Re: Qualification and identity

Dear Fabio,

On Tue, Jan 4, 2022 at 10:57 PM Fabio Vitali <fabio.vitali@unibo.it> wrote:
>
> Dear Niklas,
>
> > On 29 Dec 2021, at 02:20, Niklas Lindström <lindstream@gmail.com> wrote:
> >
> > Hello all,
> >
> > I sometimes wonder, aren't there any exdurantists [1] on this list (or people from the twicely fictional Tlön [2], perhaps)?
>
> Love the reference to Borges, and I totally agree it fits here perfectly.
>
> Also, I read your proposal with great interest.
>
> First of all, let me counteract with another quotation:
>
> «With our concept-making apparatus called "the brain" we look at reality through the ideas-about-reality which our cultures give us. The ideas-about-reality are mistakenly labeled "reality" and unenlightened people are forever perplexed by the fact that other people, especially other cultures, see "reality" differently.
> It is only the ideas-about-reality which differ. Real (capital-T) True reality is a level deeper than is the level of concept. We look at the world through windows on which have been drawn grids (concepts). Different philosophies use different grids. A culture is a group of people with rather similar grids. Through a window we view chaos, and relate it to the points on our grid, and thereby understand it. The order is in the grid. That is the Aneristic Principle.»
>    -- Principia Discordia (Malaclypse the Younger & Lord Omar Khayyam Ravenhurst).

Thank you for your response and interest!

Our position in what we trust is a sensible reality can certainly
render us doubting, left with once clear terminologies and
descriptions now in a tangled mess of contradiction. (I haven't read
the Discordia, but I gather it could be a source of zen moments; I
appreciate the quote! Concepts certainly have a tendency to cloud our
judgments, while clinging to them we seemingly must.)

> Secondly, I am happy to note that there is a growing consensus that we urgently need to find a way to represent in general (i.e., without domain-specific constructs) all the types of complex situations where the basic, plain triples (s p o) are not sufficient, and where time, location and subjectivity all have a role in the correct representation of the facts.

I do wonder if there is so much consensus as a continued frustration
with the limitations inherently present in our much-needed reduction
of complexities. I agree that we continue to attempt to find a common
ground, but it seems we're often talking past each other even with
simple triples, and our attempts to augment clarity by various
qualifications (identifying ever more detail and nuance) may certainly
increase our cognitive distances, contrary to our intentions.

Not that that should be a deterrent for specialization. Domains do
grow in depth as we continue to explore them. It's just that I think
we need to acknowledge the limitations of description logic as a means
of capturing fidelity to some "state of affairs", and that explicit
contextualization should be seen as a way of delineating
idiosyncrasies, and any mechanics for increased granularity would be
most valuable if it can also be discarded when shifting context.
"Folding" indirection is one of the simplest means of doing so (as in
an event becomes a flat relation), albeit it must often be combined
with some kind of collapsing of type as well (as in the nature
(intended state) of the subject and/or object become less qualified,
more general, endurant or whatnot), which isn't always
straightforward. I think both property chain axioms and to some extent
perhaps reification are useful tools here, but only continued practice
will tell. What I do know is that rdf:value left at least me with more
to wish for. Also, my interpretation is that RDF-star *side-steps* the
issue entirely (unless "gamed"). I'm coming around to see the
practical merit in the side-stepping though.

> Third: you seem to propose to create a class whose instances take the place of the original entity that you want to constrain temporally or otherwise. Thus you plan to replace:
>
> <Abingdon> :partOf <Berkshire> .
>
> with
>
> <Abingdon> :partOf [
>   a quid:QualifiedIdentity;
>   quid:qualifies <Berkshire> .
> ]
>
> , and then apply all sorts of constraints to this new blank node.

Yes, this is the gist of the "QuID" approach.

> This blank node is assigned to a class that must be, if I am not incorrect, a subclass of whatever class is the class of <Berkshire>, in order to use it as object of the :partOf predicate. Not only that: you have to require that the object of the quid:qualifies predicate is of the same type as the object of the :partOf predicate. These two constraints apply to basically all possible types of constrainable statements, which (in my humble experience) are, basically, most existing statements.

Yes, this is the intended, conceptual constraint. Note that the "node"
(the thing) is not assigned to a class, it explicitly or implicitly
*belongs* to a class. You are correct in that the presence of the
quid:qualifies relation doesn't formally say anything about its
subject (other than that it is "more specific" in some regard than
what it is a qualification of). Nor does it *have to*, but we're kind
of left with short change here, I'll give you that. The presence of
the other predicates may or may not fill in the blanks (pun intended),
i.e. provide sufficient information to deduce e.g. that the node is in
this case "a place" of the same kind as <Berkshire> (which is blank in
the example, but certainly need not be).

> Am I correct? How can you have a class that is a subclass of any class? And how can you restrict the value of quid:qualifies instance by instance? And, even if this is possible, is it possible to express these requirements in OWL and/or SHACL?

The QualifiedIdentity isn't this subclass of any class, it is just one
(possible) base class of each particular class to which an instance of
a qualified identity also belongs. I believe there is no way to define
these rules in OWL (that would amount to defining class templates or
generics, if you allow me to use a strained programming analogy). You
may for each class of interest define rules like:

    schema:Person rdfs:subClassOf [ a owl:Restriction ;
            owl:onProperty quid:qualifies ;
            owl:allValuesFrom schema:Person
        ] .

