Hi Pat, (01)
Many thanks for that. That's made a lot of things clearer for me. I
won't play the "hairy engineer" card since although I am one I am
committed to making the financial services ontology as correct (and
interoperable) as it can be. I've been hoping to get to the bottom of
this Extensional/Intensional and 3D/4D thing for some time, and I
thought that the exchange between Chris and John offered some clues,
even though it seems I misread some of them. (02)
The short answer to your question is that I have been a lifelong four
dimensionalist in terms of how I see the world (the clue is in my domain
name :-) ). However, as such, it did not occur to me that an ontology
view which considers a continuant thing with a past and a future, would
be called 3 dimensional. The past and the future are, after all, to do
with time. So my use of some of these terms may have been different to
how others use them. (03)
In that sense, it seems I am one of the missing 3-dimensionalists I was
looking for. The ontology I am building for the EDM Council does indeed
have continuants and occurrents. Recently, we have started work on the
kinds of terms which vary over time - things like that variable interest
rate I was talking about. I am still trying to work out if I have found
the best way of identifying the facts about these. For example, I have
defined an archetypal class of "Thing" which is a time-varying
parameter, from which we have derived things like market interest rate,
debt interest payment amounts and percentages, amounts of capital
outstanding and so on. I started off with the idea that for each such
parameter, there was a "Dated Value", which had indefintely many past
values, a present value and (for some parameters) a projected future
value. However this makes for very unreadable models. (04)
To me, the art of ontology is the art of not designing something, which
is why I am wary of baffling ideas that can't be presented to business
domain experts in terms of logic and set theory. If the real world as we
understand it is such and such, then the challenge for me is to find the
most straightforward way of representing such and such in terms of
things and facts (classes and properties, to use the language I try to
shield the domain folks from). So if something that varies over time has
past values and a present value, that's what it has. In the data models
that correspond to these kinds of thing, a market data provider would
supply a "time series" of data points representing those values on
different days or at different times. (05)
So in my simple view of the world, there are things which are and things
which happen. However anyone who knows that world knows that the things
which are, may enter different states during their life, or (if the
thing is a parameter) have different values at different times. These
facts are all fundamental to financial services data. (06)
So yes, it did not occur to me that this was known as a "3 dimensional"
view of the world, but I will try and respect that terminology,
especially as it is clearly distinct from a different way of
representing 4 dimensional reality. Thanks for the pointers. (07)
Anyway, some more thoughts interleaved... (08)
Pat Hayes wrote:
>
> On Aug 7, 2009, at 12:31 PM, Mike Bennett wrote:
>
>> John and Chris,
>>
>>> God rarely testifies on the witness stand.
>>
>>
>> I think the insurance companies are still looking for him.
>>
>> Seriously though, I've been trying to get to the bottom of this business
>> about 3D/4D incompatibility assertions, and the apparently related issue
>> of extensional versus intensional definitions and why they go together,
>> and I'm picking up clues in this exchange but would like to know if I've
>> understood it right. Is it about the ability to define a state of being
>> intensionally, versus having to go back in time to an event in order to
>> define that state of being extensionally?
>>
>> I've always assumed that a definition is (by definition!) intensional,
>> whether that definition is a written one or a set of logical statements
>> about what it is to be a member of that class. If an extensional
>> definition is of the form "all the States in the USA", then surely you
>> are describing an extent, not defining a thing (the definition would be
>> framed in terms of what it /means/ to be a State). In any case I wonder
>> if that sort of extensional description can be framed for every kind of
>> Thing? Similarly if a role type has an extensional definition like "all
>> the people who are customers of X", then the state of "being a
>> customer" would require an intensional definition - does that put the
>> concept of "being a customer" outside the scope of what can be defined
>> in an extensional ontology? Unless, perhaps, you go back in time and
>> observe them /becoming/ customers? Is that the issue here?
>>
>> I'm trying to work out if this is where the perceived need for a 4D
>> ontology comes in. I've yet to see a 3D ontology.
