discussions on other projects
Have there been any discussion about this software on the other big Wikipedia like Italian, Chinese, Portuguese & Dutch? John Vandenberg 01:09, 1 October 2010 (UTC)
Hi, at he.wikisource we would like to change the settings of Flagged Revisions so that only for pages rated "excellent" in all categories does the user see the last approved version as default. Can anyone give us some advice for how to do this? (E.g. request at Meta, file a bug report requesting exactly what...)
Comparison of some Wikipedias
- (copied from ) ruWP has Pending Changes (PC, or Flagged Revisions or Patrolled Revisions or Gesichtete Versionen or ...- i'll just call it PC) since 2009 (kind of, cp. Flagged_Revisions#Flagged_Revisions_on_Wikimedia_projects) and i don't think ruWP is a good example to follow. Much better example is polishWP and they are comparable to ruWP in both number of articles and number of active editors (see pl:Special:Statistics vs. ru:Special:Statistics). PC on plWP works quite good, like on deWP. PC on ruWP is terrible, on arWP too:
- ru:Special:ValidationStatistics: average review delay: 217 days; average review delay for IPs: 12days 15h; median: 13days 4h (1462 user with PC right)
- pl:Special:ValidationStatistics: average review delay: 10 days 10h; average review delay for IPs: 16h 1min; median: 1h 35min (2625 user with PC right)
- de:Special:ValidationStatistics: average review delay: 6 days 22h; average review delay for IPs: 11h 7min; median: 1h 32min (13.358 user with PC right)
- ar:Special:ValidationStatistics: average review delay: 122 days; average review delay for IPs: 41h 39min?; median: 26h 48min? (536 user with PC right)
- en:Special:ValidationStatistics: PC enabled on 447 articles only.
- So, why the big differences?
- 1) Status assignment procedure. On plWP and deWP users get PC right automatically at some point (number of unreverted edits+XYZ, I'll leave aside the difference of active or passive reviewers for now), plus users can request PC right for themselves or somebody else. On ruWP and arWP users have to ask for PC right, discuss and be eventually denied. Not surprising: There are far fewer users with PC right, and bigger delays. Reason for different procedure is a bigger (perceived?) vandalism problem on ruWP and arWP (well, egg/chicken first ... see RuWiki_History/English) Quite a power trip, denying PC status ^^.
- 2) Review policy. On deWP users with PC right only have to look out for vandalism. Not quality control, not "endorsing" edits, not reverting unsourced edits (except unsourced sensitive info on BLP). Users apply higher personal standards for their review anyway, but the rule is simple, low standard: just revert vandalism. Higher standards, more rules for PC -> less adoption, less users reviewing PC, higher delays. I read that ruWP applies higher standards for PC than deWP.
- This is IMHO, of course. See also relevant analysis/ discussion on deWP: ,  (mostly german, some english, you'll understand with googletranslate). Hope this helps. --Atlasowa (talk)
- Indeed, FlaggedRevisions is severely broken on several Wikipedias; comparing more wikis will give you even more interesting results, in my experience. What's worrying is that nobody in the community seems to care and, also, a lot of effort (and money) is being spent on some of those projects (e.g. WMF on ar.wiki), without investigating whether FR is killing community health making all the rest pointless.
- The 2010 analysis by Erik Zachte, simplifying, proved that there was no noticeable reduction of vandalism on de.wiki following FR implementation. However, the data is very old; hopefully we'll get Bug 46212 – Dump stats: update edit/revert stats monthly fixed soon and then the community will have more tools to understand what's going on. --Nemo 07:35, 19 March 2013 (UTC)
- fi:Special:ValidationStatistics: average review delay: 7 h 45 min; average review delay for IPs: 1 h 43 min; median: 18 min 35 s (937 user with PC (editor) right + 262 with autoreview -right and 58% of articles are reviewed).
