Re: [ontolog-forum] Endurantism and Perdurantism - Re: Some Comments on

[William Frank <williamf.frank@xxxxxxxxx>]

Chris,

I do not think the issue is that there is no difference between things that happen and things that stay around for a long time.   The issue is, as you say, people have to **decide** how most effectively to CAST the concept, from among some useful set of categories, given what they are trying to do.    Only the decision is a human one, not one forced by nature. 

I am so **perplexed** by how what is easy in any rich logical representation with an expressive power even slightly close to that of a natural language, becomes a 'problem' that seems to plague ontologists.   The problems are a result of starting by choosing an unexpressive language.

just for many examples from this thread,  the excellent Mathew West says:

MW>] It sounds like you favour things like the President and such like as classes. I won’t say that you can’t do it, but I find it unattractive. I would want to be able to say that the President of the United States signed a treaty, but classes do not make good actors as abstract objects, so there are more contortions to be gone through.

*the* President would indeed be a poor class

but

'x is president of y' is a nice relation, with two roles,

by adding the implicit constraint that

at time t, for each y, there can be at most one x who is president of y 

then,

using the iota operator (that object such that - an analog to the lamda operator  for functions)

let P =

that object o such that at time t, o is the president of the United States

and use the relation 'x signs y', so that

for treaty T,

at time t, P signs T.

Tells this story, to me, completely naturally.   Not a contortion, but an 'unpacking' of the complex, dense way in which people talk.

Any such language will have from 5 to ten different categories of being, (hard to beat Aristotle's, I agree). 

And any good data model I have seen recently will let you say these kinds of things.  And, if an 'ontological language' can't say them, or makes it hard to say them, then how is the ontology going to allow us to easily *translate* between these lowly data models?

It is possible, in any such rich language, to RECAST something in one category into another.  This does not mean they are 'really' the same. or that there is 'really' only one kind of thing.  (when anybody says 'really', if you are a practical person, run.) Why would it be beneficial to recast everything into **one** category.  I may be wrong, but I don't think anyone suggested that this was a good idea.  Should we forget about money and integers and strings and only talk hexadecimal?  Given a problem domain, casting things one way or another will prove to be the most useful.   Sometimes, we want color to only be an attribute, say, talking about bears, sometimes, we want colors to be things, say, talking about the effects of colors on moods. 

So, the only 'problems' I see are that people 'code first, think later'.

Let's start by creating a Procrustean bed ,an ontological straight jacket, in which, for example, you are not allowed to speak of times, maybe no states that things can be in during a time period, (such as a bannana that will be ripe for a given period of time, then rotten, then completely gone as it was consumed by microorganisms), where we are not allowed to treat the process of ripening as much of a first class citizen as the bannana, or are not allowed to treat the relation between the bannana and the tree it is growing on as a first class citizen, where probably no instances of relations are allowed, so that we are not allowed to refer to Tom and Linda's marriage, where being the officiator of that marriage can't be treated as a role in that marriage, or where the relational type marriage can't have roles like officiator, where for some reason, there are either no sets or no composites of other things - you have to choose which,

Then we can happily say 'oh gee' we've got a real problem here -- how are we going to deal with the fact that customers come and go, and that when you say 'customer' you always mean the customer of something, and it becomes a 'problem' that the same company might be both a customer and a vendor.    This can keep us employed solving 'problems', I guess.

These are **manufactured** problems.    I see similar kinds of 'problems' increasingly discussed  among those following the fashion for 'Dogma Driven Development'.

Wm 

On Wed, Mar 18, 2015 at 8:33 PM, Chris Mungall <cjmungall@xxxxxxx> wrote:

On 18 Mar 2015, at 14:30, Pat Hayes wrote:

> On Mar 18, 2015, at 1:15 PM, Chris Mungall <cjmungall@xxxxxxx> wrote:
>
>>
>>
>> On 18 Mar 2015, at 9:23, Pat Hayes wrote:
>>
>>> On Mar 18, 2015, at 9:17 AM, Thomas Johnston <tmj44p@xxxxxxx> wrote:
>>>
>>>> Could those of you who are involved in this thing vs. event
>>>> (endurantism vs. perdurantism) exchange explain to me why, except
>>>> as
>>>> a bit of interesting metaphysical speculation, those of us building
>>>> formal ontologies should be interested in it?
>>>
>>> Because if one is using almost any conventional representational
>>> notation, from classical FOL through RDF and/or OWL, each of these
>>> positions requires one to adopt an axiomatic style which is sharply
>>> incompatible with the style the other one forces you to adopt. If
>>> your
>>> ontology encompasses both kinds of entities, then you are obliged to
>>> enforce a rigid top-level distinction between them, since all
>>> assertions about one kind have to be made in one axiomatic style,
>>> different from those mentioning the others. Basically, one style
>>> prohibits attaching a temporal parameter to entity names (because
>>> "continuants do not have temporal parts"), the other prohibits
>>> attaching temporal parameters to relations (because relations are
>>> timeless.) One sees this very sharply in the amount of work involved
>>> in the OBO ontologies to keep continuants separated from occurrents.
>>> The typical result is that many general facts have to be stated
>>> twice,
>>> one using each notational style.
>>
>> Can you give an example from some widely used OBO ontologies? I'm not
>> really aware of work specifically to keep these separate.
>

> ?? The basic ontology of every OBO ontology requires that they be kept
> separate, and the OBO discussion forums are constantly awash with
> discussions and questions about whether or not some new idea is a
> continuant or not. Just check the email archives.