>
> In the sense we are (I hope) using it here, OBO and DOLCE are both
> examples of 3D ontologies. In fact, if anything, this style is more
> common than (what we are calling) 4D. (09)
See above - this was not the sense I was using, but I appreciate the
difference now. (010)
> Both of them treat time, of course, and things being extended in time:
> but the 3D style insists that some physical things, called
> 'continuants' - roughly, objects as opposed to events - are inherently
> 3D in their nature, have no temporal extent or temporal parts, but
> rather retain their identity *through* time. These ontologies also
> typically admit 'occurrents' which are more like 4D things, in that
> they have a real temporal extent and can have temporal parts; but such
> ontologies insist upon a sharp, exclusive top-level distinction
> between these two categories, which are thought of as 'embedded' in
> time in different ways. In particular, an occurrent is inherently a
> 3-D entity which 'continues' to exist as time passes, and is *not* a
> 4D entity. In Simon's phrase, it is wholly present whenever it is
> present. This means that one is obliged to distinguish for example Pat
> Hayes the continuant, who is typing this message *now* and is the same
> Pat Hayes who will exist when you are reading it in the future (the
> present when you actually read it, of course), from a very similar
> entity which might be called Pat Hayes' life-history, which is
> genuinely extended in time and has parts called Pat's childhood, Pat
> on his sixteenth birthday, etc..; this latter being an occurrent
> rather than a continuant. At any moment in my life, there is a rather
> intimate connection between me-the-continuant and what might be called
> the time-slice of my history at that moment; but, importantly, this
> relationship cannot be identity, since two of these history-slices at
> different times are necessarily distinct things, whereas I am the very
> same - identically the same - person-continuant throughout my life
> (even though, of course, my properties may change: which is why, in a
> conventional formal ontology, they must be treated as what McCarthy
> called *fluents*, ie as having an extra temporal parameter.) (011)
Thanks - this is something I want to try and get to grips with. Fluents
sounds like an interesting concept. Aren't all continuant things fluent
though, in the sense that virtually all things are capable of changing
through time, apart from occurrent things, which just happen. (012)
For example a financial security, which we treat as a "continuant"
thing, clearly has a lifecycle. I've tried to model some basic concepts
of lifecycle, but as you suggest the connection between the phases in
the lifecycle of the thing, and the continuant which represents a
security that has a security master file record about it, is not an easy
one to make. An interesting challenge I'm looking at at the moment is
the formal terms attached to a fund unit, before that fund unit is
issued (in many cases it may never even be issued). Normally we model
terms as legal terms, which are facts about a legal contract, and all
financial securities are some kind of contract. However, before it is
issued, it doesn't exist as a contract. (013)
So it seems to me that there are phases in the life of any continuant
thing (i.e. anything which "is" as opposed to something which
"happens"). The security has the same identifier for most of its life
(though not before it is issued, and sometimes there is "grey market"
trading in things before they are issued, using temporary identifiers).
Like the Pat Hayes who wrote this email, they have the same identity for
most of their life, much as the river we never step in twice is still
called the Thames (I stepped in it once and believe me you wouldn't step
in it twice). (014)
Therein I think is an important feature of this "Continuant" concept,
and one which I'm sure people smarter than myself have already got to
grips with. (015)
>
> A lot more has been written on this idea, but I hope this is enough to
> persuade you that the notion is not entirely incoherent, though IMO it
> is severely flawed and the 4D framework is superior. (016)
Agreed - now that I realise this is what is meant by 3D, it is not
incoherent, but suffers from the sort of challenges you have outlined
here. I need to think about this some more, but it seems to me that as
long as we are talking about things which have some identity or "being",
surely it's a matter of adding and refining the level of detail in which
we choose to try to represent things along that dimension we call time.
