[WikiEN-l] Re: Accuracy Heuristics

[Sj]

On 06 Sep 2004 11:26:00 +0200, Erik Moeller <erik_moeller at gmx.de> wrote:
> David-
> 
> choose high ratings. These articles then attain a false notion of being
> authoritative. Similarly, controversial articles might never gain such
> status because some people don't like their content.

Yes, an abiding problem with ratings.  I think ratings / flags are
mainly useful as a tool for isolating targets for cleanup, and as a
distributed vehicle for VfD-style comments (it would be nice to
separate out as metadata the review/VfD/controversial status of an
article) .  As for source annotation and fact checking :

> There's an interesting project going on here:
> http://en.wikipedia.org/Wikipedia:WikiProject_Fact_and_Reference_Check

Indeed!

> "??..??" means that this part of the article needs a source. Using CSS,
> all passages marked with "??" could be highlighted or not, depending on
> personal preferences.
> 
> or like this:
> 
>  ^+The inscription is approximately 15 metres high by 25 metres wide
>  [[Source:Behistun, p.84]]

Or [[Source:Behistun-1992|p. 84, see image caption]], where the
source: reference produces the base of the footnote/reference, and the
rest is specific information to be added at the end of the reference.


> Using a method like this, we have real semantic information about
> individual facts and can easily make statements like
> * 80% of the facts in this article have sources
> * 40% of the sources we cite are of high quality
> * Source X is used in Y articles

also
* 20% of articles with <<NPOV>> tags reference more than one source
* average # of (marked) significant verifiable facts per article

> Of course these claims themselves could be faked. But together with
> stable-revision flagging and a consensus-based peer review process

Well, this gets us closer to a zero-knowledge proof of article validity and
user reliability, as one can very directly and specifically check the accuracy
or sincerity of another's work.

+Sj+