The latest PC Pro has a 5 page spread on Wikipedia, the analysis is very fair. They did WPs ability to tackle deliberate or unintentional misinformation by posting ten small rewordings from an IP address (a single address!), all were corrected within an hour (suprised NOT). Later that week they tried the same thing from different addresses - 10 were correct within an hour and two went unnoticed for a week and they say they were self corrected. The article isn't online yet and I only managed to read it in the bookstore.
It's mentioned in peer review http://en.wikipedia.org/wiki/Wikipedia:External_peer_review
In http://en.wikipedia.org/wiki/Reliability_of_Wikipedia it covers it as such
A further informal assessment by the popular IT magazine "PC Pro", for its 2007 article *Wikipedia Uncovered*[8]http://en.wikipedia.org/wiki/Reliability_of_Wikipedia#_note-uncoveredtested Wikipedia by a similart device to those described above, by introducing 10 errors that "varied between bleeding obvious and deftly subtle" into articles (the researchers later put right the articles it had edited). Labelling the results "impressive" it noted that all but one was noted and fixed within the hour, and that "the Wikipedians tools and know-how were just too much for our team". A second series of another 10 tests, using "far more subtle errors" and additional techniques to conceal their nature, met similar results: "despite our stealth attempts the vast majority ... were discovered remarkably quickly... the ridiculously minor Jesse James error was corrected within a minute and a very slight change to Queen Ann's entry was put right within two minutes." Two of the latter series were not detected. The article concluded that "Wikipedia corrects the vast majority of errors within minutes, but if they're not spotted within the first day the chances ...dwindle as you're then relying on someone to spot the errors while reading the article rather than reviewing the edits."
I don't condone vandalism of this kind, but I think it proves we have a finger on the pulse.
Mike
Interesting. I'll see if I can snag a copy, at some point.
In my own experience, I tend to agree that things are usually caught pretty quickly (minutes), after a few days or weeks, or essentially never. The "regular" RC patrol and anti-vandalism bots are excellent at catching sudden and large changes which are obviously problematic, but more subtle changes usually get caught by editors diligently checking their watchlists for articles they've worked on or particularly care about. In my opinion, anyway.
Problems can come up, when two vandals hit the same page, whether in quick or slow succession; if the first change is subtle, and the second is blatant, then more often than not a bot or RC patroller will revert only the later change, and will neglect to even check the first. We need to encourage people to check over the pages they revert, to see if they may have missed any lingering vandalism.
Speed is good. Our main advantages, in my opinion, are in numbers and tools. We should use them to our best advantage. Increase awareness of tools, even just of page history, user contribs, and the ability to report vandals, and even common laypeople can help us keep things clean.
I personally am interested in two main areas, here: improving our mechanisms to help experienced users detect vandals, and educating the public as to the available tools and options for dealing with vandals.
Bit of a ramble, I guess. I keep telling myself I'll provide some useful coding, once I have a bit more education in that area, so hopefully I'm still involved with WMF when that day comes.
-Luna
On 28/06/07, Luna lunasantin@gmail.com wrote:
Interesting. I'll see if I can snag a copy, at some point.
In my own experience, I tend to agree that things are usually caught pretty quickly (minutes), after a few days or weeks, or essentially never. The "regular" RC patrol and anti-vandalism bots are excellent at catching sudden and large changes which are obviously problematic, but more subtle changes usually get caught by editors diligently checking their watchlists for articles they've worked on or particularly care about. In my opinion, anyway.
Problems can come up, when two vandals hit the same page, whether in quick or slow succession; if the first change is subtle, and the second is blatant, then more often than not a bot or RC patroller will revert only the later change, and will neglect to even check the first. We need to encourage people to check over the pages they revert, to see if they may have missed any lingering vandalism.
Speed is good. Our main advantages, in my opinion, are in numbers and tools. We should use them to our best advantage. Increase awareness of tools, even just of page history, user contribs, and the ability to report vandals, and even common laypeople can help us keep things clean.
I personally am interested in two main areas, here: improving our mechanisms to help experienced users detect vandals, and educating the public as to the available tools and options for dealing with vandals.
Bit of a ramble, I guess. I keep telling myself I'll provide some useful coding, once I have a bit more education in that area, so hopefully I'm still involved with WMF when that day comes.
-Luna _______________________________________________ There are some great tools like VP where you can wipe out 10 quick "Adam Gardner dances like a girl" type edits, but most of those get spotted by ordinary ppl in RC before you have chance to hit a button. The subtle changes that dont flag so easily are hard to spot with any tool. most ppl are suspicious of IPs and new users but its the silent vandalism that is a problem, a few minor changes and a summary with "ooops typo" are the vandals greatest tool. a few clean edits and bang - 5 edits before x has changed aquited to convicted!
Mike