Re: Assertional status of published information

On Mar 09, 2004, at 12:42, ext Chris Bizer wrote:

>
> --DS3235082598710344942
>
>> On Mar 09, 2004, at 07:11, ext Pat Hayes wrote:
>>>> 1. Graphs published on the Semantic Web are not asserted
>>>
>>> Well, the trouble is that most folk think that they are being
>>> asserted, at present. So we have to preserve this understanding of
>>> current normal usage.
> Well, I don't know. Taking everything found on the Semantic Web as 
> asserted
> seams like an invalid oversimplification to me.
>
> We are basically having the following situation on the Semantic Web:
>
> A information provider
> a) publishes information which he beliefs to be true, meaning he sees 
> the
> information as asserted

:someGraph { ...  x:thisGraph x:isAsserted x:true . }

or simply

:someGraph { ... }

and presume that unless otherwise stated, it's asserted.

> b) might want to publish information about which he isn't sure if it 
> is true

:someGraph { ...  x:thisGraph x:isAsserted x:true . x:thisGraph 
x:accuracy "30"^^x:percent . }

;-)


> c) might want to publish information which he knows is wrong in order 
> to
> achieve some advantage.

:someGraph { ...  x:thisGraph x:isAsserted x:true . }

(i.e., deliberately telling a lie)

> d) might want to quote somebody, not assuming a shared 
> conceptualization (de
> dicto)

:someGraph { ...  x:thisGraph x:isAsserted x:false . }


Note that 'x:true' and 'x:false' apply to whether the graph is or is not
asserted, not whether the statements in the graph are or are not true.

>
> A information consumer
> - wants to use the published information for something (decision making
> ...).

Can decide, based on the authoritative qualifications of the graph,
whether to take the statements in the graph into account.

> - might have a different view of the world than the information 
> provider.

Well, that's a more general problem, and one not directly IMO bound to
the issue of assertion or trust.

>
> Which is exactly the situation we are facing in daily offline life. We 
> don't
> take everything we hear as a fact, but evaluate if we belief it using 
> the
> background knowledge we have.

> Thus I think that on the Semantic Web, it's
> also the job of the information consumer to decide which information we
> wants to treat as *asserted* and that a infrastructure supporting him 
> should
> be open to a wide range of mechanisms for this decision.
>

Well, so much for beginning to become convinced...

I can't accept this last, er, assertion ;-)

Whether something is or is not true, is quite different from whether 
something
is or is not asserted as true by someone.

We may choose to disbelieve what is asserted, but it still remains 
asserted.

> Taking case a) and different world views between the information 
> provider
> and consumer, we have a situation where a statement is asserted for the
> information provider but not for the information consumer.

???

Rather, we have a case where a statement is 
accepted/trusted/believed/relevant
for an information provider but not accepted/trusted/believed/relevant 
by
a consumer -- but assertion is something else.

Assertion IMO affects whether the statements are reasoned about at all. 
I.e.
are they in the domain of discourse or at some other level.

> In case b) the
> information provider needs some mechanism to express that he isn't sure
> about the information (Statement about the graph within the graph or
> somewhere else).

Right. So the information provider qualfies the graph (statements) in 
some
manner to indicate a possible degree of inaccuracy or precision. But 
again,
that is something other than assertion.

> In case c) the information consumer needs a trust mechanism to filter 
> out
> wrong information.

Right. And the consumer uses "authoritative" qualifications, based on
the bootstrapping machinery, to select which graphs (statements) are
to be considered 'trusted' (or for that matter, relevant to whatever
application -- one could similarly use such qualifications to provide
a generalized scoping mechanism by which 'trust' is merely one kind
of scope).

> In case d) the information provider needs a mechanism to express the de
> dicto quotation, e.g. using literals.

I think we miss out on alot of functionality by being forced to use
literals and/or reification for de dicto. E.g. one might want to
reason along the lines of "what if" someone actually asserted a
set of statements -- and to do that optimally, one must have triples.

By being aware of which graphs are or are not asserted, one can move
in and out of such hypothetical "what if" scenarios without losing
track of the "real" semantic web (that ever changing virtual
mother-of-all-graphs corresponding to the sum total of all asserted
statements accessible on the web).

>
> The question raised by Patrick if a information consumer can be held
> responsible for what he publishes is on a different level, which I 
> think
> strictly requires digital signatures and PKIs.

Absolutely, but digital signatures and PKIs are merely forms of
graph qualification. Having a bootstraping vocabulary/semantics
to hook such machinery onto then allows for folks to have a clear
and explicit basis for determining that someone (a) explicitly
asserted some statement and (b) who that someone is -- which are
the key elements for accountability.

Cheers,

Patrick


>
> Chris
>
>>
>> True. It's probably most practical to allow folks to presume that 
>> unless
>> explicitly stated that a graph is or isn't asserted, that it is
>> asserted.
>>
>> My recent proposal for a bootstrapping intra-graph vocabulary reflects
>> the opposite
>> presumption, that unless explicitly marked as asserted, it's not, but
>> that's
>> easily fixed.
>>
>> Patrick
>>
>> --
>>
>> Patrick Stickler
>> Nokia, Finland
>> patrick.stickler@nokia.com
>>
>>
>>
>
>
>
> --DS3235082598710344942--
>
>

--

Patrick Stickler
Nokia, Finland
patrick.stickler@nokia.com

Received on Tuesday, 9 March 2004 07:04:58 UTC