Re: questions on assertion

Charles McCathieNevile wrote:
> On Tue, 9 Jul 2002, Giles Hogben wrote:
> ...
> >An assertion is traditionally a statement that you believe something to be
> >true.
> 
> No, an assertion is something that is stated as if it were true. An important
> tool in checking proofs and dealing with logic is the ability to make an
> assertion and not believe it.
> ...

We should distinguish two possible meanings of assertion: internal and external.  These will
have, probably, different semantics or at least conventions, associated
with them.  If you are talking about an internal event of an agent asserting a sentence into
one of its KBs, then the act need not imply that the agent believes the sentence.  As
you suggest, the agent may be doing hypothetical reasoning ("What if the wumpus *is* in
cave (3,4)?  what would then be true?").  However, the situation is different for
external assertions -- speech acts in which an agent tells another agent (or group of them)
a set of sentences.  This has usually been formalized as having the precondition
that if the telling agent asserts P and is honest, then the telling agent believes P.  See,
for example, the semantics for KQML (http://citeseer.nj.nec.com/labrou94semantics.html)
or the FIPA ACL (http://citeseer.nj.nec.com/36892.html).

I think we should develop a way of associating an agent (human, organization or software)
with a set of RDF assertion either conveyed in a message or published on a web page.  The
semantics would then be that, if such an agent is specified, than all of the RDF triples
in its scope can be taken as beliefs of the agent.  We would probably also need to tie this
to some explicit or implicit time point (e.g., the time when the message was sent or the
web page last modified).

Received on Thursday, 11 July 2002 06:04:29 UTC