[Top] [All Lists]

Re: [ontolog-forum] Two ontologies that are inconsistent but both needed

To: "[ontolog-forum] " <ontolog-forum@xxxxxxxxxxxxxxxx>
Cc: Pierre Grenon <pierre.grenon@xxxxxxxxxxxxxxxxxxxxxx>
From: Bill Andersen <andersen@xxxxxxxxxxxxxxxxx>
Date: Fri, 8 Jun 2007 21:14:34 -0400
Message-id: <6D857996-B040-47DC-9680-372039414DA8@xxxxxxxxxxxxxxxxx>
Hi Pat...

Wow..  That was a lot of writing.  Forgive me if I fail to generate so long a discussion -- it is simply outside my ability.  A couple of comments, though.  And a pre-apology for top-posting, but it's simpler to say what I want as systemic comments on your post rather than a point-for-point response.

1) You called on Barry to consider adopting Quine's doctrine for existence.   To borrow your example from below:

(ist Morning (that (inside Fritz Bratwurst)))

By this, you are asking the user of your ontology to accept propositions, as they do fall under the quantifiers in IKL.  While I don't have a problem with that (and in fact really like that feature of IKL), this is hardly a benign, pedestrian, work-a-day kind of entity to introduce into the kinds of engineering ontologies that people actually build, cf:

Ive never seen any convincing pragmatic or 
engineering argument for insisting on this as a 
rigid distinction

But while you find the continuant/occurrent distinction distasteful, you don't seem to mind propositions, the existence of which by the Quine doctrine you are definitely committing to.

2) If the continuant/occurrent distinction doesn't make any difference insofar as our use of terms to refer to temporal phenomena is concerned, then we run into another problem.  So you don't like that distinction.  How about presentism?  Your philosophy is very close to empiricist and verificationist, so I'm guessing (and I admit I could be wrong) you will lean toward presentism.  If you want to construe say, me, as a 4-dimensional object, then presumably you don't believe in the existence of any of my future or past time slices.  Would that be fair to say? 

3) As far as "convincing pragmatic or engineering argument" goes, would not such arguments be empirical and/or anecdotal by nature?  If so, then accept the work our ontologists at OW have done as such evidence.  We use every day a bi-categorial upper level of the kind exemplified by BFO, Cyc, and DOLCE, and have found it quite useful, both in teaching and in practice.  If you'd like to see some of these ontologies, we'd be glad to show them to you.

4) Again, following Quine's doctrine, what are these things for which the continuant/occurrent distinction is incoherent?  Presumably, you'd want to be quantifying over some class of objects for which you'd like to state some axioms governing, e.g., property change over time, however that comes out in your favorite formalism.  Then, by Quine's doctrine you're committing to the existence of those things.  Let's call them Continuoccurrents.  Now you have to elaborate your theory of Continuoccurrents and distinguish them from temperatures, numbers, properties, propositions and all the kinds of other things you have in your ontology.  Doesn't sound to me like that project is any less problematic than the defense of either bicategorialism, 4D, or any other metaphysical framework.  By this I mean on a practical, engineering level.


On Jun 8, 2007, at 17:39 , Pat Hayes wrote:

An example is described here:


Hey, nice survey. Utterly wrong in its 
conclusions, but nice :-). I'm happy to welcome 
you to this debate which many of us have been 
involved in for quite a long time. (see for 

But you come to the wrong conclusion. These two 
'irreconcilable' ontologies ARE reconcilable, if 
one does things right. The basic error is to 
assume that what a philosopher means by 'exists' 
has to be rendered into the logical existential 
quantifier. That is good form, perhaps good 
doctrine, when the game is to use formal logic to 
sharpen philosophical debate; but that is not (or 
at any rate should not be) what we are trying to 
do here. The only sensible engineering attitude 
to take towards the logical existential 
quantifier is that it means "is an entity which 
can be referred to", i.e. an entity which is the 
denotation of a logical term; which as long as we 
are using a reasonably classical logic is 
essentially vacuous, of course. In a pluralistic 
ontological framework, this cannot usually be 
interpreted as any philosopher's notion of 
existence. Those notions have to be treated as 
classes or properties. Yes, existence IS a 
predicate, when there are many notions of 
existence to be considered. It has to be in any 
logic which is intended to support 
interoperability. (See the regrettably brief 
discussion at 
. Sorry, I know that to say this to a philosopher 
is like farting in church.)

