Wikipedia talk:Flagged revisions

Also see the discussion at Wikipedia talk:Flagged revisions/Sighted versions.

A list of blog posts on the issue. Sdedeo (tips) 22:37, 9 September 2007 (UTC)

List of Sub Pages

/Trial

/Trial/Votes

/Trial/Proposed trials

Automatic sighting after a certain period

I think it would be appreciable from a number of points of view that, when an edit hasn't been sighted for x time, for example 24 hours, then the software automatically sights it. Primarily because Wikipedia is free for editing, and there is no reason that a backlog of non-sighted edits delays a good edit for days. Secondly because vandalism is generally reverted within minutes, or hours on lesser watched articles, so it wouldn't particularly reduce the expected effects of sighted revs. Of course, it would reassure users worried that it could impair the openness of Wikipedia. We could put the delay in a message so that we can modify it based on consensus. Cenarium ^Talk 02:15, 27 October 2008 (UTC)

This could be a good idea, since it avoids edits appearing for either (a) 24 hours, or (b) until sighted. That said, I have a number of concerns with this implementation which I'll need to think about. – Thomas H. Larsen 22:45, 31 October 2008 (UTC)

Reasons to automatically sight edits after a reasonable amount of time:

• Avoid huge backlogs of edits to sight that would dramatically reduce the openness and free editing nature of Wikipedia, and in the long run that would mean a decrease of anonymous and new contributors.
• Vandalism is generally reverted within minutes (even seconds) on high profile pages, and within hours on most pages, so it won't significantly decrease the efficiency of sighted revisions.

We could use different delays before automatic sighting, for example 24 hours for IPs, 18 hours for users, 12 hours for autoconfirmed users and 5 minutes for 'established' users, and of course immediate for surveyors. 'Established' would be a group between autoconfirmed and surveyor with only this specificity that edits are automatically sighted faster, that is granted automatically after certain conditions, and can be assigned and removed by sysops. Cenarium ^Talk 01:18, 8 November 2008 (UTC)

• Automatic sighting seems very largely to negate the purpose of the whole exercise, and the time-frames suggested seem extremely short for the generality of articles or even of FAs. Some older FAs seem barely to be on anyones watchlists. Johnbod (talk) 22:10, 13 November 2008 (UTC)
  • You forget that anti-vandalism tools, like Huggle, can catch all articles and vandalism is generally reverted within minutes, and to the extreme, within hours on most pages. If some edits are delayed by days, weeks or months because editors haven't some 'surveyor' rights, then it's the end of open-editing and will inevitably reduce the appeal of Wikipedia to new editors. Confirming that "any edit will be visible within 12-24 hours" will make flagged revisions acceptable. Without this, we'll never have support for flagged revisions, it's clear. Cenarium ^Talk 03:32, 14 November 2008 (UTC)

FWIW this is bug 4397[1], currently marked as WONTFIX. - (User) Wolfkeeper (Talk) 16:50, 14 December 2008 (UTC)

Due to lack of support. If we have consensus for flagged revisions with expiration on the English Wikipedia, but not for flagged revisions without it, then this will probably be revisited and worked out. Cenarium (Talk) 18:52, 18 December 2008 (UTC)

On full implementation: expired revisions

Main page: User:Cenarium/Flagged revisions/Expiration system

Still trying to address the problem of backlog for a full implementation, with acceptable edits getting unsighted for long periods. I proposed above that edits be automatically sighted after a certain period of time, however it should be properly distinguished from manual sighting, and doing so technically may be a bit cumbersome. With a similar goal, a system of expired revisions could be created to show to IPs revisions when they are old enough, and are not positive to an automatic vandalism check (see below). The principle is that when a revision is old enough, it is marked as expired and the latest expired revision is showed as the stable version if no later sighted revision exists. The rules are listed there. This system could be deactivated by admins on a page. Since there is still the, albeit unlikely, case where a vandalism edit could expire and be showed to IPs, we could use the abuse filter to prevent a revision matching certain filters from expiring. There seems to be an appreciable use of the abuse filter in conjunction with sighted revisions, for example preventing a surveyor from sighting certain revisions identified as 'very bad', since there is still the case where a vandal could access surveyor rights. Cenarium ^Talk 04:00, 29 November 2008 (UTC)

There is a special page to maintain the queue of outdated reviewed pages. I don't see why you'd need or want anything else.  — Mike.lifeguard | ^@en.wb 06:19, 29 November 2008 (UTC)

I suppose you are referring to Special:Oldreviewedpages. But all the problem is that they won't be sighted for long periods of time and thus good revisions will be long delayed to public view. This is with the lack of openness the number one opposition to flagged revisions. If revisions that are old enough and automatically verified as unlikely to contain vandalism are showed to IPs, this is a step in the good direction. Our editing volume is so important and the number of regular editors so small in comparison to the total number of editors that if we don't use this kind of system, we'll be quickly overburden. Using automatic processes like the Abuse filter, and semi-automatic tools like Huggle is, of course, necessary if we want that FLR works on Wikipedia. And even so, we won't be able to keep up with the number of edits, so an expiration system is inevitable if we don't want weeks-old or months-old unreviewed pages. Cenarium ^Talk 14:47, 29 November 2008 (UTC)

Just adding a note that this will probably affect anonymous readers only. They can still see the draft if they want - either clicking the link, or creating an account.  — Mike.lifeguard | ^@en.wb 06:22, 2 December 2008 (UTC)

