On Sat, Aug 28, 2004 at 04:16:43PM +0200, Marco Krohn wrote:
On Saturday 28 August 2004 06:06, Jens Ropers wrote:
If you doubt the standards of our editorial
review mechanisms, go try
and introduce some decidedly un-encyclopaedic (unproven, contentious
and/or unacademic, etc.) information into an article of your choice.
Then check back and see how long your contribution will remain in the
article. My confidence is high that -- depending on how much this
contribution falls short of encyclopedic standards -- you will find
your contribution challenged on the respective article's discussion
page (where you will likely be asked to provide references for your
claims) or outright removed.
we just had someone on the de mailinglist who purposely modified four articles
and introduced a mistake in each of them. He also told us which articles he
modified and claimed that none of the mistakes was detected by now. I checked
three of the articles he was right with his claim.
In the german article about "consumer surplus" the error was there for about 9
(!) days before I removed the nonesense. In other articles the errors were
there for more than 9 days.
I agree with most of what you wrote, but I think it is a mistake to believe
that we have any kind of review system which is on par (wrt error
elimination) with a real peer review. At least my experience is that the
probability for finding a mistake in Wikipedia is by far higher than for
Britannica.
I'm seeing a lot of mistakes in Wikipedia too. It seems to be have a lot more
mistakes than it used to.
Imagine a wiki without a Recent Changes page - the error can stay in it for
days or even weeks before it gets corrected. Such a wiki wouldn't work well.
But that's almost the situation in Wikipedia today - because the Recent Changes
is so huge, very few people will check all articles, and as the topics of articles
become more specialized, chances that non-trivial error will be spotted in RC
are getting lower and lower.
Articles monitoring helps, but not that much.
I think we just have to divide RC into reasonably-sized parts.
We should group categories into some related standarized sets (without standarization
some of the categories would be much more likely to end underchecked), provide RC
for each of those sets, and a huge RC for articles without categories,
from which the edits in the articles would be categorized for later review
by people competent in given area. RC in all categories could stay,
but it would probably be only used for things like obvious vandalism
and newbie experiments (not like it's much different today).