- Status assignment procedure: right is mostly given to editor which edits are ok by administrator without any request or prior notification. Users can also request PC right for themselves or somebody else. Biggest problems with this are users which edits are generally good or very good, but they have also problematic edits too like copyright violations. Also the distribution of the editor/autoreview rights doesn't seem to run by itself and and somebody will need to specifically do it or too few users will get the right. (this is bad thing). Good thing is that with the non-automatic procedure admins can thank users for the good work and it seems that users like that. :) Manual procedure is also pretty accurate. 2) Review policy. Target is same as in dewiki; just revert vandalism. However as in dewiki users apply higher personal standards for their review anyway and there is statistical increase in reverts. I think that unsourced edits are reverted easier than before. Also in fiwiki there has been target that review backlog should be small so it will not start accumulate. This is main reason why review delay is so low in fiwiki compared to dewiki or plwiki. Zache (talk) 20:09, 19 March 2013 (UTC) Corrected the number of autoreviewers. --Pxos (talk) 20:29, 19 March 2013 (UTC)
- Thank you Zache, very interesting. I see that fiWP enabled PC in November 2011, relatively recent. How do you know about the statistical increase in reverts? (Link?) Regarding thanking new users/autoreviewers/editors when notifying of new rights: That is excellent, this is not done systematically on deWP, afaics. Regarding review backlogs, deWP sometimes does "review calls" on Kurier (Signpost), you can see the effects in the review stats. This works rather well, because the backlog does not grow beyond 10.000 articles (7-8.000 mean) and because every "editor" (user with active PC right, not just autoreviewer) has a backlog on his watchlist. And there are a couple of tools, like:
- Deep out of sight unreviewed articles in chosen category, by Magnus Manske
- Random out of sight 20 random unreviewed small edits, by Magnus Manske
- Hannes’ Tool similar to Magnus' tool, more features
- List of users by number of unreviewed edits by Hannes Roest
- List of articles that have never been reviewed, by age by Hannes Roest.
- This is the good stuff. --Atlasowa (talk) 14:22, 20 March 2013 (UTC)
I think PC is a great tool. It's very, very effective against vandalism and spam. I read a study about enWP showing that "~10% of damaging edits are viewed by 100+ persons. Deeper analysis shows that many of the associated survival times are quite short, and these are often the result of damage to extremely popular articles. ... While practical writings have often focused on the time to revert of damaging edits, we argue that the quantity of persons who view it is the more relevant metric. Vandalism that survives for days on an obscure article is effectively harmless if no one visits that article." This doesn't happen with PC, readers don't see unreviewed/vandal edits. Only logged users and the vandal (cookie) see the vandal edits, until they are reverted minutes later (or hours, days). This also means that the "recent changes" patrollers have less pressure and a less nervous "revert by huggle finger" - other editors will see the edit in their watchlist and will often be more competent on the article subject and on the merits of the edit. This is also very good for quality, articles get better, not worse. It discourages bad editors and POV pushers in practice. But it also frustrates good new editors, to some point, unfortunately. And yes, with long backlogs this is an even bigger problem on ruWP and arWP:
- A german user about ruWP: "I sometimes make small edits in the Russian Wikipedia and PC there bugs me mightily. The PC backlog is long, long, and the requirements for PC are far higher (not just "no vandalism"). Therefore, it is almost a rule that pages that I am working on stay unreviewed for months if I did not explicitly ask for screening."
- Another german user about ruWP:"For instance, although I had for a long time already collected more than 100 edits, I had to apply myself to get passive PC right there."
- A commentator about arWP in october 2011: "the main problem i have with Arabic Wikipedia is their stupid adoption of flagged revisions since 2009. They don’t have enough active members and editors to make this work. A lot of articles and great contributions are blocked because no one has the time to review them. I made a few edits back in 2010 and some more in 2011 and 90% are still pending and waiting for one of the few editors to take a look at them. I hope this issue will be discussed on some level at this meeting because on Arabic Wikipedia it is completely ignored by the admins. Apparently it pleases them to hold this kind of power and they refuse any sort of dialogue about it. Add to this, the fact that most editors/admins are from Egypt/KSA meaning that you have almost no chance to have your article approved if it concerns other Arab countries like Sudan, Tunisia or Morocco. Why? Because they simply don’t care about some historical event or some town in Sudan or Algeria."
