| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
|
Ian, I am glad you think it is useful. I think one of the
problems here is that the issues seem to me to be obvious through familiarity
(perhaps), but they are not to you - so need more explanation. If we do
(eventually) manage to understand one another, I would agree with you that this
would be v. useful. My comments below marked CP3> Chris -----Original Message----- From: Ian Niles [mailto:iniles@teknowledge.com] Sent: 14 June 2002 19:37 To: 'Chris Partridge';
standard-upper-ontology@ieee.org Cc: Adam Pease Subject: RE: RE: Metaphysical choices - position.
mereology and constitution Chris, Comments below. I think this is turning out to be a
very useful discussion.
I've done much editing of previous messages to keep the overall size down. -Ian > I'm not sure that this is true. Why couldn't (part QueenElizabeth > EnglishMonarch) be true in an atemporal
sense? Even part > relations that are > blessed by Peter Simons are going to hold
between two > physical objects, > right?
If so, then, assuming that all physical objects have > a limited life > span, it would follow that all part relations
are parallel to the one > involving Queen Elizabeth. > > CP2> Sure you can have atemporal parts -
like my hand being a > functional > part of my arm being a functional part of my
body - and so > on. In fact, in > the traditional 3D view objects' parts are
atemporal - and > for this reason, > in principle, objects do not have temporal
parts. As the SUMO > documentation > says (copied from my earlier extract below) 'in
other words, > an &%Object > cannot have 'parts' which are separated in
time, such as the first and > second halves of a football game.' But what you
have here > contradicts that: > Queen Elizabeth and King George are *definitely*
separated in > time. QED they > cannot be 'parts'. I don't understand why you think the hand/arm case
differs from the Queen Elizabeth case. It is true to say now that my hand is part of my arm, but, if I had an unfortunate accident at some point in
the future and my hand had to be amputated, from that point on my hand would
not be a part of my arm. Thus, it still seems to me that all part relations
that hold between physical objects are inherently temporal. CP3> I cannot make sense of this. Some part
relations, even of physical objects are not temporal. The centrifuge (facility)
in a centrifuge pump is atemporally part of it. The various centrifuges that
are installed in the pump to undertake the work of the facility are part of the
pump for a period of time. There is clearly a difference between the two. In
this passage you seem to be referring to a hand that can be cut off and
amputated - and so is a part temporarily. There is also my hand, the functional
atemporal part of my body - which I talked about in the original example. CP3> I am also a bit lost as to what you are
arguing. I seem to recall the point you make last time was that parts are
atemporal. Now you seem to be saying the opposite, that 'all part relations
that hold between physical objects are inherently temporal'. Have you changed
your mind - or is there some continuity that I am missing? > CP2> Similarly if X and Y swap hearts in a
heart transplant, > then the heart > X was born with has a stage which is part of X
until the > heart transplant > and then the heart Y was born with has a stage
that is part > of X. It makes > no sense to me to say that these are
atemporally part of X - > surely this is > a example where they are quintessentially
temporarily part of X. > > CP2> Also as the SUMO Object documentation
says: 'An &%Object > is thought of > as continuing through time, but at any
particular time is all > there is at > that time, in contrast to something that is
thought of as > being divided into > stages (contrast &%Process).' I suspect a
typo (or > misunderstanding) - it > should read 'is all there at the time'. So
taking this and > your claim at > face value, at any particular time that the
EnglishMonarch > exists its parts, > Queen Elizabeth and King George, do also. Which
seems > counter-intuitive. I'm not sure I follow part of this comment. What difference in meaning is there between "is all there is that time"
and "is all there is at the time"? CP3> The typo seems to have re-appeared. "is
all there is that time" is true of processes. Whatever stage of the
process that is happening at a time is 'all there is at that time'. However,
the whole process is not 'all there at the time' - different (temporal) parts
of it exist at different times. As for the rest, I agree that the documentation of
'Object' and 'Process' is overly laden with metaphysical niceties that
probably have no business in a general-purpose ontology. CP3> Of course, I take the opposite view. Perhaps
the best course would be to retain the distinction between 'Object' and 'Process', but
eliminate the language in the documentation strings which implies the 3D
orientation. In other words, I think we can have our cake and eat it too. We can assume that there is, in the majority of cases anyway, a clear distinction
between objects and processes, since people by and large agree on which
is which, but we can also refuse to stake a strong philosophical claim
about what this difference amounts too.