After thinking and arguing about endurance and 
perdurance for longer than I care to remember, I 
have come a rather mundane conclusion which can 
be summed up as follows: the continuant/occurrent 
distinction is basically a distinction between 
*how we use names* when talking about 
spatiotemporal entities. It should not be seen as 
a fundamental ontological distinction: it is 
merely a linguistic distinction between modes of 
_expression_. Things we call continuants are things 
for which we tend to use the same name at 
different times, so it is natural to encode 
changes to their properties by attaching the 
temporal parameter to their properties and 
relations rather than to them: we write things 

(inside Fritz Bratwurst Morning)

but we don't tend to talk of Fritz having 
temporal parts. Special terminologies are used to 
distinguish these temporally-sensitive relations 
and properties: "fluents", "roles".

Occurrents, on the other hand, are things that we 
do tend to speak of as having temporal parts or 
'episodes', so it is natural to formalize 
temporally-relative talk of those entities by 
attaching the temporal qualifier to the name 
itself. If Fritz and the Bratwurst were 
occurrents, we might write

(inside (episode Morning Fritz)(episode Morning Bratwurst))


If one puts all philosophical discussion aside 
for a moment and asks for a purely formal, 
syntactic, way of distinguishing these ways of 
describing things, what it seems to amount to is 
where to attach a temporal parameter to a 
time-free assertion. One might pose it as a 
challenge: given that

(inside Fritz Bratwurst)

is true during a time-interval


invent a systematic way of encoding that fact by 
incorporating the temporal parameter into the 
logical _expression_. There are basically three 
places it can go: attached to the entire 
_expression_ (the 'ist' version:

(ist Morning (that (inside Fritz Bratwurst)))

using the paraphernalia of context logic), or 
attached to the relation symbol (the first 
option) or attached to one or more of the 
argument terms (the second option). These 
correspond respectively to the 
hybrid/context-logical, continuant and occurrent 
ways of treating time.

So, can these co-exist? Yes, of course. One can 
use both (in fact, all three) modes of _expression_ 
in a single ontology, and in a reasonably 
expressive logic (like IKL) can even write axioms 
which relate them systematically. One does need 
to use some discipline, to keep things straight. 
One has to use even more discipline to use them 
both (or all) in ways that respect the 
philosophical prejudices of all users. For 
example, if someone insists, as you do, that it 
is incoherent or irrational to talk of temporal 
parts of a continuant, then one will probably 
need some kind of mechanical check to ensure that 
no entity is ever spoken of in both temporal 
styles. Such code could be written, but I 
personally see no practical use for it, and large 
amounts of harm caused by insisting upon the 
distinction it would be there to check.

The continuant/occurrent distinction seems to be 
of no actual value in real ontology 
engineering[1]: on the contrary, in fact, recent 
discussions on this very list and on 
public-semweb-lifesci@xxxxxx seem to illustrate 
what I have always found to be the case, that as 
soon as one gets away from nice homely examples 
like Fritz' bratwurst, the distinction becomes 
more and more tenuous, intuitions regarding it 
dissolve, and the insistence on its being a basic 
distinction rapidly becomes more trouble than it 
is worth, causing long and pointless debates and 
tending, if anything, to produce new, artificial 
barriers to interoperability rather than help 
with our practical goal. The real world is full 
of entities which are both 'continuant' and 
'occurrent', both thing and process: ocean waves, 
storms, weather fronts, the Olympic flame, a 
cumulus cloud, the interior of a Bessemer 
furnace, the Krebs cycle, a tomato ripening on a 
sunny windowsill, a cell expanding because the 
sodium pumps in its membrane are insufficient to 
oppose the osmotic pressure. The list goes on and 
on: and the Brentano/Chisholm doctrine of mutual 
incompatibility forces one to make all these 
pointless and harmful ontological distinctions 
between things and their lifespans, distinctions 
which arise solely from the artificiality of this 
doctrine of ontological apartheid.