There could plausibly be a bot that automatically sights revisions if they have been in the Special:Oldreviewedpages list too long.- (User) Wolfkeeper (Talk) 21:58, 14 December 2008 (UTC)

For the trial maybe, but not for a full implementation. That would be too much server load compared to a built-in, automated process. And it would mark an edit as sighted, while it is not exactly what we want since it has not been reviewed by a user. While marking a revision as expired, will recognize expired revisions as such, and will be dealt with differently. I'm thinking about creating a subpage where we can discuss the issue, generally and more technically. Cenarium (Talk) 18:58, 18 December 2008 (UTC)

RfC opened

I have opened an RfC on the /Trial implementation: Wikipedia talk:Flagged revisions/Trial#RfC on implementation. Happy‑melon 17:59, 14 December 2008 (UTC)

I wanted to open a RFC (or something like it) myself, but you beat me at this! Ruslik (talk) 18:20, 14 December 2008 (UTC)

Thanks. I'd like to announce this via watchlist notices as well, but I'm not sure if this would be the best path of action—suggestions, anybody? – Thomas H. Larsen 02:23, 15 December 2008 (UTC)

I think this should be a gradual 'ramp up': we'll start with the RfC, then spam the boards with it. Then we can add it to {{cent}} and by the time we actually come to have a poll a Signpost/watchlist combo will be appropriate. At the moment making it too visible will do more harm than good: we need time to iron out any problems before they're jumped on by the rampaging hordes... :D Happy‑melon 13:30, 15 December 2008 (UTC)

That's a good point. – Thomas H. Larsen 03:50, 16 December 2008 (UTC)

Wikipedia:Protecting BLP articles feeler survey

FYI, based on a conversation on Jimmy Wales's talk page:

• Wikipedia:Protecting BLP articles feeler survey

Your feedback is appreciated. rootology (C)(T) 22:40, 19 December 2008 (UTC)

Question

I have tried to quickly find the answer to this question in a few different places, including this page and its archived pages (without, admittedly, reading all of the thousands of posts), but I haven't found it. Can someone please explain, in relatively simple terms, what this "flagging" thing is all about, as it relates your average everyday editor or reader of Wikipedia? And please try to do it without doing things like using "flagged" and "sighted" to define themselves (which is one of the reasons I need to ask the question in the first place.) And please, please try to do it without sending me to a Wikipedia in a language I don't understand. (I almost fell off my chair when I saw the tag at the top of this page, suggesting I do exactly that.) In other words, let's say this proposal or a trial of it are implemented. Let's also say I am a regular editor, no special privileges, but not a new editor, and I make an edit to an article. What happens to it? Do people still see it right away? Or not? Or what? And then change the hypothetical to, I am an IP editor, and answer the same questions. And then change it to, I have just created an account and am making my first edit, and so on. And any other categories that apply. And then, from the standpoint of the reader, I get the impression that different readers are going to see different things, but I do not know who is going to see what, and when. 6SJ7 (talk) 23:20, 19 December 2008 (UTC)

Here's the quick version: It allows us to pick a version of a page and mark that version as "sighted". This means that that version will be the version of the page people see when they go to the article (although logged in users may pick if they see the newest edit, or the "sighted" version). This means that if someone comes along and edits a page to add "LOLZ" all over it, that version won't be shown to the public. New edits are "sighted" if they are not vandalism, although the exact process of all this is still being worked out here. --Falcorian ^(talk) 02:21, 20 December 2008 (UTC)

Ok, so it is called the "sighted" version because it is the version that is designated to be seen (or "sighted", I guess) by any reader who is not logged in, or by any logged-in editor who chooses to read that version. Is that correct? Thank you for the explanation, Falcorian. 6SJ7 (talk) 05:14, 20 December 2008 (UTC)

That's the gist of it! We're still trying to hammer out pretty much all the other details though... --Falcorian ^(talk) 08:35, 20 December 2008 (UTC)

Then perhaps a section should be added to this page to include an example of how flagged revisions and sightings works, describing who gets to see what and when. The current version doesn't actually provide much explanation. Dl2000 (talk) 18:36, 28 December 2008 (UTC)

That's entirely dependent on the configuration that's chosen. Mr.Z-man 18:43, 28 December 2008 (UTC)

You can try FR yourself on en.labs. Ruslik (talk) 18:49, 28 December 2008 (UTC)

/Trial progress

I think we are now approaching the stage where we can consider moving forward with this proposal. We have numerous comments on the RfC, which are generally favourable. I have asked for input from our friendly local 'crats, who have declared happiness with their part in the process. I have spent a few hours today mucking around with en.labs to make its implementation as closely as possible mirror the proposal here; I think it's pretty good. Have a play here and tell me what you think. I'm inclined to wait now until after Christmas, but maybe we should be thinking about opening a poll in a few days time? It will need to be open for a couple of weeks, so well into the new year, to gain the necessary consensus. Thoughts? Happy‑melon 14:41, 24 December 2008 (UTC)

Hi Happy-melon, sort of discussion is running on Wikipedia:Protecting BLP articles feeler survey. cheers Mion (talk) 15:21, 24 December 2008 (UTC)

I am looking at that, and I generally like what I see. However, any presentation to the developers must demonstrate a very clear consensus for the exact configuration proposed, which does mean a specific straw poll "shall we implement config X?". Happy‑melon 15:53, 24 December 2008 (UTC)

I noticed two possible technical problems:

1) 'autopatrolother' right assigned to reviewers and rollbackers. It will enable autopatrolling of the edits that these usergroups make to non-reviewable pages. Currently only sysops have this ability. Should it be removed?

