Re: post-structuralistm and (formal) ontologies

(copying Karl, with whom I've discussed some of this before)

Jeremy Carroll wrote:
> 
> 
> I am looking for pointers for thinking about whether some ontological 
> constructs, maybe owl:complementOf, are patriarchal.

> I am wondering whether work by people like Foucault or say some 
> feminists could be used as a basis for this.

There is a worrying trend in leftist, progressiveish and feminist 
thought in which logic is seen as an extension of the penis.

I'd rather not encourage this; progressively minded people are already 
too likely to treat ideas such as "evidence" and "science" as tools of 
the phallocentric industrial-military complex. Which unfortunately 
discourages them from working to share and aggregate information about 
what's going on in the world.

I'd nudge people towards the likes of ...

http://www.amazon.co.uk/Intellectual-Impostures-Alan-Sokal/dp/1861976313 
http://en.wikipedia.org/wiki/Andrew_Potter
http://www.amazon.co.uk/MUMBO-JUMBO-CONQUERED-WORLD-FRANCIS-WHEEN/dp/0007140967/ref=sr_11_1?ie=UTF8&qid=1202129964&sr=11-1

> A sample argument, in sketch form, would be that a political affirmative 
> definition of gender would be by positive qualities of 'feminine' and 
> positive qualities of 'masculine', without an assertion of the 
> disjointness, or the definition of say female as not male.
> 
> In contrast, social structures constructed by the powerful, define the 
> powerful in-group (us) and then define the out-group as (not us).
> 
> To illustrate the point relating to sexism, at least traditionally, the 
> concept of man has being associated with a number of attributes e.g. 
> (strong, heterosexual) - with woman defined as not-man, then we see a 
> homosexual male being called 'a woman' in a derogatory fashion - which 
> is insulting to both homosexuals and women, as well as simply being an 
> error. (Although the error, as hypothesised is in the classification 
> system, the T-Box, rather than the A-Box)

What we have in our hands here is a language for making crude, cartoon 
like, claims about the world. But just as a cartoonist can fashion works 
of subtlety from a thick marker pen, on the Semantic Web we have 
choices, and can express rich information through these simplistic and 
simplifying mechanisms. Taking examples from the FOAF / Social Graph 
scene, ... we can write ":x bestFriend :y", or instead we can publish 
information about photos, collaborations, event attendence, ... the 
evidence friendship leaves in the world. We have a tool (RDF/OWL) that 
encourages oversimplification and the ommission of context. And we have 
an environment where technology developments are led by predominantly 
white, rich, middle-class, penis-wearing people from the West. But that 
said, the technology is also highly decentralist, supportive of free 
speech, annotation and alternative points of view. Sure we can define 
terms that encourage people to say dumb, false or misleading things, but 
the architecture of the system that this is done within is one in which 
those terms constitute one dictionary amongst many. The DTD / XML Schema 
approach, by contrast, is more an all-or-nothing system, and so getting 
buy-in for a minority perspective may be harder there, due to the 
difficulty of freely mixing independently developed elements and 
attributes.

I made a set of tradeoffs in drafting 
http://xmlns.com/foaf/spec/#term_gender but if others don't like that, 
they can always define, use and promote their own. Just as Perl has 
"there's more than one way to do it" as an official motto, 
(http://catb.org/jargon//html/T/TMTOWTDI.html) we might use "there's 
more than one way to say it". In general this has been a blessing and a 
curse to RDF, but its pretty central to the technology we've built so we 
may as well start making it explicit". Re foaf:gender, I imagine the 
aspect you might think most suspect there is that we declared it an 
owl:FunctionalProperty. But there's nothing to stop you writing 
foaf:gender="female male".

One important point here, that relates to OWL 1.1:

In FOAF, it is important to us to have the notion of a "Static inverse 
functional property". That is, to go beyond "inverse functional 
property" (IFP) to take a higher view about different RDF graphs, some 
of which might be true at different times. The idea is that, for 
example, foaf:mbox is a static-IFP, because if some value eg. 
<mailto:danbri@w3.org> is ever the foaf:mbox of some thing, then it is 
never the case that it can in a different graph truly be ascribed to 
some other entity's foaf:mbox. OK that's clumsily put. Another way: some 
properties can at different times have different true values. Others are 
somehow static or timeless; an example might be the dateOfBirth of a 
Person. This concept helps when one is aggregating information from 
multiple RDF graphs, collected from different points in history. The 
graphs might not all be simultanously true, yet the aggregator still 
wants to figure out which real world entity they are talking about and 
disagreeing about. OWL currently has nothing to express such 
information, AFAIK.

But if it did, OWL users would have another tool in their hands for 
over-simplification: we could say that a property like :gender is one 
that doesn't change.

However this can be done using many technologies. There was a thread 
recently on the MusicBrainz list in which some were insisting that the 
artist Wendy Carlos ought to be listed under a name she once used, 
Walter Carlos, due to those early works originally being published under 
that name. The technology base here is just SQL and a "name" field, but 
that's enough technology to express simple minded worldviews.

In other words, you don't need fancy technology to dumb-down the world.

I have high hopes that SemWeb might help people understand just how 
complex and diverse the world around them really is, by letting 
information flow to them that makes this clear. That information largely 
won't be raw triples, however; it will be - as ever - prose, photos and 
perhaps statistics, maybe prettily displayed. And as above, the issues 
around stats are well known. However they're not solely tools of The 
Man. I cling to the old fashioned belief that counting and measuring are 
treasured human inventions rather than tools of The Man. Just be aware 
who is doing the counting, how they see the world, and what things they 
might not be counting...