- arWP had a poll about flagged revisions in october 2011 with 77 users, and they decided in November 2011 to cancel the automatic grant of "editor" right.
Frustrating good new editors is indeed a problem, not just on ruWP and arWP, also on deWP. In this respect I think there is still a lot of room for improvement:
- a) If a new user made at least 150 good edits in 30 days (for passive PC right: autoreview) and 200-300 good edits in 60 days and fulfills all conditions (many conditions! on deWP) he gets PC right automatically (editor). Do you think this user gets congratulations, an explanation of his new status, a welcome? No, he doesn't even get a notice about it. Nothing, unless some editor does it. Really. A dozen a day in the log, not even an automatic message.
- b) Reviewers get a link on top of their watchlist, to a feed of "unreviewed edits of pages on your watchlist". You practically get a feed of newbies and IPs editing in your area of interest! Yet we're not using this systematically, there are only individual editors that make an effort to reach out to newbies through this.
- c) Newbies systematically get negative feedback through PC, not positive feedback: A new user makes 10 good edits. 9 good edits get reviewed and pass without edit comment by the reviewer (because he can't comment when accepting the edit, technically). 1 good edit lacks a citation and is rejected by the user with PC right with an edit summary explaining the incompetence of the new user. So this good new editor gets 9 x no feedback and 1 x negative feedback (assuming he sees the reviewers edit summary or just sees his edit disappeared). Frustrating. And this good newbie needs at least 150 good edits in 30 days to have passive PC right (autoreview), a long way to go... and did I mention that he won't even know about this big step, never mind get the inherent positive feedback? (->b)
- d) What terrible things would happen if we relaxed conditions for PC rights, from 150 good edits to 80 (autoreview), or from 300 good edits to 150 (editor)? What good would it do? We don't know. Active editors immediately forget how things worked before they had PC rights. When PC started in 2009, deWP had 6500 users with PC rights, it's 13000 users now in 2013. Shouldn't this be more? Maybe this is OK on deWP, but on other WP, arWP? How to find out?
Again, I think PC is great, deWP would be worse without it. But it could be made even better. Regarding c), I'm actually working on a proposal for an EditLove button (or WikiLob, Edit-like, WikiThanks, whatever ;-) to give quick positive feedback to IPs and new users from the article diff.--Atlasowa (talk) 15:44, 20 March 2013 (UTC)
- Thanks for your thoughts. I had read that article on en.wiki pageviews, but it didn't calculate how many pageviews each vandalism likely got, it only esteemed aggregate results: punctual (?) stats would be very hard because we only have hourly stats and we can't know exactly what cached version a user was served (cache lasts 30 days for anonymous users, parsing time varies etc.). As for reverts, I think our amusing Finnish colleagues have their own statistical tools, if I remember correctly? But yes, we are all craving for better statistics. If you found some German user/dev interested in knowing more, maybe he could submit patches for WikiStats, or try and run the scripts on the dump alone. --Nemo 18:19, 20 March 2013 (UTC)
- And regarding revert stats, Erik Zachte of Wikimedia is asking Which single Wikimedia metric would inspire you most? Revert stats may be not exactly inspiring, but very useful, seeing how much new users dislike unexplained reverts (arab readers above, and all WP readers). --Atlasowa (talk) 16:08, 16 April 2013 (UTC)
Erik Zachte has produced new edit and revert stats, but I don't really see a correlation of revert ratio with the introduction of pending changes:
- Edit and Revert Trends: German
- Edit and Revert Trends: Russian
- Edit and Revert Trends: Polish
- Edit and Revert Trends: Arabic
- Edit and Revert Trends: Finnish