2) Bots usergroup. Should 'review', 'autoreview', 'unreviewedpages' and 'movestable' rights be assigned to this group too?

Ruslik (talk) 16:49, 24 December 2008 (UTC)

Ooh, good catches. Certainly autopatrolother shouldn't be granted any more widely than it is now. I don't think we should give review to bots, but autoreview and movestable will be utterly crucial, otherwise we'll have to go round sighting all of Sinebot/MiszaBot/etcBot's edits! Happy‑melon 17:33, 24 December 2008 (UTC)

I thinks bots and sysops still need to have 'autopatrolother' (they currently have 'autopatrol' right). So added it to these user groups. Ruslik (talk) 19:26, 24 December 2008 (UTC)

AFAIK, the main namespace is pretty much the only one that gets patrolled anyway, giving some trusted-ish users autopatrolother shouldn't really matter. Mr.Z-man 00:34, 25 December 2008 (UTC)

Editor view only

I have a question regarding an editor making a contribution and as I understand it that editor is the only one seeing the article complete with the attribution, all other visitors see the last sighted version. Editor A makes a contribution, now comes editor B and makes a contribution which is only live shown to editor B, and not to editor A ? or is editor A upgraded to see the version of editor B as well (actually as soon as you make a contribution to an article flagged revisions is shut off for you and other editors on the article ?)Mion (talk) 12:18, 30 December 2008 (UTC)

All logged-in users see the most recent version of a page, no matter who edited it and when/whether it was sighted. So if editor A makes a contribution, editor B will see the page with that contribution whether the edit was sighted or not. Let's say that editor A is not a 'reviewer' and so cannot sight edits. He edits the article foo; to editor A, editor B and editor C, the article now appears with that new change immediately visible. To the IP 012.123.234.123, however, the article appears as it was before A edited it. Now let editor B come along and make another edit to the article, which she then sights. Now to all viewers the page displays the content after A and B's edits. Happy‑melon 13:20, 30 December 2008 (UTC)

But if editor A, and B are IP users ? not uncommon with a 33% IP user base, does editor A see the edits of editor B, or is editor A given the impression that the contribution was allright and still watches version A, which is actually changed to the version from editor B ?(until the sighter comes by) Mion (talk)

I see your question. The answer is that editor B will view the page, and will see the original page (before editor A's edit). When editor B clicks 'edit this page' (it actually says "edit draft") the edit box appears containing the current code (after editor A's edit), plus a diff of editor A's changes, plus a notice explaining what's going on. When editor B has finished editing, the next editor will see the up-to-date wikicode plus the diff of both A and B's changes. By far the easiest way to explain is with an example: this page on en.labs is currently in the state you suggest after an editor "A" without sighting permissions has made an edit to it. Have a look at what happens if you try to edit as an IP. Hope this helps. Happy‑melon 13:53, 30 December 2008 (UTC)

You're not a Sighter, then you're a Vandal

As this question is explicit about flagged revs, i'll ask it here:

Further testing on DE shows some effects. The setup is as followed

1 Administrators

2 Sighter

3 Registered trusted users

4 Registered untrusted users

5 Unregistered untrusted users (IP)

I have 2000+ edits on DE, and so i'm automaticly promoted to Sighter, but I only contribute content and corrections to DE, no patrolling, dont run a Cluebot, so I requested to have the Sighter status removed (de:Wikipedia_Diskussion:Gesichtete_Versionen#Sichter_abstellen , back to the former state. What happends is that my edits from now on are not shown to the public until a Sighter comes by, my edits end up in the hidden queue, you can see that on [[2]]. Mion (talk) 08:01, 1 January 2009 (UTC) In effect, my status on the German wiki changed from trusted user with 2000+ edits to vandal with 2000+ edits, regardless of the type of article, is there an option implemented that I can keep my current status ? Mion (talk) 09:25, 1 January 2009 (UTC)

No, your status reverted to a "registered trusted user" as you describe it; are your contributions to de.wiki being arbitrarily reverted? Why would you want to have sighter status removed? More importantly, why would you then complain about not having the benefits of the 'sighter' group? :D Happy‑melon 11:27, 1 January 2009 (UTC)

I'm not a patroller on DE, so i want to go back to group 3 Registered trusted user, the status i had before, so they did keep me as an autoconfirmed user, but in effect, i'm downgraded to group 5, as a not trusted user, and no, i'm not reverted, the edit is still handled as an edit in the sandbox [[3]] , until a sighter comes by. Mion (talk) 11:33, 1 January 2009 (UTC)

You have to logout to see the effect. Mion (talk) 11:40, 1 January 2009 (UTC)

There is no difference between 2 Sighter and 3 Registered trusted users. When you wanted to have your editor rights removed, you became a registered untrusted users, because there is nothing else. And you are right, that's a nuisance both to you and the german wikipedia, so please ask to get your rights back. --P. Birken (talk) 13:11, 3 January 2009 (UTC)

So, in effect i'm a vandal with 2000+ edits ? Mion (talk) 15:59, 3 January 2009 (UTC)