But unless I'm overlooking some intricate pattern using e.g. OWL
restrictions and owl:hasSelf to build some kind of "meta class" here,
I don't think there is a way to formally state that all things linked
with a quid:qualifies belong to the exact same class (just as the
definition of RDF lists leave some with the very same yearning for
generics).

In fact, they may very well be essentially unsame, given that there is
this obvious difference of qualification. To go far out on that limb,
<Berkshire> could, in theory that is, belong to a perfect Platonic
class which cannot in a given world be further generalized. While that
appears nonsensical, who am I to judge? (Aren't we all qualifications
of Brahma, according to some?) My *personal* notion of identity itself
is that it is infinitely qualifiable (I won't even rule out cyclical
qualifications, if only to affirm the limit of my rationality). I
guess that effectively means that I equate it with owl:Thing; although
I'm not so formally audacious in QuID.

I believe herein lies a possibility of better semantics. In a world
where QuID would become widely adopted and effectively used, this
notion could certainly become a candidate for inclusion in future core
semantics (especially then its flexible indirection, but ideally also
this "class template" notion, I suppose). I just don't see it as
likely to happen at this time.

> Not only that: the derived entity must be different from the original replaced entity for your method to work. How do you plan to work with singletons, i.e., classes of which there must be at most one instance in the universe?

It appears that those would by definition be impossible to qualify.
They may still appear as objects of a quid:qualifies relations though
(Brahma is one prime example, I suppose).

> Just curious.

Without curiosity we lapse into void. Keep it up!

Cheers,
Niklas


> Fabio
>
> --
>
> >
> > Just the other week I wrote up this simple mixed essay/ontology/critique, spurred by my reading up on RDF-star and all the surrounding discussions. I called it "QuID - The Qualified Identity Ontology":
> >
> >     https://w3id.org/quid/ [3]
> >
> > I wasn't fully motivated to bring it up then (it isn't a finished thought product for one), but given the recent discussion on the list, I felt that perhaps it might provide some kind of perspective (or prompt someone to give me constructive feedback). I am not out to recommend this "QuID" approach over RDF-star annotations (they could be orthogonal, albeit here poised as contrasts), I just need a way to clarify on the one hand what I believe is viable regarding qualification, and on the other what I worry RDF-star is and isn't, to hopefully gather further insights here.
> >
> > The motivation for it was the impression I'm getting that RDF-star appears to be set up to be somewhat *gamed*. It appears to me (as is also currently debated) that there may be some conflation of statement and "asserted but implied event" in use cases here and there, and that this may be a potential problem going forward (at worst heading towards an httpRange-14 situation but for every triple...). I wanted to contrast this with a notion I've had for some time about the limited nature of the *identities* we use in RDF, and what that implies in relation to various forms of qualification, including this "gaming" of RDF-star, specifically its annotation form (which admittedly I'm drawn to for practical reasons, to the point of wanting to game it...).
> >
> > I also have some concerns that this pursuit of reification (becoming "triples within triples") might make the simple substrate of triples in RDF much harder to grasp. But I may be wrong there.
> >
> > Perhaps disqualifying me from the fully rational, I am actually fairly comfortable with ambiguity and broken semantics (though not necessarily with conflation); and if this "gaming" I am worried about is considered sound and as intended, I will probably continue to pursue it in some fashion (I've already gone down that route [4], if only to avoid inventing something.) It might be that annotations are "mixed quotations" [5], and the use/mention distinction cannot be readily applied. That might be a semantic bog in the making of course, but I suppose any applied semantics eventually strays from the picture and breaks cohesion. I put my graphs under names and certainly don't trust the giant global graph of the semantic web to be cohesive (one stray owl:sameAs and the entire castle comes tumbling down). It's all just maps, skewed, with varying symbols, granularities and abstractions. They're not beyond interlinking, even ontology-wise, and that's workable enough. I'm not a fan of entropy, and endless complexities (still, alas, I find myself an agent of it, time and again), but that's the game of nature, and just striving to navigate it may be enough to get by.
> >
> > I do think clarification and harmonization of (or dare I wish, even unification of) what is happening here on the quoted triple level with named graphs as a vehicle for triples and provenance would be the most valuable route towards standardization. And I do miss a clear stance on qualification (rdf:value has been around long enough without catching on, so something's amiss). As these are the papers and inks we're all working with, and we're now about to get a new kind of colored marker to work with, to clarify how these map making pieces are to be grasped together would be wise (lest we get lost in the map making process and forget to navigate our realities).
> >
> > Sincerely,
> > Niklas
> >
> > [1]: https://en.wikipedia.org/wiki/Perdurantism
> > [2]: https://en.wikipedia.org/wiki/Tl%C3%B6n,_Uqbar,_Orbis_Tertius
> > [3]: Snapshot reference of the article code for future readers of this list: https://raw.githubusercontent.com/niklasl/quid/565a15b4263398d9d20b36e2c2af5b204430d6a2/index.html
> > [4]: https://github.com/niklasl/ldtvm/blob/master/examples/Spec.md#qualified-relations-as-reifications
> > [5]: https://plato.stanford.edu/entries/quotation/#MixeQuot
> >
>

Received on Sunday, 30 January 2022 20:24:49 UTC