[WikiEN-l] Improve quality by reviewing all new articles

Fri Dec 16 14:08:42 UTC 2005

On 12/16/05, Jimmy Wales <jwales at wikia.com> wrote:
> Anthony DiPierro wrote:
> >>Yes, it won't necessarily lower the workload, but it would certainly
> >>improve the quality of our articles.
> >>
> >>Mgm
> >
> >
> > How?  The quality of the articles is exactly equal if someone puts
> > them into a queue and you delete it or someone creates it and you
> > delete it.
>
> Without making any comment on the merits of the specific proposal at
> hand (it sounds great to me, but then again I haven't thought it through
> enough yet), I can answer this question...
>
> The question is what happens in the case of an overlooked article?  What
> happens when the New Pages Patrollers get behind or there is a gap in
> coverage?
>
> The default failure mode of the patrolling process today says "Anything
> we don't look at, we assume is good enough to go on the site" ->
> Seigenthaler.  The proposed failure mode of the patrolling process says
> "Anything we don't look at, we leave it in the queue until someone gets
> around to it."  If the queue of new articles by newbies is time-stamp
> sorted, then we will often be only a few minutes behind, sometimes maybe
> an hour.
>
> Now, this need not necessarily be done with a queue or a gateway model.
>  Another idea in the same general area is the long-desired 'check off'
> model for collaborative RC/new pages patrolling.
>
> --Jimbo

You're right in theory, but in practice I don't think new articles are
coming in fast enough that there are some that no one looks at at all.
 Someone *did* look at the Seigenthaler article, it's just that the
mistakes were subtle enough to not be obvious.  I don't know if it
contained any references or not, but it certainly could have, just not
references for every single fact.  If you're going to require
*extensive* fact checking before releasing something from the queue,
then I could see the purpose.  But this proposal was about "basic
quality standards", standards which the Seigenthaler article
apparently met (I haven't actually seen it, I'm going by the
statements of you, Seigenthaler, and others, here).

A "check off" model might help a little, but it would be rather easy
to search the database for articles which were created and not edited
by some subset of "trusted users" within a certain timeframe.  This
would accomplish pretty much the same thing as a check off model, and
it'd allow for different people to have different definitions of who
is trusted.

Say you defined "trusted user" as a user who has at least 500 edits. 
It would be no problem to present a list of articles that haven't been
edited by such a "trusted user".

Anyway, I guess a properly implemented queue wouldn't be a bad thing. 
For instance if "publishing" and "unpublishing" were a simple toggle
switch.  But I see it as rather unnecessary.

As an aside, would the user who clicks the "publish" button find
herself legally liable for the content of that article?  I certainly
could see an argument for that being the case.  Whether or not that'd
be a good thing or a bad thing is another unanswered question.

Anthony