No, you're an "untrusted user"; or rather, someone who has asked to be treated as an untrusted user. You'd be a vandal if people weren't sighting your edits, were reverting them, and were thinking of blocking you. Hyperbole only clouds the issue. Asking for rights to be removed because you weren't planning on using them is stupid (look at my block log to see how much I use that permission), doubly so when you then realise that you've actually tied your shoelaces together in the process. Go to wherever the de.wiki equivalent of WP:RFR is and ask for your flag back. Refusing to do so and then complaining that you don't have the permissions associated with it is, I have to conclude, rather silly. Or am I missing the point of your comments? Happy‑melon 16:52, 3 January 2009 (UTC)

Legal liability under a sighter system

A few people in the proposed trial are referring to libel, and the review/prevention of libelous contributions. I just wanted to know if anybody has seen the legal ramifications of one editor 'sighting' another editors contributions as ok, when in actual fact it is libelous, be discussed before. As I understand it, reviewers are not going to be much more than trusted users, so where is the expectation that they can accurately judge what is and isn't libel coming from? And what does the presence of a more formal 'moderating structure' have on the legal position as I understand it, that Wikipedia itself is immune from claims of libel, which instead are filed against the original contributor. If I sight a libelous version under the Wikipedia sanctioned system, do I become a co-accused in any claim? Or is there simply no legal difference between libel in the public versions available to all, and libel in the unsighted versions available to registered users only?

As a side note, has anybody commented on the inherent bias it could produce in Wikipedia review process if reviewers who are more aware of the US libel laws (which differs significantly from the next largest superst of contributions, the UK) end up sighting more articles than other reviewers? How well (if at all?) do the Germans know the US libel laws (I have not looked to see if they actually have the presence of libelous material in their sighting criteria, so if its irrelevent just scratch it) MickMacNee (talk) 16:19, 3 January 2009 (UTC)

See Wikipedia:Protecting_BLP_articles_feeler_survey#Would flagged revisions make Wikipedia legally liable?. Ruslik (talk) 17:14, 3 January 2009 (UTC)

I've e-mailed Mike Godwin and asked him to comment; don't hold your breath, but it would be nice to have his opinion. In the interim I think Newyorkbrad's opinion in the thread Ruslik linked is the most authoritative (his credentials are well known, and his integrity is beyond dispute). Happy‑melon 18:04, 3 January 2009 (UTC)

I agree its probably all speculative, but I think the fact that other forums have moderators, it probaly indicates that the Foundation is not a risk by having organised systems. What I would like clarification on (or an official foundation statement of 'no idea') is, 1. Is it legally relevant whether libel is not visible to the general public, if it can still be seen by millions of registered editors?, and 2. Is a reviewer in any way even likely to be held as liable for sighting a libelous edit? MickMacNee (talk) 20:11, 3 January 2009 (UTC)

I think the general consensus is that implementing FlaggedRevs can't be detrimental to the Foundation's legal position; it is at worse neutral and at best a good-faith attempt to improve their handling of libel etc. I guess there are some parallels to deleted revisions, which are still visible without logging to 1,500 unvetted, unidentified users; from the fact that the give-non-admins-view-deleted discussion was axed by Godwin I suspect there might be some mileage in making questionable material less visible being a Good Thing. As for the legal liability of individual editors, I expect (hope certainly) that we're still no more liable than we would be if we made an unrelated edit to the page (and thus were unquestionably 'there') but failed to remove the libel or whatever, that is, guilty (perhaps, in a tenuous fashion) of failing to react to it, but probably not complicit in its addition. But then I'm not a lawyer, so don't put too much on these ramblings. Happy‑melon 21:00, 3 January 2009 (UTC)

Mike Godwin has allowed me to post his response, which was as follows:

“ So far as I know, there has never been a case that directly addressed the issue of whether review of a wiki article, followed by its being labeled "sighted" or "flagged", creates liability for the labeller. I can imagine a circumstance in which someone might try to base a theory of liability on such review -- the plaintiff might argue that the review somehow endorsed or validated the allegedly libelous statements in the article -- but that circumstance hasn't come up yet.

Note also that many non-American jurisdictions tend to be more willing to find creative ways to impose theoretical liability on editors.

The counterbalancing factor is that most plaintiffs want to be able to reach the money at the Foundation, not individual editors. And "sighted" revisions won't help plaintiffs do that, in general.

--Mike

”

Typically conservative; the general message seems to be "no one has a clue". It seems to implicitly support the assumption that the system won't make the Foundation any more liable; and that, given the limited compensation to be gained by suing an individual editor rather than the Foundation itself, plaintiffs are extremely unlikely to try given the uncertainty over their legal position. I don't think legal liability is a major concern of his here. Happy‑melon 16:59, 4 January 2009 (UTC)

I've been deeply concerned about this issue as well. The Seigenthaler incident involves an article which could easily have been flagged by a user who is not completely alert. (I would estimate that it passed under the radar of at least a few new page patrollers who saw the article but didn't spot the offensive sentence.) Under an flagging scheme, I think the media would come pretty hard down on a volunteer who gets fooled into approving an edit like that, even if the flagger were not legally liable. I wonder if a possible solution is to make the flagging process totally anonymous; no record of any sort should be made on who flagged what edits. There should be no log which a small priveleged group can see as is the case with checkuser and oversight, I mean no log whatsoever. The only person who can identifiably be associated with an offensive edit is the person who made the edit, not the person who approved the edit. Sjakkalle (Check!) 07:21, 6 January 2009 (UTC)

This sort of anonymity is deeply against wiki philosophy; transparency and open-ness is an absolutely fundamental tenet of what we do. There is no such thing as "no log whatsoever"; if a legal case was brought, the wikimedia developers would be subpoena'd; they would tear the server logs apart and extract the evidence of who did what from the raw data if they had to. There is no way to conceal from a court the fact of who sighted what edit unless the entire system was reconfigured with that express purpose in mind... and I suspect a court would view such an effort with immediate suspicion, which is the last thing we want. So given that a court would have the data one way or another, what's the point of witholding it from everyone else?

