[sc34wg3] Subject Schizophrenia

Lars Marius Garshol sc34wg3@isotopicmaps.org
Wed, 25 Aug 2004 17:08:51 +0200

* Bernard Vatant
| In the current debate, folks seem to take for granted that subjects
| have some absolute existence and absolute properties that could be
| represented/proxified/reified/indicated (pick your choice, I don't
| mind) or otherwise "captured" in a topic map or any other kind of
| representation/proxification/reification/indication, but which
| somehow exist out there in the blue, independently of and beyond any
| representation/proxification/reification/indication or other kind of
| capture.

I agree with you that this assumption is questionable, and this is
part of my reason for wanting to just say "topics represent subjects"
and leaving it at that.

In my view of what a topic map is and does and means, when an author

  [something ...

in LTM and thus creates a topic with the ID "something", that topic is
created with something in mind (what ISO 13250 called the "invisible
heart of [the] topic"), something it is intended to represent, but
what it represents is of course the creator's notion of something. (I
guess this is what Murray meant by calling this "humans expressing
themselves, making speech acts", and if so I agree with him.)

So the subject is always in one sense a private one, and when we merge
we do it on the assumption that the subjects represented by two
different topics are, for our purposes, equivalent.

But whether this helps us decide which term to choose I am very much
uncertain of...
| So I tend to think that 
|   1. This assumption is completely wrong (although I am ready to
|      agree with a consensus viewpoint that it is at least undecidable),
|   2. To enable any kind of language, knowledge and communication, we
|      generally act "as if" the subjects exist,
|   3. Actually we always deal with representations or proxies without
|      being able to know is there is something to be represented and
|      proxified, let alone that this something have properties isomorphic
|      to the ones we have given to the proxies, 
|   4. Proxies can have a longer life than whatever they are supposed
|      to proxify. Their changing subject, if any, is better defined by the
|      way they are used than by anything else, and is notoriously subtly
|      changing over time, as Stella very well describes in the quoted
|      post.


| So I suggest to avoid any explicit reference to this questionable
| subject, 

I think the notion of a subject isn't questionable; I think what is
questionable is the idea that subjects pre-exist, that they are
universal, and that they are independent of the minds of those who
imagine them.

| and stick to the proxy level we are able to manage, by replacing:
| "Those two proxies proxify the same subject"
| by a more agnostic
| "Those two proxies are equivalent"
| (if you really believe in subjects, read "the subject is the same"
| if you like)

Actually, TMDM doesn't say anything like this at any point. It
describes merging as the removal of redundant topic map constructs and
leaves it at that.
| The best we can achieve then is to define agreement on
| - Rules under which two proxies have to be considered as equivalent
|   in a specific model : this is TMDM.


| - Language enabling the expression of such rules for any other model
|   : this is TMRM.


| - Process that may/should/must be applied when two proxies are found
| to be equivalent under such rules : this is Processing Model, which
| might be or not in the scope of the standard (I tend to think it
| should not)

It's in TMDM, and it needs to be, because you can't define the syntax
without defining this. (You can't really define the semantics of the
[source locators], [subject locator], and [subject identifiers]
properties without it, either.)

| Martin's remark tends also to make me think that two proxies can be
| found to be equivalent under certain facets, or at a certain level
| of granularity, or under specific rules, and distinct under other
| facets or at a finer level of granularity, or under other rules. Of
| course, this is questioning the very foundation of the TM
| fundamental objective : one proxy <=> one subject ...

Yep. This is pretty much Ann's idea, that there is really an
equivalence relation that forms equivalence classes, and that one
could imagine applying several different such relations to a model.

Lars Marius Garshol, Ontopian         <URL: http://www.ontopia.net >
GSM: +47 98 21 55 50                  <URL: http://www.garshol.priv.no >