RE: log:notIncludes (conclusion?)

> > Why do people want non-mon?    For defaults, yes?    Anything else?
> 
> Defaults is one. The other reason is that it is impractical (and often
> infeasible) to specify all the necessary negative facts. This 
> is the main
> reason why NAF is used in databases. In fact, even in LP I 
> think the more common use of NAF is due to that second reason.

Yes, this is the issue of "closed" predicates (classes and 
properties), which are quite common in all kinds of administrative 
domains (such as being a resident of Boston, or being an official
document of the W3C, etc.). The extension of a closed predicate
is completely represented in the KB of the "owner" of such a
predicate. Therefore, NAF is equivalent to classical negation
in these cases.

N3 has the consruct "definitiveDocument" to handle closed
predicates, while CycL has the "#$completelyAssertedCollection"
consruct for the same purpose. There is no sensible treatment
of this distinction between open and closed predicates in
classical FOL (using circumscription for this is like breaking
a fly on the wheel). 

But there are further sources of nonmonotonicity. In particular,
nonmonotonicity also results from "defeasible" knowledge, a 
form of heuristic knowledge that may lead to contradictions
without this having the effect of completely discarding the
entire KB (as in classicial FOL). NAF can also be used to
express heuristic knowledge, which is quite common in
empirical knowledge domains.

Business rules experts also agree that there are defeasible
rules, though this has not been considered in SBVR.

-Gerd

--------------------------------------------
Professor Gerd Wagner 
http://www.informatik.tu-cottbus.de/~gwagner
Email: G.Wagner@tu-cottbus.de
Tel: (+49 355) 69 2397
Institute of Informatics
Brandenburg University of Technology 
at Cottbus, Germany

Received on Sunday, 28 August 2005 23:41:37 UTC