Ive never seen any convincing pragmatic or 
engineering argument for insisting on this as a 
rigid distinction. There are plenty of purely 
philosophical arguments, but then there are also 
plenty of purely philosophical arguments in the 
other direction. As you know, there are almost no 
uncontroversial, universally accepted positions 
in philosophy. Academic philosophy has no "normal 
science", does not come to widely accepted 
conclusions, and does not progress by a kind of 
accumulation of evidence, where the task of each 
new theory or argument is to account for 
everything that earlier theories have done, but 
to do so better. Philosophy is an ongoing 
argument, where professional competence is 
demonstrated by the ability to find a new flaw in 
someone else's argument (which itself might be 
the finding of a flaw in someone else's argument, 
and so on for many layers). This means that while 
almost any nontrivial philosophical position can 
be bolstered by a long list of impressive 
references, it can also can be attacked by an 
equally long list of authorities who have argued 
the opposite. This is why I have often said that 
while philosophy can be of use to ontological 
engineering, the appropriate attitude to take 
towards a philosopher should be rather like one 
adopts to a pet dog: they need to be housebroken, 
properly trained and fed well, but it is most 
important not to let them feel that they have the 
upper hand. (I personally find the 
'intuition-pump' (in Dennett's phrase) that your 
paper obliquely uses, which I tend to attribute 
to Simon's definition of "continuant" as 
something which, when present, is wholly present, 
quite unpersuasive because it is circular. If I 
have temporal parts, then I am NOT wholly present 
now. So am I wholly present now? In a sense yes, 
in another sense no. I can run my intuition 
either way.)

One pragmatic argument I have heard is that the 
distinction provides a kind of conceptual 
scaffolding, an ontological discipline which 
helps users render their intuitions more clearly 
by requiring them to think more clearly, 
basically. While this general idea certainly has 
some merit (as for example in the successful 
"Ontoclean" notions) it seems to have no real 
purchase when applied to the continuant/occurrent 
distinction, since the only purpose of making 
this distinction is to maintain the distinction 
itself. If one simply denies it then nothing is 
thereby lost: the only result is that 
distinctions, equally artificial, which have been 
produced by this splitting (such as the required 
distinction between Fritz and Fritz's lifespan) 
are themselves no longer needed. The resulting 
wave of simplification and unification rolls 
through the ontology like a kind of global 
relaxation into a simpler, and yet ironically 
more expressive, ontological framework. So the 
'discipline' which this framework requires serves 
only to maintain the framework itself: it is like 
a parade-ground exercise of marching in step.

I don't mean to argue that the intuitive 
categories of 'enduring thing' and 'event' are 
vacuous or useless. To the extent that they fit 
with ontological intuitions, and with linguistic 
usage, they are useful and important. But one can 
admit all that, and even include them as 
categories in a formal framework, without 
requiring that they constitute a rigid taxonomy, 
so that every physical thing MUST be in exactly 
one of the two categories and as a matter of 
logical necessity CANNOT be in both. Things can 
be in both, and there is no need to be concerned 
about this or try to forbid it. One can be 
noncommittal about the category. Sometimes it is 
useful to speak of temporal parts of 
continuant-like entities. I had red hair as a 
child. Why should one not be able to render that 
by speaking of the child-temporal-part of me, and 
attributing the color 'red' to its hair? If that 
treats me as a process, I am perfectly happy to 
be regarded as a process when that is useful. For 
some purposes, indeed, it is difficult to see me 
any other way than as a process (as for example 
when we learnt that I lose and gain cells at what 
might otherwise be an alarming rate.) The logical 
sky does not fall when a temporal parameter is 
attached to a continuant-like name. It is 
perfectly clear what it means, even to those who 
feel that it ought to be meaningless. One can (in 
CL) even state conditions which translate this 
form of logical description to the more 
continuant-like form:

(forall (r (x Continuant)(t TemporalInterval))(if (r (x t)) (r x t) ))

Perfectly consistent, with a clear meaning, and it works.