Whether we look at slices of what is defined as a variable parameter, or
phases or states in the life of a thing such as a business entity, a
security or a person - at what level of refinement does this become a
"4D" view, given that we are presumably adding facts about things in any
and all of the dimensions that are deemed relevant to a particular
application domain, be it time, mass, charge, length or whatever? Is
there some leap that has to be made when we move to the 4D view, or are
we adding more facts about the thing until we end up with what is a de
facto 4D model? To put it another way, are they fundamentally incompatible? (017)
> My point in reply to Chris was only that strict extensionalist
> identity criteria can be used in either this or a 4D ontological
> framework, and indeed can be translated between them: essentially the
> same criteria can be stated using either ontological framework. (018)
That's what I would have thought. I had begun to think from the way the
conversation was going that there were some different definitions of
extensional / intensional in play, and I'm glad to see I was mistaken. I
certainly don't see how the use or otherwise of extensionalist criteria
have anything to do with the dimensionality of a model. Come to think of
it, I don't see how something which is extensionalist can even /be/ a
model. (019)
>
>> The world as we know
>> it has things which have an extent in space and time, and have mass,
>> charge, temperature and so on - the basic dimensions of the "good
>> enough" Newtonian physics we all learnt at school. So that's what I'd
>> expect to model in an ontology.
>
> But not all ontologies need be about the physical world; and even
> those that are, need not conceptualize it the way that physical
> sciences do. As many folk in this forum will attest, our own human
> language carves the world up in ways that do not make scientific sense. (020)
Indeed so. Many of the things we need to model for financial services,
or in many other business domains, are essentially made of "information"
as a kind of physical quantity. I have been wondering whether it would
be better to define things made of information as a partition in their
own right, distinct from say concrete and abstract things, to cater for
things like contractual rights and obligations, equity, debt, accrual of
interest and so on. However, this adds little to what the domain experts
need to think about - the main point is whether or not we can "ground"
the meanings in terms of legal obligations and the like, which means
tying them back to legal and financial primitive terms. Once we've done
that sort of grounding, then we don't need to think about issues like
mass and electric charge, however the time dimension really is seen in
the same way as we see it for physical things. A debt might not weigh
less over time, but it consists of less money, which is very analogous
to that. A share has a price now, a price an hour ago, and a price at
close of trading yesterday, for example, and interest has an amount
which accrues every day. So we have a set of problems which is similar
to, or perhaps in some cases a super-set of, the problems which exist in
the physical sciences. That is why for example I've been keen to try and
identify those facts about units and measurement which are not unique to
the physical sciences, as distinct from those which are. The various
"time" related terms in the FpML standard for derivatives messaging for
example sets out a wider range of time related concepts that would be
found in a basic ontology of time itself (perhaps more related to the Q1
through n that have been talked about here). (021)
>
>> If a given application is only
>> interested in things at the current point in time, then the implied
>> ontology of that individual application will clearly be a 3 dimensional
>> (or less) ontology. But as soon as we start to define an ontology that
>> can work across a whole business unit or supply chain, one would want to
>> start asserting facts in different dimensions, especially time, in order
>> to disambiguate terms.
>
> Of course; but (3D plus time) need not mean 4D. One could characterize
> the 'continuant' framework in fact by saying that it effectively
> refuses to allow continuants to be embedded in time *considered as a
> dimension*. In this, I would add by way of keeping an open mind on the
> subject, it is in common cause with some current theoretical
> physicists, who have been casting doubt on the very notion of time
> being a dimension like space: see for example Lee Smolin's
> entertaining book "The Trouble with Physics". (022)
That is very interesting. I personally would never refuse to allow
continuants to be embedded in time "considered as a dimension", since
that's what time is, for all practical purposes. Does that mean that my
use and definition of continuants is different from that of these folks?
Like I said, I just want to look at the way the world is, and then put
that down. I am always suspicious of any framework that requires some
additional cleverness on the part of the user - at the very least, it
pushes business subject matter experts out of the process of reviewing
and validating the business ontology. To me, being understandable is a
fundamental requirement of an ontology for the development of any system
or data models. Eschew obfuscation, as they say :-) (023)
>
>> For example, a bond has an interest rate, and as
>> far as one application is concerned, that interest rate is "obviously"
>> (i.e. contextually) the interest rate right now, while to another
>> application it's just as "obviously" the interest rate defined when the
>> bond was issued. For a bond with a variable interest rate, one of those
>> is a number and the other is a formula for determining that number.
