[sc34wg3] Reification or Representation?

Murray Altheim sc34wg3@isotopicmaps.org
Thu, 19 Aug 2004 17:33:10 +0100

[Was: [sc34wg3] "TMDM doesn't specify what is reified?" I thought it
deserved a new thread since it might be considered off-topic by some.
Due to or despite its length this is a first draft, so my apologies
for any errors.]

Lars Marius Garshol wrote:
> I think this was a helpful reply that exposes a lot of the
> misunderstandings between us, as well as some weaknesses of the TMDM
> text.
> * Steven R. Newcomb
> | 
> | What I'm missing, in the TMDM, are clear, explicit answers to the
> | question:
> | 
> |   "What are the *subjects* that are being reified?"
> | 
> | By "reify" I mean "represent via a surrogate", and "provide with a
> | proxy".  
> I think we can identify one source of confusion here immediately.  ISO
> 13250:2000 and XTM 1.0 both use the term "reify" in a different sense
> from the one it was coined for by the artificial intelligence
> community, and I think you continue to use and understand it in that
> sense. 

One of the things I've noticed over the past few years is that the
Topic Map community has (seemingly unwittingly) been delving deeper
and deeper into epistemological issues surrounding what it is exactly
that Topic Maps *do*. To look back on the KR/AI community's use of
"reification" or "representation" as any sort of correct usage would
be a mistake; they have been hopelessly muddled in their use of
language for many of the same reasons, particularly that their
community has been typically made up of engineers, not philosophers.

> There are three problems with this usage of the term:
>   a) it confuses people who know the original meaning of the term,

I would suggest that there are a variety of "original" meanings to
the term, and likely few who actually do understand the term and its
ramifications. This would seem to be important if one were planning
to build an epistemological view around it. As a simple idea of how
Idea is transformed into Thing (i.e., made real), it is deceptively
simple. I've read my fair share of Hegel (any amount of Hegel is a
fair share), remember reading Gabel's "False consciousness: An Essay
on Reification" back in the 80's, and have since read some Construc-
tionist theory, but I wouldn't claim to understand the term. I would
suggest that the majority of people in the KR/AI community likely do
not know or are confused by the term, so I'm not too worried about
confusing the already confused. I think we should focus on being
accurate with our use of language.

>   b) it scares people, because "reify" and "reification" are
>      difficult-sounding terms, and so most people go blank whenever
>      these words are uttered, and

Yes, "representation" and "represent" are less scary. They're also
very misleading, i.e., their ontological commitments are unclear to
most people. Is it better to use a rather common but erroneous term
over an unusual but rare one? How many people have a sufficient
background in semiotics to understand the ontological commitments
of representation?

>   c) it leaves us without a term for distinguishing "reification" (in
>      the original sense of the word) from "reification" (in the ISO
>      13250:2000 sense of the word), and we do need such a term.

Are you prepared to distinguish "representation" from the variety
of meanings it has? Or do as most do, just leaving people to believe
what they will?

> So for all these reasons, and also because there appears to be no
> reason to change the meaning of the term when the term
> "representation" will do what we want, TMDM changes the meaning of the
> term. This has been discussed at ISO meetings, and the current draft
> uses and defines the terms carefully, but I suppose if you have gotten
> used to your own meaning of the term for many years that isn't
> necessarily enough to clear things up.

I believe you actually mean that the TMDM will change its choice
of terms, not the meaning of any given term. We ought not be
creating new meanings for existing terms.

For me to accept that "representation" does what you want, you'd
first have to explain in greater detail what it is you believe
"representation" means, and in greater detail what it is you want
to say. Sowa heroicly spends a whole book on the subject, and I
dare say he's not particularly clear. But I'm honestly not trying
to engage anyone in a debate here, simply express an opinion. I
just think it's unfair to characterize Steve's use of reification
as somehow "his", without admitting that your use of representation
will not be "yours". You're both in the same soup.

> So, here's an attempt to do that.
>   A topic REPRESENTS its subject. The topic is an electronic symbol
>   which stands in for, represents, the real-world thing that is the
>   subject.

But I must caution you that "representation" is no better than
"reification" in terms of accuracy. If the TMRM is an attempt to
develop an underlying model for Topic Maps, then it too needs an
underlying, explicit epistemological model, a TMEM. For without
that, the community will continue to flounder along the beaches of
philosophical tides, unsure of meaning because it is unsure of how
it knows. This is the same problem that has bedevilled KR, with the
Semantic Web community seemingly even more ignorant of the issues.

