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

Mon Dec 19 19:43:03 UTC 2005

David Gerard wrote:

> Jimmy Wales wrote:

>(re: [[John Seigenthaler Sr.]])

>> Any good editor _with enough time_ who looks at an
article 
>> about a living person which makes claims as
transparently 
>> outlandish as these will know to remove those
claims and 
>> insist on a source.  It would have taken 15 seconds
of 
>> googling to see that the claims were in no way
supported by 
>> any obvious source.

>Newpages patrol typically doesn't take or have the
time to do 
>that. It's a firehose of slush-pile quality
information. It's 
>about a first cull.

It shouldn't be. Jimbo's point about the "failure
mode" is absolutely right - at the moment we're
letting in "a firehose of slush-pile quality
information" which doesn't even meet basic standards
(I'll come back to this point). Our "failure mode"
right now is to let everything in and hope that
someone improves the "slush-pile". What happens if
they don't? We simply don't have enough time to make
this happen.

>> The Seigenthaler article didn't pass _any_ basic
quality standards.

>It passed *basic* ones, which is how it survived
Newpages 
>patrol. Basic quality standards for Newpages means
more or less 
>that's it's shaped enough like something that might
be a 
>Wikipedia article not to shoot on sight.

I don't know how you can argue that it met *any*
standards other than that it was written in English.
Let's consider what was wrong with it. It was an
orphan article, with no wikilinks, no references, no
formatting and hopelessly (actually maliciously)
incorrect information. What was right with it?
Ummm....

Note that in my original proposal I said that
including references should be part of the basic
criteria. I expressly didn't say that the article
should be accurate in every regard. 

There's an important point here - it's simply not
feasible to fact-check everything that goes into
Wikipedia and nobody should fool themselves that this
can be done. But the point about inaccurate
information is that much of the time, it's something
that someone has misremembered, it may be speculative
or, as in the Siegenthaler case, it may simply have
been made up. 

If the Siegenthaler article submitter had been
required to reference his allegations about his
subject, the article would probably never have been
accepted into the main namespace in the first place.
Insisting on referenced sources is the first and best
line of defence against bad information. 

>What it would need is something like what someone
else mooted, a
>biography patrol. Which is a damn fine idea, I think.

It might help with existing biographies, but it won't
solve the problem for newly created ones and it won't
do anything at all for all the new non-biographical
articles (which is most of them). Your idea of a
"prefilled" article template is a good one, and I
think it should be implemented. But it won't make much
difference if people are still free to create junk
articles and we still rely on those articles being
caught.

You said earlier in your post that "Newpages patrol
typically doesn't take or have the time to [check
articles]." This is true - we have too few patrollers
and too little time. Our method of checking new pages
is effectively equivalent to random sampling; we're
dipping into the firehose to see if what's coming
through it is any good. This might work for an
automated production line, but it's fundamentally the
wrong model for a setup where each and every one of
the products is "hand-made". You have to do quality
control on each article, not on a representative
sample.

I think we have three variables here: the number of
new articles, the number of patrollers and the time
required to review each new article. We might be able
to reduce the number of new articles a bit by
requiring users to log in before creating them, but it
doesn't seem to have made a great deal of difference
so far. The number of patrollers is something we don't
have any control over and it's never likely to keep
pace with the demand anyway. 

The only variable we *can* really change substantially
is the time we have in which to review articles.
That's where the change in "failure mode" comes in.
Right now, we have literally no time at all to review
articles, because as soon as they're published they're
in the namespace, and they scroll off the Newpages
list within minutes or hours. Putting them in a review
queue would give us all the time we needed to review
them. 

It would also greatly reduce the attractiveness of
submitting Siegenthaler-style hoaxes and "look
Ma!"-type junk because such articles would have a very
high chance of never getting into the public website
in the first place. Right now we're getting a huge
amount of similar junk because the submitters know
that it'll be visible immediately. Take that away, and
you take away a lot of the rationale for
hoaxing/vandalising in the first place.

One other point. People are understandably getting
jittery about biographies because that's a point of
legal vulnerability. But that's missing the point: we
just don't have a problem with bad biographies, we
have a problem with a broken process that lets in all
kinds of bad information, not just biographies.
Jimbo's experiment in stopping anonymous editors from
creating new articles won't stop that, any more than
your proposal of patrolling biographies. The
underlying process needs to be fixed, not just a
particular category of user privileges or content.

- ChrisO

___________________________________________________________ 
NEW Yahoo! Cars - sell your car and browse thousands of new and used cars online! http://uk.cars.yahoo.com/