I'm on many of those lists. I'm aware of many upper ontology discussions
that could be construed as time wasting. The people building the
ontologies are generally quite practical and may not be participating in
some of the more obscure discussions.

What I'm actually after is not waffle on a mailing list, but concrete
examples of a pair of classes P and C, taken from widely used OBO
ontologies, where P is a process and C is a continuant, where
maintaining the distinction between P and C involves work/effort, and
where there are pragmatic advantages to collapsing these. The advantages
may be conceptual (easier to comprehend to domain scientists), elegance,
advantages in inference, etc.

>
>> I might agree that there is sometimes a general tendency of some
>> ontologies to over-inflate the number of entities into different
>> aspects
>> corresponding to upper level categories. But the process vs physical
>> entity distinction would be very hard to avoid (or perhaps we are too
>> engrained in the mindset of biologists and bioinformaticians).
>
> If you find it useful, by all means use it. But there are many cases
> where it is hard to classify something on either side of this divide.
> Is the flame of a gas torch a thing or a process? It has aspects of
> each. How about hurricane Katrina, or a tsunami wave, or a ripening
> banana?

I have no opinion on tsunamis, flames.

But bananas I like, because your example corresponds to classes in two
ontologies I work on:

"fruit ripening" from the process branch of the Gene Ontology:
http://amigo.geneontology.org/amigo/term/GO:0009835

"fruit" from the plant anatomy ontology (continuants):
http://purl.obolibrary.org/obo/PO_0009001

(banana ripening and banana would be trivial subclasses of these)

AFAIK nobody has ever argued for merging these together. Or for
analogous merges, between anatomical structures and developmental or
physiological processes.

I mean obviously I see how this *could* be done, I have no problem
conceptualizing the banana as a 4D worm and the ripening process as a
slice through the worm, and I could probably come up with the required
axiomatization (there are subtleties, and some constraints imposed by
OWL). I have a harder time figuring out how to make the results
practically usable, and how to make my banana worms palatable to both
the ontology editors and users. I expect continuants would be
reintroduced through the back door, as some kind of 'maximal worm slice'
class.

Maybe this is just a conceptual prison we've locked ourselves into, but
if so, I'd argue this is *not* due to philosophers imposing some
categorical distinction from above, but rather through the perspectives
and modalities of different disciplines such as developmental biology
and classical anatomy and systematics.

> All of these and many more can be viewed in either way; and
> prohibiting this, forcing users to make an artificial distinction
> between the flame-thing and the flame-process, serves no useful
> purpose. The point is not that the distinction is bad, but that being
> *forced* to make it can be harmful, or at the very least,
> time-wasting.

Conversely, being denied the framework to make useful distinctions may
be harmful and time-wasting.

But this is all rather hypothetical. If there are particular cases in
the OBO world where we are wasting time making spurious distinctions
between physical entities and processes, I will be with you all the way
in seeking to eliminate this.

>
>> It may
>> _seem_ to be the case that facts are stated twice: for example, the
>> GO
>> ontology of biological processes (Occ.) will state about classes
>> 'limb
>> development', 'forelimb development' and 'hand development', whereas
>> an
>> anatomy ontology will also have facts stating facts about classes
>> 'limb'
>> 'forelimb' and 'hand'. Similarly GO will include occurrent
>> hierarchies
>> of biochemical reactions and signaling pathways, and these will be
>> related to but modeled distinctly from continuant ontologies about
>> molecules and protein complexes and cell parts. What may appear to be
>> redundantly stated facts are actually inferred from one hierarchy to
>> another.
>
> I know that this all does actually work, and if one has become
> comfortable thinking of everything in this strictly segregated way
> then it may even seem obvious, but there are many ways in which the
> strict segregation is harmful. Other ontological standards use a basic
> framework which treats all spatiotemporal entities uniformly as far as
> their temporal description is concerned. Each, I am sure, seems as
> natural and motherhood and apple pie to those who have lived within it
> for an extended period.

I have absolutely no problem with this, and I am happy for people to
model their domain in whatever way is most appropriate. I just responded
because you said we were doing work (implied 'busy work') in the OBO
world maintaining a separation. If we can do less work and get a leaner
and more intuitive model then I'm all for it. But it will take concrete
examples to convince those of us doing the work. We need more than a
handful of edge cases like flames and tsunamis.