>> Creating an ontology that goes beyond one part of the business workflow
>> obliges us to take account of the time dimension in order to
>> disambiguate terms that were meaningful only in the implied 3D
>> ontologies of individual systems.
>>
>> If you set out to define a comprehensive model of "how things are" then
>> I would argue that you need some kind of lattice of basic types of thing
>> that partition the model - continuant v occurrent, independent v
>> relative and so on - and I accept that these need not be the same
>> partitions or that any one partitioning is "right", though the above two
>> are pretty fundamental even if they are not explicitly stated in a
>> particular ontology.
>>
>> There is a question of terminology as to whether these top level
>> distinction are framed as "things" (Continuant Thing and so on) or
>> "partititions" in the UML sense (Continuant v Occurrent maps neatly to
>> the UML Structural versus Behavioural partitions), or as something else,
>> representing a kind of ontological commitment or a theory, and I think
>> people sometimes talk at cross purposes about what sits at this top
>> level of most ontologies. In defining an ontology with a Continuant
>> partition, one is making an ontological commitment to the existence of
>> classes of "Thing" which exist at some period in time and therefore
>> necessarily have a beginning and an end
>
> No. A continuant can be eternal. (024)
Agreed. I was being hand-wavy and vague there. (025)
>
>> - and therefore of course a
>> commitment to the 4th dimension.
>
> No again. Time need not be a *dimension*. In the case of a continuant,
> it is logically required to NOT be a dimension: continuants *do not
> have* a temporal extent, by definition. (026)
Real things have a temporal extent, so I still struggle to see how
things that are real can be modeled as some kind of thing which has some
such restriction placed upon it which is clearly not placed on the real
thing in question. But it's not a framework I'm asking you to defend, of
course. (027)
Time need not be a dimension?... As an instinctive four dimensionalist
I am struggling with that. My admittedly pragmatic definition is that a
dimension is that along which we measure some variable which is
orthogonal to all variables measured along any other such. In the
Newtonian world that most business systems inhabit, time is something
along which we measure days, seconds and so on, which makes it as much
of a dimension as length, mass, charge etc. Maybe time need not be
/modeled/ as a dimension, I can see how that might be. But change
happens, and time is what we define change against, and most systems
(and all business units i.e. anything across multiple systems) will need
to deal with that, so there would need to be ontology concepts which
deal with the reality of change over time. So I am very wary of the
concept of continuant that you are describing here. (028)
Incidentally I'm not at all interested in modeling the full range of
human concepts and perceptions, just the range of concepts that
businesses inhabit. I realise others have different ambitions. (029)
>
>> Like people, roles, states (in the
>> USA), states (in state diagrams) etc. There are things outside that
>> partition, that do not continue to be, but simply happen - events, state
>> transitions and so on. Then you have one, mutually exclusive and
>> completely exhaustive pair of partitions. Continuant things continue to
>> be, even if they change over time, as surely they must.
>
> Careful. Their *properties* change, but they do not change, in a
> sense. They retain their identity: they are the **same** thing at
> **different** times. (030)
Agreed. That is a much clearer way of putting in - thanks. (031)
> This is not 4D: it does not even make sense in a 4D framework. (The
> same thing exists at various times but does not have a temporal
> extent? If that seems incoherent to you, then you probably think in
> 4D; but then you really should not be using the continuant/occurrent
> terminology.) (032)
It seems incoherent me. I think entirely in 4D and always have, which is
why I had never suspected that calling a thing which continues to exist,
was anything other than a 4D item of which we might have various 3D
"slices" of information in different data sources at different times. To
me it seems fundamental to deal with the relationships among those
slices, states, values and so on as they vary over time, if we are to
model anything beyond a single data source. The real question, which as
you see I am struggling with, is how to do it. I did not realise I had
stumbled into such a terminological minefield in terms of how we talk
about the real problem of how best to achieve that. (033)
>
>>
>> All of which is both 4 dimensional and intensional. By using a lattice
>> of theories, one can define the partitions of a model in terms of the
>> ontological commitments made towards the "God's eye view" as novelists
>> call it, of what are the intensional definitions we are going to give to
>> things. This would cover as many things as are of interest to the
>> application or business, in as many dimensions are are relevant. I don't
>> see any reason to deny that to the ontology modeller.