How does this approach strike you?
CP3> It seems to me a big backward step, if you
want to try and pick up Bill Anderson's PAR 3) point. One of the prime benefits
of ontologies that I have found in my data integration work is that they help to
resolve ambiguity. If you are not doing any kind of integration - probably to
the extent of not worrying about helping people to reach a common
understanding, then, of course, it does not make sense to do the work of
regimenting this kind of stuff. However, if you want to honour claims about
ontologies helping out with data integration, then this is work that needs
doing. And the benefits of doing it are great. In my (and others) experience of
the integration of largish operation data systems - this work is useful and
sometimes even necessary. > > > And during that lifespan (part
EnglishMonarch QueenElizabeth) > > also would > > also hold as improper parts are allowed. > > I don't follow you here. > CP2>Maybe it is easier to understand if one
spells out the > indirectly stated > holdsduring - holdsduring Queen Elizabeth's
lifespan (part > EnglishMonarch > QueenElizabeth). Under most merelogies this
would, of course imply > identity - but I cannot see how this pans out
in SUMO - maybe you can. I agree that the assertion '(holdsDuring (WhenFn
QueenElizabeth) (part QueenElizabeth EnglishMonarch))' should be true in
the SUMO. I'm not sure why you think this should imply an identity between
'QueenElizabeth' and 'EnglishMonarch'. The assertion (holdsDuring (WhenFn MyHand) (part MyHand MyArm)) is, barring any bad accidents, true, but
that doesn't imply that there is an identity between my hand and my arm. CP3> Earlier you stated (part QueenElizabeth
EnglishMonarch), I added holdsduring Queen Elizabeth's lifespan (part
EnglishMonarch QueenElizabeth). In many mereologies if A is part of B and B is
part of A then A = B. If you allow part to be temporalised, you need to modify
this. I cannot see this being done anywhere in SUMO. > Well, I don't think you've established that
we're adopting a > 4D tactic here, > for the reason given above. > CP2> Firstly, see reply above. Secondly the
4D thesis is, in > one of its > definitions, that ordinary physical objects
have temporal > parts - see, for > example, "Parthood and Identity Across
Time" by Judith Jarvis > Thomson. So as > you are proposing that objects are subject to a
temporal part > relation - > this must be, *by definition*, a 4D
thesis. Thirdly, the > name 4D tactic or > X tactic is not particularly relevant, the
question still > holds. To repeat: > 'I would be interested in knowing in what cases
of > co-location you would not > adopt this position. And how this is decision
is formalised within the > SUMO.' Note - I cannot find anything on
co-location using the > SUMO browser. > I presume this means that the SUMO has not
explored the ways in which > co-location can be interpreted - leaving this
open to the people who > implement domain ontologies to do in as
heterogeneous way as > they wish. I think the last sentence expresses a fair
assessment. We haven't taken a position on co-location, and we've made the
assumption that conflicting interpretations of co-location will be rare, and
conflicting interpretations that actually result in, for example,
interoperability problems will be even rarer. CP3> This issue is commonplace. I suggest that
you look at the way in which 'tag' and facility' have been analysed in the
engineering industry. I believe Bill Anderson described a project where this or
something similar was a problem. You can also look at the EPISTLE website. You
may also wish to consider any kind of description of responsibility in
contracts , and ...., and..... I would only agree that it is rare that people
bother to analyse the problem in terms of metaphysical policies for
co-location. > > > > > (NB: you seem to have ignored completely
my arguments that > > Queen Elizabeth > > is not any of the types of agent in the
SUMO - though I agree > > with you that > > in the SUMO ontology it would be an object
and not a > > process.) > > I don't recall you giving an argument to this
effect. I > think you made this > claim, but I don't remember seeing anything to
substantiate > it.
In any > case, Queen Elizabeth would be a
'CognitiveAgent' in the SUMO. > CP2> In ordinary language we talk about
Queen Elizabeth > making a commitment, > but if you look at this in (legal) detail this
is shorthand for the > EnglishMonarch (when Elizabeth is Queen). To
see the same point from a > different viewpoint, take the X Y heart
transplant example > above. Are the > stages of X and Y's hearts also hearts. I
suppose one could > make a case for > this (and as usual a philosopher has) but it is
unusual and > counter-intuitive. Or to argue another way, is
Queen > Elizabeth during the > first decade of her reign a CognitiveAgent, or
Elizabeth Windsor as a > child - and if so, what temporal parts, if any,
do not qualify as > CognitiveAgents. Well, it may just be a fact about the way the
concepts work that temporal parts of 'Agents' are themselves 'Agents', while
temporal parts of other things, like 'Organs', are not themselves 'Organs'
(or at least not the same kind of 'Organ'). CP> And this has not yet been regimented in SUMO.