>> I'm all for collapsing distinctions and reducing the number of
>> ontologies. I can imagine syntactic tricks that would allow classes
>> from
>> one category to proxy for classes in another, e.g. limb for limb
>> development, or vice versa.
>
> The fact that you insist upon treating these as fundamentally
> different suggests to me that you may be entirely happy in one way of
> thinking (not meaning to be ad hominem). It is possible to think
> differently, however. I think it is notable that apparently many users
> of OBO do not find the notion of continuant at all obvious, and indeed
> have extended debates about it, some of which are only finally
> resolved by an ex cathedra stipulation from Barry Smith.

Right, I have no problem with the statement that users find notions with
strange labels like 'continuant' unintuitive. But I think it's a leap to
go from this to the fact that users find the distinction between a
physical entity (with a certain set of properties such as mass, and
spatial parts) and a process (with properties such as rate, and process
parts) unintuitive, or to state that this distinction is unnaturally
forced on them from above by philosophers. (Personally, I think the
"picture" vs "movie" distinction is the only clear distinction in any
upper ontology, sometimes I'd happily throw the rest out).

I mean we could try an experiment. You could make a request in CHEBI
(the OBO chemical entity ontology, continuants) for "phosphorylation"
using the tracker here:
https://sourceforge.net/p/chebi/curator-requests
After all, phosphorylation is just part of the life of a molecule, and
the distinction between process parts and spatial parts is an
unintuitive one foisted on us by philosophers.

It won't be a philosopher that rejects the request, it will be a
biochemist.

> At the sub-cellular level, classifying between physical entities and
> processes becomes close to impossible to do on a principled basis, as
> far as I can see. Is a molecular pool (in SBGN) a continuant or an
> occurrent? (Does the question even make sense?)

Well I guess if you want to be literal the question doesn't make sense
in that SBGN defines squiggles and shapes on a diagram. And what the
diagram depicts exactly is not as straightforward as it might seem. But
I'm not sure what the difficulty here is. Most pathway models, including
SBGN boil down to bipartite graphs, with nodes in either the physical
entity category (e.g. ATP, or pools of ATP) or event/process categories
(e.g. phosphorylation). The node pools in an SBGN process diagram (or
corresponding BioPAX or SBML model) would typically be annotated using a
continuant ontology like CHEBI, and the transitions/events/processes
using GO or a vocabulary of reaction types. No one has a problem with
that as far as I know, and no one wants to start mixing these things up.
There is no huge effort to maintain artificial distinctions here, as far
as I'm aware. I mean in SBGN it's called a process diagram, it has
different properties than a structure diagram used to depict a chemical
structure (continuant).

>
>> But I don't think anything would be achieved
>> here, other than completely baffling users. In addition, when you get
>> down to the level of cells and below there are some subtle issues
>> involving time and identity that could confound the use of the
>> ontologies in query answering.
>>
>> Perhaps I just have a hard time envisioning a different model. As I
>> say,
>> I believe the distinction is fairly hardwired into people's thinking.
>> I'm not talking about ontologists here. Look at any datamodel that
>> captures something like biochemical or signaling pathways, at its
>> core
>> you will see a bipartite graph of physical entities and occurrents.
>> I'd
>> be genuinely interested in seeing a non-handwavy attempt to model
>> things
>> differently.
>
> With some colleagues I am trying to do this exactly on cellular
> pathways in oncology research. It is natural to write equations
> describing qualitative dynamics of chemical pools (abstracted from
> several sources including SBGN graphs) and then to instantiate the
> terms in such equations with temporal arguments when required, so one
> can speak of the quantity of-MEK hydrogenase at a certain time by
> simply adding a temporal argument to the relevant term denoting
> quantity of MEK-hydrogenase. All of which is elegant, natural and
> formally correct, but violates the OBO ontology conditions, since
> these quantities are classified, for reasons which are entirely opaque
> to me, as continuants.

I couldn't care less if it violates current OBO ontology conditions. If
your formalism is elegant, natural and formally correct then I'm far
more interested in that. I'm genuinely interested to hear more, either
off list or after you're ready to publish on it.

>
> Pat
>
>>
>> _________________________________________________________________
>> Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
>> Config Subscr:
>> 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 join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J
>
> ------------------------------------------------------------
> IHMC                                     (850)434 8903 home
> 40 South Alcaniz St.            (850)202 4416   office
> Pensacola                            (850)202 4440   fax
> FL 32502                              (850)291 0667   mobile
> (preferred)
> phayes@xxxxxxx       http://www.ihmc.us/users/phayes

_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/
Config Subscr: 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 join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J

_________________________________________________________________
Message Archives: http://ontolog.cim3.net/forum/ontolog-forum/  
Config Subscr: 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 join: http://ontolog.cim3.net/cgi-bin/wiki.pl?WikiHomePage#nid1J    (01)