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).