However, it one wants to be able to describe the commitments that exist in
commercial contracts it helps to know about these things. The risk management
systems I have worked on have tried to capture some of the key aspects of this.
It seems to me unlikely that temporal parts of Intentional Agents are also
Intentional Agents (one may have a better argument for Agents) . If so, I could
argue that it is 'me in 2001' that signed up for the credit card - and so has
to pay the bill, not 'me in 2002'. > > >If the SUMO > > wants to do a proper job of regimenting
the notion of a > > position as agent it > > needs to clarify what type the
position-occupying stage is. > > Well, I suggested making it a 'subrelation' of
'part' in my > last message to > you.
I'm still unconvinced that this results in some sort of > incoherent > melding of 3D and 4D perspectives, which is
what you seem to > be claiming. > CP2> This seems to be a misreading of the
point. The question > asked what the > object the 'position-occupying *stage*' was -
not the > position-occupying > *relation*. Both arguments in the 'occupiesPosition' relation
are required to be instances of 'CognitiveAgent'. CP3> So SUMO claims that every occupation of a
position creates a cognitive agent. ](documentation CognitiveAgent "A
&%SentientAgent with responsibilities and the ability to reason,
deliberate, make plans, etc. This is essentially the legal/ethical notion of a
person. Note that, although &%Human is a subclass of &%CognitiveAgent,
there may be instances of &%CognitiveAgent which are not also instances of
&%Human. For example, chimpanzees, gorillas, dolphins, whales, and some
extraterrestrials (if they exist) may be &%CognitiveAgents.")] - so my
very brief occupation of the position of director creates a SentientAgent with
responsibilities and the ability to reason, deliberate, make plans in addition
to me and the position. This is a strong claim. I would be v. interested in an
example that substantiated this. I suspect that SUMO needs some revision. > CP2>My original point was that this variety
is not useful as > it compromised > the intended use of the ontology to facilitate
interoperability of > information systems. Allowing there to be two
interpretations > with different > objects and different numbers of objects
clearly does so - > there cannot, in > principle, be a one-to-one mapping. So can I
infer that you > (now) accept > that SUMO allowing two interpretations
compromises interoperability as > described by Bill Anderson as item 3) from the
PAR. > The advantage of embodying such variety is that
it > makes the standard palatable to a wider class
of users. It > would certainly > be a shame if some users declined to use the
ontology just > because it didn't > accord with their. I'm not sure I follow you here. I guess your general claim is that one ontology (call it O1) can conform with another
ontology (call it O2) only if the existential claims that follow from O2 (call them E2) can also be derived from O1 (call these claims E1). Is that what you're saying? If so, I don't buy it. Even if O1 and O2 make different claims about what exists (i.e. even if E1 and E2 are not identical), the two
ontologies may agree about how to classify what exists in the real
world. For it may be that the class of things that actually exist in the world (or
even just the things that we are interested in describing) are a subset
of both E1 and E2. If that's the case, then the two ontologies agree for
all practical purposes, even though they conflict as regards la-la land. CP3> I do not understand how what you say relates
to the point I made. Are you saying that whenever any two ontologies (with the
same interests) differ in their commitments regarding exactly the same parts of
the world - the standard commercial data integration problem - it is because
the things that the commercial systems are describing are not real but exist in
la la land. Even if this was true, which seems unlikely, the enterprises that
own these systems seem to think that they exist sufficiently to warrant
integrating them. > > CP2> This does not seem to have been an
overriding consideration when > Teknowledge took the decision to use a 3D
ontology. > CP2> I agree there are trade-offs. However
for most > commercial projects, > worries about accommodating 'preferred
metaphysical theory' > are low down the > list compared with enabling interoperability -
unless the > preference gives > at least a comparable business benefit. I'm glad we agree about this. CP3> So we agree that it is not important to
worry about accommodating people's pet theories - it is more important to
regiment the theory that enables interoperability. |