This sort of issue is precisely why we don't require people to edit under their own name; if the thought of media ridicule concerns you, edit under a pseudonym that you can throw away, as you do now. Happy‑melon 11:01, 6 January 2009 (UTC)

OK, the foundation is not at risk by implementing FR, we have no grounds to say what the liability of sighters will be, so the only thing left I want to see cleared up is whether a libel is legally less actionable if it is only visible to registered logged in users. i.e. is there any actual legal benefit to implementing FR with regards protection from libel claims? MickMacNee (talk) 17:11, 6 January 2009 (UTC)

I've asked Mike again, hopefully he'll respond, although don't hold your breath. My distinctly non-legal Opinion is that there is probably some crossover with the view-deleted debacle. That viewing restriction was rather more pronounced - it was proposed to rescind the status quo which was to hide a lot of material with a high libel content from the vast majority of readers - but the situation is similar. Godwin said that rescinding this viewing restriction would be legally "disastrous". I can imagine that implementing FlaggedRevs would have a more moderate but still arguable positive effect in comparison. Happy‑melon 17:36, 6 January 2009 (UTC)

21 days?

The major argument for this proposal has always been "But it works on the German Wikipedia". In fact, this translated report says that the German WP is, with great effort, keeping the median time between sightings down to 21 days. This means that half of all sightings take longer than three weeks.

This is a tolerably good description of failure, on a smaller project far more committed to this idea. Can we declare this historic now? Septentrionalis PMAnderson 19:54, 3 January 2009 (UTC)

If you can persuade a sufficient number of the 96 people who patently don't share your opinion that this is sufficient new insight for them to completely reverse their position, then yes we can. Work within the system, please. Happy‑melon 19:59, 3 January 2009 (UTC)

How many of that 96 (and which 96 are we talking about?) have any idea that this "successful implementation" is a scandalous failure? Septentrionalis PMAnderson 21:37, 3 January 2009 (UTC)

It's 106 now. Since I assume they all read the relevant discussions surrounding this topic before declaring their support, they're all familiar with exactly what's going on at de.wiki. It's not exactly a secret. Happy‑melon 21:51, 3 January 2009 (UTC)

You do? What WP have you been editing? In most polls, the !votes are based on the question and sometimes the few voices expressed immediately before their own; it is the exceptional !vote which is based on the entire poll, much less any outside information, like that in another section of the same page. This naïveté may explain why Happy melon insists so vigorously on an idea which assumes a level of activity incompatible with experience and with human nature. Septentrionalis PMAnderson 22:06, 3 January 2009 (UTC)

I don't want a 3 week backlog. A reviewer bot that sights revisions (that don't contain anything on a special blacklist) after 48 hours would be a good idea. That way, a few pieces of vandalism would get through, but it would be a lot less than now. Dendodge Talk^Contribs 22:16, 3 January 2009 (UTC)

The moment you start talking about bots and flagged revisions, you've completely undermined the ostensible purpose of flagged revisions. –Outriggr § 07:57, 4 January 2009 (UTC)

This thread is an example of misunderstanding of the statistics from German Wikipedia. Majority of edtis on de.wiki are sighted without minutes or hours and never show up in the list of unsighted pages. 1 week median or 21 days maximum only represent the average "ages" of those few revisions that were not sighted quickly and ended up in the list. So backlogs in German Wikipedia are not so awful as they seem. Ruslik (talk) 09:06, 4 January 2009 (UTC)

Please read median. The report may be false, but it says that 50% of sightings take place a week or more after the last one; that is not "few". Septentrionalis PMAnderson 17:19, 4 January 2009 (UTC)

the median time articles edited by non-sighters wait is unfortunately currently not recorded, but if of course lower - most edits are checked be RC patrol, watchlists or wikiprojects/portals within minutes to hours. The time articles wait is not known. 1 week is median of the waiting time of the articles listed there (in the list of unreviewed articles). All this thread is based on the misinterpretation of the statistics. Ruslik (talk) 17:52, 4 January 2009 (UTC)

I have just asked for, and received, "Sichter" rights on de.wiki, which means I can now see the contents of de:Spezial:Seiten mit ungesichteten Versionen, the list of all articles with unreviewed edits. The very first thing I notice is that the top article, that is, the article that has been waiting the longest for review, is marked as being "Raketengrundgleichung (-232) (review) (20 days) [sighted] (15 users watching)" (my emphasis). I believe there has been a mistranslation. I will endeavour to determine the actual median as we understand the term. Happy‑melon 22:45, 5 January 2009 (UTC)

There are 11,983 articles on de.wiki currently with unreviewed versions. The 5,992^nd article is de:Tauhid-Moschee, the oldest unreviewed edit being at 19:35, 1 January 2009 (UTC); four days and three hours ago. Although not optimal, this seems to me a much more reasonable length of time. Happy‑melon 23:01, 5 January 2009 (UTC)