>>
>> Then for example the concept of ownership is a concept which can exist
>> within such an ontology, and can be given an intensional meaning -
>> written and/or with properties, e.g. ownership must always have an
>> autonomous, self-directing being that does the owning, and a thing which
>> is owned. You might also model the temporal event of coming into
>> possession of something, but the state can be defined intensionally
>> without that, surely.
>>
>> In other words I don't see why there is any need, practically or
>> conceptually, to go back in time to some event which brought about the
>> state so described, in order to be able to define the state as a state.
>
> I wholly agree. Does anyone maintain the opposite view? (034)
Apparently not. This was part of my trying to understand Chris' thinking
about the relationship between extensionalism and 4D. What John has been
saying makes perfect sense to me, and there must be an opposite view or
there would not be an argument? I wondered if Chris was saying that you
needed to observe some event in the past in order to define a state in
the present, but happily it seems not. Clearly ISTM a state of something
at some time (such as the state of ownership, or of owing, or whatever)
can only be defined intensionally. So I still don't know what the
opposite view of that actually is. (035)
Many thanks for the pointers on this. I still have a lot to think about
and read up on, as you can see. Hopefully it's now clearer where I was
coming from, given that I was using terminology that was based more on
my own common-sense notions of things rather than on the way specific
terms are used by others. I still have a way to go in terms of finding
the most complete and intuitive way of defining facts about a thing
which has an ongoing existence, where those facts relate to changes over
time. (036)
Mike (037)
>
> Pat H.
>
>> Sometimes it is of practical use and relevance to the target
>> application(s) that you connect events in the past with states in the
>> present or at some other point in time, for example to model the
>> relationship between a legal event or ceremony and a legal state.
>> However, I don't see any reason why this should be a necessary
>> pre-condition of being able to model the thing at all, if indeed that is
>> what is being suggested. Just as an event has a cause whether we know it
>> or not, so a continuant thing has a beginning whether we know it or not.
>>
>> Chris: Does the need for a (specifically extensional) 4D approach follow
>> from the extensional approach, and if so how does it differ from the 4+D
>> approach implicit in most intensional ontologies? For instance the Units
>> of Measure work is proceeding just fine in creating intensional
>> definitions in all the dimensions in which we can measure things.
>>
>> On a more practical note, is there some fundamental dichotomy of
>> theories that would sit /above/ the lattice of theories of a
>> multi-dimensional intensional ontology, to distinguish if from a
>> BORO-style extensional ontology? Can this theoretical dichotomy be
>> dissected in such a way that two ontologies, one in each of these
>> styles, can be related to one another, or are terms in extensional
>> ontologies not reusable? I'm thinking in terms of what sort of meta-data
>> can be defined about an ontology or its terms or partitions, that can
>> help people in different domains identify and reuse ontology material
>> that has definitions for terms that are of interest to them. If we can
>> formalise the relationship between these approaches, perhaps we can
>> formalise the required meta-terms?
>>
>> Mike
>>
>> John F. Sowa wrote:
>>> Chris,
>>>
>>> The term 'Cambridge property' was introduced by Peter Geach in the
>>> derisive phrase "mere Cambridge property" without giving a definition.
>>> Barry Miller (in the Stanford page you cited) gave a definition:
>>>
>>> A Cambridge property is the referent either of a relational
>>> predicate or of a purely formal predicate.
>>>
>>> That's fairly clear, but it's at the level of Quine's dictum:
>>> "To be is to be the value of a quantified variable."
>>>
>>> I use the term 'role type' for a very widely used type of word
>>> or concept that everybody uses daily:
>>>
>>> Author, Player, Spouse, Buyer, Seller, Pet, Employee, Manager,
>>> Contractor, Pilot, Driver, Mathematician, Brother, Sister,
>>> Student, Weed, Nuisance, Tool, Prize, Reward, Favor, Example...
