I haven't finished the article, but I got to the section about verification and the following system poped out to me as an elegant solution.
So, you have 3 user levels: Unverified, Verified, and Credentialed.
All users start out as unverified. As unverified all edits made by these users are reviewed (by a verified or credentialed user). However, if a user wants to make edits in real time without oversight review they can immediately choose to go through the potentially lengthy process of getting verified. This would be the postcard approach as described in the article or whatever other system you like.
Finally, once Verified (or during the process of becoming verified) a user can choose to become Credentialed by going through whatever process is deemed sufficient to establish credentials.
Within Each level there are also sub-rankings. So an unverified user who has submitted 100 reviewed edits can be more trusted than a verified user who has only submitted 1. Thus there could even be a de facto verification process for unverified users after a certain point.
So for a fresh unverified user it would go as follows. All of the first 10 edits are reviewed. Then only every 3rd-5th edit (chosen randomly so vandals can't game it) from 10-50 edits is reviewed. From 50 say 1000, edits are only put up for review sporadically (every 100 or so). After 1000 non-contested edits an unverified user becomes a verified user. Also, verified/credentialed users in good standing can contest past edits made by a user, even those that were not reviewed. If these "black marks" add up then the user is banned. It's like Ebay, once you earn your ranking it has value so you don't throw it away.
Even Credentialed users can be peer reviewed by other credentialed and (with lesser effect) verified users, with like a white/grey/black mark system. That way some crackpot academic who signs off on every article under the sun, perhaps without ever reading it, can be singled out as untrustworthy, while a credentialed user who reviews a lot of articles and provides much feedback/corrections to the writers gets white marks for being diligent.
Finally, all users, no matter what, should be monitored for suspicious behavior. Making 100 edits in 10 minutes when your account has averaged only 1 a day for the past few months is suspicious, or making a few major edits of 1000+ words when your account averages only 10-50 word edits is suspicious. Any suspicious edits are held for review, even if it is a highly regarded credentialed user (after all their account could have been stolen).
Slashdot Mirror
I haven't finished the article, but I got to the section about verification and the following system poped out to me as an elegant solution.
So, you have 3 user levels: Unverified, Verified, and Credentialed.
All users start out as unverified. As unverified all edits made by these users are reviewed (by a verified or credentialed user). However, if a user wants to make edits in real time without oversight review they can immediately choose to go through the potentially lengthy process of getting verified. This would be the postcard approach as described in the article or whatever other system you like.
Finally, once Verified (or during the process of becoming verified) a user can choose to become Credentialed by going through whatever process is deemed sufficient to establish credentials.
Within Each level there are also sub-rankings. So an unverified user who has submitted 100 reviewed edits can be more trusted than a verified user who has only submitted 1. Thus there could even be a de facto verification process for unverified users after a certain point.
So for a fresh unverified user it would go as follows. All of the first 10 edits are reviewed. Then only every 3rd-5th edit (chosen randomly so vandals can't game it) from 10-50 edits is reviewed. From 50 say 1000, edits are only put up for review sporadically (every 100 or so). After 1000 non-contested edits an unverified user becomes a verified user. Also, verified/credentialed users in good standing can contest past edits made by a user, even those that were not reviewed. If these "black marks" add up then the user is banned. It's like Ebay, once you earn your ranking it has value so you don't throw it away.
Even Credentialed users can be peer reviewed by other credentialed and (with lesser effect) verified users, with like a white/grey/black mark system. That way some crackpot academic who signs off on every article under the sun, perhaps without ever reading it, can be singled out as untrustworthy, while a credentialed user who reviews a lot of articles and provides much feedback/corrections to the writers gets white marks for being diligent.
Finally, all users, no matter what, should be monitored for suspicious behavior. Making 100 edits in 10 minutes when your account has averaged only 1 a day for the past few months is suspicious, or making a few major edits of 1000+ words when your account averages only 10-50 word edits is suspicious. Any suspicious edits are held for review, even if it is a highly regarded credentialed user (after all their account could have been stolen).
Anyway... just my 2 cents.