(BTW, I suspect that nothing in the case which 
started this thread comes anywhere close to this 
degree of complexity or intensity of 
philosophical debate.)


[1] PS. I know that your framework and Dolce both 
use it, and are both used by real people in real 
settings. But that in itself is not evidence that 
a similar but simpler framework which does not 
have this distinction in it might not be even 
more use.

PPS. Although I am doing all this emailing on 
borrowed time, this issue is important enough 
that I will make the following challenge. If 
anyone has two actual ontologies (of a reasonable 
size, in a reasonable formalism) which satisfy 
Waclaw's **criterion** below for the reasons 
outlined by Barry and Pierre, then please send 
them to me and I will undertake to produce a 
single ontology, written in CL or at worst IKL, 
which is consistent but into which they can both 
be translated so as to preserve entailments. That 
is, my ontology may (will :-) require one or both 
of them to be rendered into a different form, but 
that re-rendering will not break any inferences, 
if used uniformly. I may need a week or two.

At 08:34 AM 6/8/2007, Waclaw Kusnierczyk wrote:
The discussion would certainly be made clearer if one could support the
claims with a simple example;  e.g., **two ontologies that taken together
are inconsistent, which cannot be reduced to a single consistent
ontology, and which both are necessary to cover the needs for all
involved in modeling the domain.**

As in mathematics, illustrative examples help in understanding dry
theories.  I sympathize with Bill, and would like to see a
counterexample to what he says.


Bill Andersen wrote:
Hi John...

On Jun 8, 2007, at 01:42 , John F. Sowa wrote:

Those are two important points, but they don't exhaust all the
options.  There are many cases where the ontologies happen to have
some features that create inconsistencies, but with some revisions
those inconsistencies could be eliminated by redefining some of
the terms.  There are also many cases where the same thing is
viewed at different levels of granularity or from different
perspectives.  Any inconsistencies caused by such methods
could also be eliminated, in principle.

However, the job of eliminating every one of the inconsistencies
that could arise could take an enormous amount of effort.  Instead
of striving for a global consistency of everything, it might be
better to adopt methods that don't require global consistency.

What I was more trying to get at was the notion of identity (or
perhaps unity) for "ontologies".  In Sean's original note, he said
something like "a single ontology cannot be used".  You just gave us
a recipe for how to make (IMO) a single ontology from Sean's
"inconsistent" pieces, via the use of reformulation of his pieces to
make them consistent, or via use of some kind of paraconsistency.

That was what I was trying to get to in my original note ­ loose talk
of "one single ontology for X can't ..." is usually based on equally
loose understanding of the terms "ontology" and "can't".   Sorry I
wasn't more explicit about this in my original note.



Wacek Kusnierczyk

Department of Information and Computer Science (IDI)
Norwegian University of Science and Technology (NTNU)
Sem Saelandsv. 7-9
7027 Trondheim

tel.   0047 73591875
fax    0047 73594466



IHMC (850)434 8903 or (650)494 3973   home
40 South Alcaniz St. (850)202 4416   office
Pensacola (850)202 4440   fax
FL 32502 (850)291 0667    cell
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes


Bill Andersen (andersen@xxxxxxxxxxxxxxxxx)

Chief Scientist

Ontology Works, Inc. (www.ontologyworks.com)

3600 O'Donnell Street, Suite 600

Baltimore, MD 21224

Office: 410-675-1201

Cell: 443-858-6444

Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Subscribe/Config: http://ontolog.cim3.net/mailman/listinfo/ontolog-forum/  
Unsubscribe: mailto:ontolog-forum-leave@xxxxxxxxxxxxxxxx
Shared Files: http://ontolog.cim3.net/file/
Community Wiki: http://ontolog.cim3.net/wiki/ 
To Post: mailto:ontolog-forum@xxxxxxxxxxxxxxxx    (01)

<Prev in Thread] Current Thread [Next in Thread>