[topicmapmail] PSIs - alternatives

[jackpark at thinkalong.com]

If one were to substitute the term "belief" for the term
"opinion", then we gain "wormhole-like" access to an
enormous body of literature on "belief networks" -- mostly
of the Bayesian kind. That, it so happens, has been the
topic of discussion in some of my close circles: "bayesian
subject maps". 

Consider this: much of what we are doing in the symbolic
world (and discussing in this thread) has been based on
set-theoretic notions and boolean logic of one sort or
another. But, at least one aspect of the present discussion
about PSIs and subject identity relates more to beliefs,
possibly less to boolean truth statements. 

And, bayesian maps would not totally answer all our
questions on domain understanding, as if anything we
manufacture can do that. For instance, it has been said
that set theory lets you talk about set membership, while
category theory lets you talk also about the social life of
those members. Category-theoretic subject maps find their
way into some of my other close circles. After all, our
enterprise is mapping the relationships of the universe;
the social lives of the entities we represent counts for
something.

It certainly strikes me that we have only now begun to
really scratch the surface of the issues we all face in
regards to wrestling representations of knowledge artifacts
and human thought into our maps.

I expect to have much more to say about this later.

Jack

On Wed, 28 Jun 2006 16:19:11 +0100
 Simon Grant <asimong at btinternet.com> wrote:
>At 10:25 2006-06-28, Bernard Vatant wrote:
>>Steve Pepper:
>>>
>>>My main point is that we should be extremely careful
>about keeping agreement
>>>concerning the *identity* of a subject separate from
>agreement concerning
>>>*opinions* about a subject. We should therefore avoid
>overloading the PRD
>>>with assertions, especially machine-processable
>assertions.
>>>
>>Could you explain how you think that asserting the
>*identity* of a subject can be cleanly separated from
>asserting *opinions* about it?
>
>Maybe it is possible to separate it, even if it is not as
>clean as one would like. More below...
>
>>First, asserting that a subject exists as a distinct
>entity with id-entity is by itself an *opinion*. A subject
>comes to existence when someone starts to declare
>something about it, the first declaration being : "Here is
>something distinct from everything else". This is the
>expression of an opinion, with which one may agree or not.
>
>Sure, but if one party doesn't agree with the existence of
>something, they won't have a PSI for it, and therefore
>that subject will not have to be merged in any dealings
>with that party. Hence the "opinion" about the existence
>of something has different qualities for our purposes from
>other kinds of opinion. (I don't mean "existence" in a
>physical sense, but in a conceptual sense.)
>
>>The second step is : "This subject will be identified in
>such and such a way, enabling to make it distinct from any
>other subject". Note that, as Patrick has pointed
>correctly, identification will generally occur in a
>certain context (domain context, or processing context),
>so this context is implicitly or explicitly part of the
>identity declaration, and its definition is also an
>opinion ... and so on ...
>>So, even if I agree on the principle, I don't think this
>separation can be as neat as you wish it to be (as Lars
>Marius would say : "The real world is broken"), so there
>is no point in trying to specify what the limits are, as
>Simon tried to suggest, or as we tried to do (without
>success) in the OASIS TC.
>
>Perhaps I ought to look this up. Does anyone have useful
>references? We could always check if any new approaches
>could be seen.
>
>>Seems to me that a pragmatic approach can rely on the
>following principles :
>>    * Whatever people strongly agree upon belongs to
>identity side, 
>> IOW identity is the "minimal common opinion about the
>subject".
>>    * Whatever people are likely to disagree upon belongs
>to the 
>> assertions/opinions side.
>
>That's a very appealing position. However, I'm not sure
>that it takes into account the fallibility of information
>and knowledge systems.
>
>Say that we all agreed that date of birth was a good
>attribute amongst a set for identifying people. The
>problem comes that if, say, one party has my date of birth
>as 1956-03-04 and another 1956-04-03 by mistake (neither
>is my actual one!), they might be led erroneously to
>conclude that I was two different people. Thus, one could
>argue, personal identity doesn't *necessarily* include
>date of birth - indeed it doesn't *necessarily* include
>anything: I could spell my name wrongly; I could have a
>signature or a photo that is easily confusable with that
>of someone else. Etc. etc.
>
>What I'm suggesting is that, firstly, there is no set of
>attributes which can be relied on absolutely to
>distinguish identity.
>I therefore follow Steve in thinking that in many cases
>there would be nothing that could be relied upon to be
>agreed on. Would it, in those circumstances, make sense to
>have a "bare" PSI?
>
>Secondly, despite the related problems with information
>and knowledge systems, we all have a concept of existence
>(even if it is not always co-extensive), I would say as
>not just one property among others. Physical existence is
>just one property among others, to be sure, but existence
>of a concept - well, to me that is something different.
>And this will be reflected in any authority's set of PSIs:
>a PSI which identified something for which you didn't have
>any concept at all would be meaningless (to you), and
>couldn't be coherently attached (by you) to any structure
>in your knowledge base.
>
>Thirdly, there is the question of how does one *in
>practice* distinguish between individuals. Of course we
>have to use properties related to the individual. But I'm
>happy to admit (though repeating myself) that no set of
>those properties is guaranteed to be effective for
>identification. Surely it's more fluid than that. If
>someone questions the validity of one set, we can move on
>to something else, until they are satisfied (or give up).
>
>Thus, lastly, I agree that the separation may not be
>"clean", but I maintain the separability nevertheless.
>Indeed, how can we have a coherent discussion about which
>properties define identity *without* a notion of existence
>and identity separate from all those properties?
>
>Simon
>
>--
>Simon Grant  http://www.simongrant.org/home.html
>Please continue to use my established e-mail address
>a (just by itself) (at) simongrant.org 
>_______________________________________________
>topicmapmail mailing list
>topicmapmail at infoloom.com
>http://www.infoloom.com/mailman/listinfo/topicmapmail