"Representation" requires the concept that there is some canonical
subject to represent, the platonic idea of a subject, e.g., that
there is some canonical, platonic concept of "cat". Two years ago
I was a believer in that approach. John Sowa and many people in
the KR community seem to be definite Platonists (of course, a large
majority simply don't deal with epistemology at all, so John is to
be given enormous credit for trying). Sowa promotes the concept of
the representation of knowledge, indeed his last book has that as
its title. While John has a lot to say that is valuable, his insights
tell only one side of that story, and I might say (without criticism
to John's contributions, which are many) that they are not in the
forefront of epistemological understanding.

My own reading would suggest an approach to the issue not as one of
representation but as one of *expression*, i.e., that these
"representations" aren't statements of platonic ideals -- regardless
of their formality -- but simply humans expressing themselves, making
speech acts. This establishes the entire enterprise as a form of
social practice rather than some kind of abstracted, truth-directed
venture. One can imagine scientists not being eager to accept this
idea, as any ivory tower is rendered concrete. (Not quite so pretty,
but hopefully fewer elephants must die.)

This doesn't sit well with the KR/AI community, who like to think
there is a "truth" returned by their logical functions. To be fair,
McCarthy and Hayes published a paper in 1969, but I can't say that
either I agree with their supposition, which seems to be that "merely
undertaking to embody the philosophy in a computer program involves
making enough philosophical presuppositions [i.e., ontological
commitments --MA] to exclude most philosophy as irrelevant." [1] In
other words, we will avoid the issue because doing our work requires
us to. This has been the modus operandi in most of AI since 1969.

The Brandom line of thought is a branch of the later Wittgenstein,
since travelled by the likes of Wilfred Sellars, Richard Rorty,
Jurgen Habermas. To those who criticise expressionism is a form of
philosophical relativism, the counter is simply that it's calling it
what it is: not an outline of universal truths but a set of human
statements of belief about reality. In moving away from belief in
universal truths towards a understanding of this as a social practice,
moving away from Descartes' "representation" towards "inferentialism",
the idea is (quoting Brandom) "the semantic content of a sentence is
determined by the norms governing inferences to and from it, and the
idea that the distinctive function of logical vocabulary is to let
us make our tacit inferential commitments explicit."  I would suggest
that we in the Topic Maps community have inadvertently found ourselves
trying to perform this same task. I believe Steve N. has been searching
for the right language, but unable to find the exact terms since
(a) terminology in KR/AI is hopelessly muddled, and (b) the field of
epistemology (from whence the KR/AI steals its language) is itself
currently undergoing a revolution, a social-ist revolution. To suggest
"representation" is to simply substitute for Steve's version of
"reification" a now-outmoded (at least to current epistemology)
concept of "representation", certainly a step backward.

Rather than consider Topic Maps (or for that matter, our latest
buzzword, "ontologies") as statements of universal truth functions as
we find in "representation", we are better off looking at them as part
of a greater social practice of "expression", of making statements.
Then, the idea of Topic Maps (and computer-based ontologies) becomes
separated from the Descartes model and more firmly grounded in the
originating concept of ontology, i.e., as a statement or set of
statements regarding what we know about a domain of knowledge, with
emphasis on knowing (i.e., or of believing) as a human endeavour
rather than some revealing of underlying truths. This also avoids
any sense of a One Ontology To Rule Them All (OOTRTA). We instead
exchange Topic Maps as communiqués, as statements of shared
understanding, grounded in the purpose for which they are developed.

I'm not suggesting any new term here, just that "reification", if
thoroughly defined within a framework of a greater epistemology,
is a more accurate term than "representation", which while perhaps
friendlier, is also because of its familiarity more misleading.


[1] Some Philosophical Problems From the Standpoint of Artificial
     Intelligence, John McCarthy and Patrick J. Hayes. Stanford
     University, 1969.
Murray Altheim                    http://kmi.open.ac.uk/people/murray/
Knowledge Media Institute
The Open University, Milton Keynes, Bucks, MK7 6AA, UK               .

    "[The US Dept. of] Justice offered up to $67 million to ChoicePoint
     in a no-bid deal for computer profiles with private information
     on every citizen of half a dozen nations. While the September 11th
     highjackers came from Saudi Arabia, Egypt, Lebanon and the Arab
     Emirates, ChoicePoint's menu offered records on Venezuelans,
     Brazilians, Nicaraguans, Mexicans and Argentines. [...] What do
     these nations have in common besides a lack of involvement in the
     September 11th attacks? Coincidentally, each is in the throes of
     major electoral contests in which the leading candidates have the
     nerve to challenge the globalization demands of George W. Bush."
                                  -- Venezuela Floridated, Greg Palast