>>>
>>> CP> However, I can easily give an extensional definition of them
>>>> - if the role of the definition is to capture the extension...
>>>
>>> The primary role of a definition is to tell people how to use
>>> the term in a way other people will understand. None of those
>>> words can be defined by observable properties of the individual.
>>> The same plant, for example, could be a weed or a delicacy,
>>> depending on somebody's intention.
>>>
>>> Brother and Sister are two of the easiest to define, but note
>>> that there may be only one person in the world who actually
>>> observed the relevant conditions, and you have to take her word
>>> for it. The 4-D proponents usually reply "Yes, but somewhere
>>> in that vast infinity of space and time, the relevant events
>>> really do exist."
>>>
>>> CP> "For example, if I buy a car, the car does not change in any
>>>> observable way." One can, of course, observe the process of
>>>> buying.
>>>
>>> How do you observe it? By watching someone go to a car dealer
>>> and negotiate? By watching the buyer go to an auction and
>>> raise a hand at the appropriate moment? By seeing somebody
>>> typing away at a browser connected to the Internet? Even
>>> if you saw those events, how would you know that the buyer
>>> didn't sell the car or the repo man didn't tow it away?
>>>
>>> Since car ownership is registered by the government, there are
>>> more reliable methods than observing the act of buying. But
>>> very few things anybody owns are ever registered in that way.
>>>
>>> For things other than cars and houses, it would be far more honest
>>> to use the traditional way of stating the criterion: "God knows."
>>> Unfortunately, when you need proof of ownership, God rarely
>>> testifies on the witness stand.
>>>
>>> John
>>>
>>>
>>> _________________________________________________________________
>>> Message Archives: http://ontolog.cim3.net/forum/uom-ontology-std/
>>> Subscribe: mailto:uom-ontology-std-join@xxxxxxxxxxxxxxxx
>>> Config/Unsubscribe:
>>> http://ontolog.cim3.net/mailman/listinfo/uom-ontology-std/
>>> Shared Files: http://ontolog.cim3.net/file/work/UoM/
>>> Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?UoM_Ontology_Standard
>>>
>>>
>>>
>>>
>>
>>
>> --
>> Mike Bennett
>> Director
>> Hypercube Ltd.
>> 89 Worship Street
>> London EC2A 2BF
>> Tel: +44 (0) 20 7917 9522
>> Mob: +44 (0) 7721 420 730
>> www.hypercube.co.uk
>> Registered in England and Wales No. 2461068
>>
>>
>> _________________________________________________________________
>> Message Archives: http://ontolog.cim3.net/forum/uom-ontology-std/
>> Subscribe: mailto:uom-ontology-std-join@xxxxxxxxxxxxxxxx
>> Config/Unsubscribe:
>> http://ontolog.cim3.net/mailman/listinfo/uom-ontology-std/
>> Shared Files: http://ontolog.cim3.net/file/work/UoM/
>> Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?UoM_Ontology_Standard
>>
>>
>
> ------------------------------------------------------------
> IHMC (850)434 8903 or (650)494 3973
> 40 South Alcaniz St. (850)202 4416 office
> Pensacola (850)202 4440 fax
> FL 32502 (850)291 0667 mobile
> phayesAT-SIGNihmc.us http://www.ihmc.us/users/phayes
>
>
>
>
>
>
> (038)
--
Mike Bennett
Director
Hypercube Ltd.
89 Worship Street
London EC2A 2BF
Tel: +44 (0) 20 7917 9522
Mob: +44 (0) 7721 420 730
www.hypercube.co.uk
Registered in England and Wales No. 2461068 (039)
_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/uom-ontology-std/
Subscribe: mailto:uom-ontology-std-join@xxxxxxxxxxxxxxxx
Config/Unsubscribe: http://ontolog.cim3.net/mailman/listinfo/uom-ontology-std/
Shared Files: http://ontolog.cim3.net/file/work/UoM/
Wiki: http://ontolog.cim3.net/cgi-bin/wiki.pl?UoM_Ontology_Standard (040)
|