[sc34wg3] SAM-issue term-subject-identity

[Ann M Wrightson]

If you want to skip the rhetoric, go directly to point 5.

1. Yes, the term is needed, since it is a "co-ordination point" for the
various arguments about when topics should be merged - including TNC, since
the argument there is:
two topics having the same name in the same scope, is sufficient evidence to
consider their subjects to be identical.

2. The notion of subject identity should be carefully and explicitly
defined, since this definition determines a v. important aspect of topic map
behaviour - including  the vulnerability of a topic map to rogue added
information.

3. The XTM definition is a good start, in that it mentions both being
identical and being distinct.
IMO arguments to date have focussed on the benefits of identifying subjects
which are the same (presumably from the viewpoint of the (author of the)
merged topic map), rather than on the risks of spurious identification of
subjects which are distinct. To quote Steve N. at XML europe: "It is the
responsibility of any topic-map-aware node to be careful what it merges"

4. With reference to the common position that topic maps are bunches of
assertions, in general I disagree (I think that they are bunches of items of
information) - except for this case, where an assertion is being made, in
the course of merging, that some-2 subjects are identical.

5. Here's my candidate for a sufficient case for considering two subjects to
be identical:

Two subjects A & B are identical if (and only if): considering all
information pertaining to A to pertain to B, entails no loss of ability to
discriminate distinct situations in the "real world" (that is, the
perceived-world of the machine-or-human agent doing the merge)
In this definition, an item of information pertaining to A is, crudely, a
topic characteristic of A; "situations" are bunches of items of information,
i.e., when modelled in a topic map, collections of
topic-characterstics-on-a-topic.

Notes:

I suspect that this definition sits more comfortably at RM level than SAM
level, and may yield a derived concept for use in SAM.

It also highlights the potentially-radical danger of merging topic maps
which have been constructed using concepts from different world-views. IMO
most of the risks can probably be characterized using classical analyses of
fallacy.

Ann W.