> I wonder whether the underlying primitives we use to construct our 
> categories (e.g. the constructs used to build the T-Box) are implicated, 
> and could do with critical review.

I'd focus more on the tools, institutions and culture around schema 
adoption. What does someone have to do to persuade the world that the 
vocabulary and it's description in RDFS/OWL (including domain name 
ownership and control) are worthy of widescale adoption? And here, the 
divisions are more to do with money, education, power, ability to speak 
English amongst western technologists and hence influence the Web 
community. Gender is also part of that (eg. have a think about who 
contributed to the works listed in 
http://ebiquity.umbc.edu/blogger/2007/09/23/top-rdf-namespaces/ ) but I 
don't think RDF/OWL is peculiarly flawed here.

Rather than frame the discussion as "is OWL patriarchal", I'd prefer to 
encourage discussion of the ways in which RDF/OWL might best be used in 
a way that promotes subtle thinking, appreciative of the world's 
complexity.

You might also want to take a look at 
http://www.amazon.co.uk/Geography-Thought-Asians-Westerners-Differently/dp/1857883535
which argues that class-hierarchy thinking is a perspective on the world 
that is peculiarly Western, rather than universal. And it does so in a 
way grounded in experimental observation.
 From the abstract,

"""When Richard Nisbett showed an animated underwater scene to his 
American students, they zeroed in on a big fish swimming among smaller 
fish. Japanese subjects, on the other hand, made observations about the 
background environment...and the different "seeings" are a clue to 
profound underlying cognitive differences between Westerners and East 
Asians. As Professor Nisbett shows in The Geography of Thought people 
actually think - and even see - the world differently, because of 
differing ecologies, social structures, philosophies, and educational 
systems that date back to ancient Greece and China, and that have 
survived into the modern world. As a result, East Asian thought is 
"holistic" - drawn to the perceptual field as a whole, and to relations 
among objects and events within that field. By comparison to Western 
modes of reasoning, East Asian thought relies far less on categories, or 
on formal logic; it is fundamentally dialectic, seeking a "middle way" 
between opposing thoughts. By contrast, Westerners focus on salient 
objects or people, use attributes to assign them to categories, and 
apply rules of formal logic to understand their behaviour."""

BTW one book (oddly enough given its central premis) that does seem 
prone to excesses of simple-minded category thinking is Samual 
Huntington's http://en.wikipedia.org/wiki/Clash_of_Civilizations ...
I can certainly imagine an OWL 1.1 appendix to that book that says

"Every :Person may have a :primaryCivilization to a :Civilization, which 
are enumerated as follows. :WesternCivilization, :OrthodoxCivilization, 
:LatinAmericanCivilization, :IslamicCivilization, :HinduCivilization, 
:SinicCivilization, :JapaneseCivilization, 
:SubSaharanAfricanCivilization, :BuddhistCivilization".

He comes slippery-close to saying that these civilizations are mutually 
disjoint. But it isn't clear from these grand, rigid-looking categories, 
what exactly is being put in them: people, countries, nations or 
strands-of-thought.

RDF/OWL certainly encourages people to get the impression categories are 
firmer than their given prose definitions and usage really justify. For 
example, during DAML we'd see occasionally a category :Terrorist or 
:TerroristOrganization in a DAML+OIL ontology. The mere fact that its in 
an ontology gives the surface feel that it has been rigorously 
("formally" as we often hear) defined. In fact it's a wooly and 
contentious and mischievously used word. A cloud disguised as a box.

A rule of thumb to promote here might well be: if you find your thinking 
on some topic can almost be fully captured in OWL statements about 
categories, hierarchies and logical membership rules, ... you're not 
thinking hard enough.

One last though re schema design (and one I was just discussing at 
http://sgfoocamp08.pbwiki.com/FrontPage ). Technologists often have a 
way of getting stuff done, by saying "ok, let's not try to perfect 
this", ... get a version out there and worry about the fancy details 
later. We do this all the time in W3C-land, but it happens with software 
and RDF vocabs too. They'll ask themselves if their efforts, in current 
state, are generally progress, maybe they'll talk about 80-20% 
tradeoffs, and they'll ship something. And hope it works for most people.

In general that's all pretty wholesome and sensible, but when they're 
creating descriptive frameworks for others to represent themselves 
through, there's a danger of cutting out those on the slopes of the bell 
curve. "Of course everyone's 'male' or 'female'" / "yeah yeah we'll 
figure out the corner cases later, ...". Unfortunately in a global 
system it isn't always so easy to change the schema later. What I think 
we need here is a bit more discipline around getting widespread review 
and feedback of schema/vocab/ontology design at the early stages, before 
it's too late to make changes.

So how do we make sure that we don't cut corners with the "corner 
cases", and that subtlety and inclusiveness in modelling isn't dismissed 
as academic detail or pedantry? That I think is more a social and 
process problem than an issue with the underlying technology.

cheers,

Dan


 > Any pointers appreciated.
 >
 > Jeremy

ps. did you see http://www.xml.com/pub/a/2001/01/31/politics.html
http://www.xml.com/pub/a/2001/02/07/politics.html
http://monkeyfist.com/articles/743 from Kendall Clark?
ps. I'm not getting Amazon points for these  bool links, they just were 
easy pointers to find :)

> 
> 
> 
>

Received on Monday, 4 February 2008 14:06:45 UTC