Saturday, January 24, 2009

FlaggedRevs Review

I've seen a lot of blog posts today about how English Wikipedia is preparing to get the FlaggedRevs extension installed. I think that's a good idea, English Wikipedia is a project that could really use it to great effect to improve it's reliability and decrease it's spam volume.

At English Wikibooks we've had that tool for a while now, and some of us are starting to question whether it's suiting our needs appropriatey. There have been plenty of complaints about our autopromotion requirements for +Editor (I can't even tell you what the current requirements are, since they are so complex). I've advocated drastic reduction of the requirements (after 10 edits or so), although some people have said that we should just tack +Editor on at the same time members get autoconfirmed, which is a simple 4 day timer. I like that idea too, although I'm not sure it is going to perform the promotion quickly enough.

We've also been worrying that we don't quite have a large enough community to support this tool anyway. The list of pages that have been reviewed is certainly growing, but it's woefully small compared to the total number of pages we have at Wikibooks. We just aren't reviewing pages quickly enough, and I'm not even sure that we're reviewing pages as fast as we are creating them. Of course, if we have more +Editors who are automatically sighting pages with every edit, that number will increase pretty dramatically.

Also, as if we need more factors to worry about, we need to think about whether or not our current grading metrics are sufficient. We have three metrics now: Composition, Accuracy, and Coverage. These are nice, but they don't necessarily cover all the things that we might want to look at. Also, I find there are lots of areas of overlap: Accuracy tends to go up as we get more coverage, and the quality of writing goes up with volume too. Being more accurate also requires more precise and higher-quality writing. In short, it's rare to get a bad grade on one if you've gotten good grades on the other two.

We have questions to ask, and questions to answer. Plenty of things will definitely change, and maybe we as a community might decide to uninstall the extension entirely. We'll see how things play out.


  1. First I should mention that I've been autopromoted to Editor (what is "+Editor"?) a while ago. Since then my personal way of using FlaggedRevs is:

    —I tend to ignore it. I don't even review the pages I'm writing.

    —I usually find myself unable to do the review. How could I evaluate the accuracy of any wikibook on a foreign language that I don't speak?

    My suggestion whould be to radically simplify the reviewing to a single yes or no question: "free of vandalism?"

  2. That is a good point Martin, it is hard for readers to evaluate a book whose subject they aren't previously familiar with. It should be possible to evaluate the quality of the page's composition, but if you aren't an expert in the subject you're not going to know how the page ranks in terms of accuracy or coverage.

    A single "Quality" ranking is probably sufficient. We can use any number of cleanup templates to mark a page as deficient in things like accuracy, coverage, text volume, formatting, etc. Also, we can use talk pages to discuss these things. If we had a single "Quality" metric, we could have the following levels: default/unreviewed, vandalism-free, and featured. If Editors marked a page as "vandalism-free" every time they made an edit, it would be quite a huge improvement over the current system. The only time somebody would need to make an explicit review would be when the book was promoted to featured status.

    Basically this system, while minimalist, gives us all the benefits of flaggedrevs without adding any additional and orthogonal workload.