I don't think that is the median time before an unsighted edit is sighted. Your sample consists of articles that are currently unsighted. This sample is biased towards edits that take long to sight, so it overestimates the mean time before an unsighted edit is sighted. A better sample would be to take the last (say) 10000 unsighted edits. -- Jitse Niesen (talk) 00:27, 6 January 2009 (UTC)

Perhaps an example helps to explain my point. Suppose that 90% of the unsighted edits take a minute before they are sighted and that 10% take a month. Then the median time before an unsighted edit is sighted is a minute. However, at any given moment almost all the articles that appear on the list of articles with unreviewed versions come from the 10%, so the median time for the articles on the list is much longer. -- Jitse Niesen (talk) 00:35, 6 January 2009 (UTC)

That's correct, an excellent point. So the median time for an individual edit to be sighted is in fact lower, probably considerably lower. Happy‑melon 11:03, 6 January 2009 (UTC)

Thanks Jitse. This is exactly the point that I wanted to make with my comments above. The real median time is significantly lower. Ruslik (talk) 12:18, 6 January 2009 (UTC)

3 weeks is the maximum time on DE. The mean wait (which is actually useful) is 6 days per aka's tool. We are working on getting that down too. Aaron Schulz 14:25, 6 January 2009 (UTC)

Although presumably that mean calculation is also based on data from de:Spezial:Seiten mit ungesichteten Versionen, so the same issues would apply. Either way, the numbers are far lower than the 21 days intially believed. Happy‑melon 15:27, 6 January 2009 (UTC)

I'm not sure the numbers have a downtrend, the numbers on articles to Sight are up de:Wikipedia_Diskussion:Gesichtete_Versionen#Meinungsbild Mion (talk) 21:25, 8 January 2009 (UTC)

I just removed this statement from the project page. Per the above discussion, it is false, and moreover it seems to be directly influencing the voting in the current poll for the trial implementation. To summarise the above points:

1. The German report said that median time for waiting is "about a week", contrary to the statement by PM Anderson above that it is 21 days which was also stated on the project page. The de:wp is aiming (apparently successfully) to keep the maximum wait to less than 21 days.
2. As also pointed out in the German report, but not very easily understood, it is crucial to realise that the mean or median age for currently-unflagged edits is much longer than the mean or median time for an edit to be flagged. This confusion is a classic blunder by statistical novices. The reason is that a list of unflagged edits is not a fair sample of all edits: it is obviously and strongly biassed towards edits which wait for a long time before flagging. For instance, if most edits are flagged immediately (because made by editors or bots with flagging rights) then the median time for flagging is zero, but the median for unflagged pages is obviously finite. (I can write down the formulae to prove this if you like).

What we really would like to know is the median time an edit by someone without flagging rights goes before being flagged (discounting edits that are eventually reverted rather than flagged, because it doesn't matter much how long this takes). This information is there in the logs if anyone on de:wp feels like extracting it. PaddyLeahy (talk) 11:48, 11 January 2009 (UTC)

Up to date and long time statistics can be found under http://de.wikipedia.org/wiki/Benutzer:ParaDox/Tabelle/noch_zu_sichtende_Artikel. The median you mention is the green curve labeled T4 in the last graph. Otherwise, we are currently at max waiting time around 16 days, still falling. --P. Birken (talk) 21:02, 20 January 2009 (UTC)

Median-time till sighting (time between editing and sighting; graph T4) is several hours declining in tendency. --Septembermorgen (talk) 15:18, 21 January 2009 (UTC)

"It's 106 now. Since I assume they all read the relevant discussions surrounding this topic before declaring their support,"

That's a hell of an assumption. And what's 106 users to the userbase of Wikipedia, miniscule? MikeLieman (talk) 21:49, 25 January 2009 (UTC)

Isn't the German project trying to implement flagged revs on all of their articles? Of course that will take a long time, and that was a silly move on their part. If we only implement it on articles that would otherwise have been protected, the back log will be 1% of the German wiki, would be easily manageable. --Arctic Gnome (talk • contribs) 22:00, 25 January 2009 (UTC)

Speaking my piece: The "This is bull" report

What the hell is the world coming to?
::-Jackie Gleason in Smokey and the Bandit

...or should I say, "What the hell is Wikipedia coming to?". Being a big supporter of freedom and Wikipedia, I think this quote proves what is going wrong here. We are implementing the big brother process on Wikipedia, which is supposed to be free for anyone to write! We are going farther and farther away from purpose here!! If you are reading this, look at your upper left corner on your screen, under the globe. It says "Wikipedia, The Free Encyclopedia". Do we really think we are that anymore? I cannot see how people still think we are, especially if we are asking for something so communist to enter the society and project we work on as a global community! This is why I have detailed my piece here as "bull", because that is my belief of Flagged Revisions. Here are many reasons and explanations to why we should not go with the track we are going.

1. Flagged revisions are not the right thing for the eminent backlogs that we already sustain. Over the past few weeks, I have been watching on Wikipedia and off-Wikipedia people talking. Several have complained that important processes, such as the Suspected sockpuppets area, or Good article nominations. I have brought these up, because they are critical areas in maintaining this project's biggest purpose: writing articles and being the encyclopedia we are supposed to be. Doing this, will only cause massive backlogs everywhere. Do we need these backlogs? One word: No.

