Re: NULLs

"JOG" <jog_at_cs.nott.ac.uk> wrote in message news:f3c8038c-1956-4325-87e5-3d1e74c2f855_at_f3g2000hsg.googlegroups.com...
> On Jan 3, 12:49 am, David BL <davi..._at_iinet.net.au> wrote:
> Those seem pretty neat definitions as far as I can tell. However there
> is then an issue as to what fully completing an extension means. Does
> the database then assume:
> (a) I possess communicated statements that represent the whole current
> truth about the world
> (b) I possess the whole current set of communicated statements about
> the world.
>
> Its a really subtle difference but its always bugged me. (b) is
> interesting in that it completely precludes a user from asking the
> database any questions about the world proper, restricting us instead
> to asking questions about propositions themselves (I said it was
> subtle...). i.e. instead of "how many employees are there", one must
> ask "how many propositions are there with employees in". This seems
> entirely sensible to myself and completely rules out 3VL, but leaves
> the user to interpret the results.
>
> I wonder whether (a) describes the assumptions of an inference engine
> and (b) that of a database.

Working in criminal justince systems cleared a lot of this up for me. If one regards a database as being something like the transcript of a trial: i.e. it is just a faithful record of testimony about the world, given by people who accurately tell the truth, inaccurately tell their "truth", or just flat-out lie, then I find it all quite easy. One focuses on the form of the facts (testimony), their internal consistency, and the reasonable derivations from the testimony. One lets go of the mistaken idea that one knows the truth from looking at the database. It is very important not to think that the database somehow *is* reality, or is even an image of reality. It is just testimony, and not necessarily true testimony. Our job is just to capture the testimony without introducing new errors, to ensure its internal consistency, and to derive only logically justified conclusions.

Roy Received on Fri Jan 04 2008 - 09:27:20 CET