a book has been written on these things, in many different
fields, and I'm hoping someone out there can advise on STATEs
and OBJECT/EVENT/PROCESS from various philosophical and strictly
practical points of view. For example, we need someone who knows
something about Whitehead's Process theory. Any takers? A good book
on these things for the beginner is "Representations of
Commonsense Knowledge" by Ernest Davis, Morgan Kaufmann, San Mateo,
Yours truly, Fritz Lehmann
GRANDAI Software, 4282 Sandburg Way, Irvine, CA 92715, U.S.A.
Tel:(714)-733-0566 Fax:(714)-733-0506 email@example.com