2. Flagged revisions just adds more bureaucracy and monarchy, along with more anarchy, to Wikipedia. Nobody wants any one of these three words alone. We are a democracy. We elected and do the best work with freedom of expression, interest, media, opinion, and decision. We don't want to be a dictatorship or communistic society that limits what they can actually do to improve Wikipedia. What the in world are we thinking??? Did communism work in the Soviet Union? Do you see Stalin or Lenin and them running Europe? No, because the democracy that is America helped stop it. We do not want to go through this, do we? It is a ridiculous thought.

3. How is Jimbo Wales god? He isn't. Not even on Wikipedia. Wikipedia has no god, Wikipedia has no dictator, Wikipedia has no Queen Elizabeth. Jimbo is nothing but a say in what goes on in this community. You can ban me for this if you'd like, but, Jimbo Wales is not our decision maker. He will never be. We, as people, are supposed to make the decisions of what is to happen on our site! Jimbo may be the head of a great foundation, that has beliefs that I very much agree with. I very much disagree by what some people have shown in regards to this debate and prior debates. I have seen some people support things, a lot of which Jimbo himself has voted support or oppose on, under the theory that If Jimbo supports/opposes it, I support/oppose it as well. Is this a way of mind, or just people giving in to a person, who on Wikipedia, has little power? As far as I am concerned, if we are to be listening to anybody's important decisions, it should be the Arbitration Committee's. They often make the final decision on what happens in disputes. We should not be making decisions only because of Jimbo. He is not god, he is not the president, he is not a supreme court judge. He is a public orator, and spiritual leader, not our boss.

To conclude, there are three major problems. They are: 1) Backlogs, 2) Communistic/Big brother way of thinking, and 3) We look up to Jimbo for way too much. We need to think what's right for us. We are not communisitic. We are not Stalin, Lennon, Hitler, or the Egyptian emperors. Jimbo is not god, Jimbo is not President Bush and/or Prime Minister Brown, Jimbo is not emperor. Jimbo, is basically a spiritual figure, such as the Dalai Lama, Buddha or Confucious. We do not follow "big brother" beliefs. We have the 1st amendment rights we are given and we should use them correctly. In retrospect, we are running as an American democracy, and that is HOW IT SHOULD BE. I will end on a quote.

• ...There is pride in every American heart and its time we stand and say....

• ...That I am proud to be an American, where at least I know I am free...

• ...And I won't forget the men who died, who gave that right to me...

• ...And I gladly stand up...

• ...Next to you and defend her still today...

• ...Cause there ain't no doubt I love this land...

• ...God bless the U.S.A....

• -Lee Greenwood in God Bless The USA (1984)

(Note: If you want to bring up any issue about this report, or want to call me out, bring it to my talk page.Mitch32^{(Go Syracuse)} 22:09, 3 January 2009 (UTC))

Well, since this is a discussion page rather than a place for personal monologues, I'll reply here instead.

We won't know if there's going to be massive backlogs until we try it. Evidence from Wikinews and the German Wikipedia suggests there won't be massive backlogs.

You say this is going to add bureaucracy, monarchy, communism, and anarchy to Wikipedia? I'm pretty sure at least some of those are pretty much impossible to add together. Also, Wikipedia is is not a democracy, its closer to anarchy really.

What does Jimbo have to do with anything? Besides the fact that he supports this, he has basically nothing to do with this.

The First Amendment does not apply on Wikipedia.

Mr.Z-man 22:49, 3 January 2009 (UTC)

Mitch, while I don't really agree with your arguments, I think your concern is well-justified. Flagged revisions are potentially dangerous for Wikipedia. They could move us closer to Citizendium, which is a bad thing. Citizendium is going to bomb - it is probably bombing already. Sanger does not believe in the brute force of 150 thousand active editors and 200 thousand edits per day (which is Wikipedia at the moment) - he appears to think that quality and volume (an encyclopedia has to have both) will come from "experts". Of course, he is mistaken. Mark those two numbers: if they start do go down, Wikipedia goes down with them, flagged revisions or not. Freedom and equality (or meritocracy, if you will) are not just values in themselves, they made Wikipedia a success; lack of freedom and equality will make Citizendium a failure. We are playing with fire here. Still, I voted "Reluctant support" just out of curiosity - I want to see it in action, only then I can draw some solid conclusions. It's early to get worked up over this. GregorB (talk) 01:27, 6 January 2009 (UTC)

Mitch, Jimbo Wales is like Queen Elizabeth II. The only difference is that Jimbo can make big decisions about how Wikipedia is run, whereas the Queen has no say on how England, Great Britain or the Commonwealth is run. --Joshua Issac (talk) 14:56, 11 January 2009 (UTC)

We'd be a hell of a lot better off with Our Liz, in my opinion. DuncanHill (talk) 00:21, 17 January 2009 (UTC)

• Actually, recruiting Queen Elizabeth or King Juan Carlos will be an enormous asset. Not some Hollywood plastic smile, but a real royalty with thousand-year-long-pedigree. Won't hurt fundraising either. NVO (talk) 19:52, 17 January 2009 (UTC)

what about POV and misleading refs and other content problems

I think one of my biggest problems with this sighting/flagging thing is that it only deals with obvious vandalism. The impression you get is that once someone with sighting privileges looks at and approves an article, that it is somehow more trustworthy. But that is very misleading. These sighters only do what the totally uninformed reader is already very well able to do: that is, they don't think that a page replaced by the word penis is real content. But the subtler content issues, the sneaky POV stuff, the refs that don't actually support an assertion, those things that would fool an uninformed reader into believing something that may not be true, they also fool the sighters. Sighters are basically uninformed about the subject of the article, and with the mass of edits they have to approve, there is no way to ask them to be informed. If what we are looking for is a way to have casual readers be more able to trust wikipedia, then obvious vandalism is not the problem and flagged revisions won't do a thing. There are a ton of anti-vandalism tools out there, what we need are some better anti-POV-pusher tools. xschm (talk) 20:49, 6 January 2009 (UTC)

Did I miss the memo about WP changing from a full volunteer effort? ♫ Melodia Chaconne ♫ (talk) 21:53, 6 January 2009 (UTC)

It's certainly true that FlaggedRevisions is not the panacea to all our problems with destructive editing. That doesn't mean that it's not a good idea. We already have ten thousand tools to combat subtle errors such as those you note. They're called editors, and currently far too many of them spend their entire wiki-lives reverting trivial vandalism. Every tool we create is intended to deal with a little bit more of that casual vandalism, to give the humans the chance to do the really difficult work. FlaggedRevisions is, in one sense, just another anti-vandalism tool. It is likely to be that much better an anti-vandalism tool, however, because it works in completely the opposite way to most. Most of our anti-vandalism tools, from Twinkle to CheckUser to ClueBot, are designed as 'force multipliers': they amplify one particular user's ability to deal with casual vandalism so that one positive editor can 'ward off' many vandals. They are ultimately tools that involve only one person. FlaggedRevisions, on the other hand, is an attempt to spread the workload, to involve all our established editors in combatting casual vandalism. The overwhelming majority of editors don't get involved in 'dedicated' anti-vandalism patrol because that just doesn't interest them; but when they see an article on their watchlist that looks like it might have been vandalised, they go check it out and revert it if it has. FlaggedRevisions acts as a 'force multiplier' for all users, turning everyone into active anti-vandals. Every time they sight an edit, they are making a positive statement about the quality of the encyclopedia, not just 'restoring the status quo' by reverting vandalism. As such, yes, it is another tool against casual vandalism, and you are quite right that it will not solve our deeper issues. But by freeing up time and energy of humans to work on those issues, it contributes indirectly to progress in that area as well. Happy‑melon 22:08, 6 January 2009 (UTC)

The more things we ask flaggers to check for, the slower they will be able to work and the longer it will take to get revisions approved. This will result in larger backlogs and will be a bigger deterrent to new and unregistered editors. Hut 8.5 22:16, 6 January 2009 (UTC)

I guess what I am saying is that sighters are not making a positive statement about the quality of the encyclopedia. All they will be saying (all they can say) is that they didn't notice any vandalism. Any wikipedia editor who notices vandalism will revert it already. What we don't need is positive statements about the same content that still contains the same level of dubiousness that it always did (and probably always will). Flagging is just just going to create busy work. It is spreading the work that the force multipliers are now doing back out among people who will have to do it piece by piece by hand. Vandalism is the cost of openness and currently openness is winning the war. It is very rare to go to a page randomly and find it vandalized. I don't understand what the big concern is. xschm (talk) 22:49, 6 January 2009 (UTC)

I have added proposed Trial 11 to see if being a Reviewer takes on any significance other than being a vandalism/nonsense/libel(?) checker. I am concerned that removing someone's Reviewer status will become 'Block Light', and that it will become the culture that as long as pages in the 'public space' are not being protected due to edit warring/POV pushing, everything is OK, or that consensus among Reviewers (rather than 'untrusted' users, registered or unregistered) becomes more important in settling content disputes in the 'public space'. MickMacNee (talk) 23:57, 6 January 2009 (UTC)

Limiting the revisions to flag for large-scale implementations

A possibility for large-scale implementations would be to limit the number of edits to review, not only by limiting the number of pages with flagged revisions, but also by limiting the number of revisions inducing a stable/draft divergence for individual pages. The soft line is: none, the hard line is: all, there are mediums. For example, an automated process could check if an edit is likely to be vandalism or not, based on the latest flagged revision and successive edits. If an edit is identified as potential vandalism, it'll be 'moved' to the draft page, and the latest revision that has not been identified as potential vandalism is set as the stable version. This would require some sort of filtering of edit, function that can be provided by the abuse filter. This would require a close collaboration between the two extensions. Examples of rules for the filter could be "redirect to non-existent article", "blank", "redirect an article which is featured, or linked from main page, etc", "mass addition", "mass removal", and other rules for typical vandalism that already use anti-vandalism bots. But those could be more inclusive, since the edit is not reverted, but deferred to the analyze of a trusted user. This would require another user right, since it is also aimed to prevent vandalism by autoconfirmed users and edits to flag would be limited and in vast majority vandalism. This would prevent readers from seeing most vandalism, that is generally reverted within minutes, but would have no effect on the vast majority of edits. This is compatible with the semi-protection implementation, the only change will be that in order to flag an edit on a 'semi-flagged' page that has been identified as potential vandalism, it requires the additional user right. If we want, we could have the option to flag all edits of a specific page, either using a filter or through direct configuration of a page, but we can also disable it. We could also use filters for specific categories, for example Category:Living people. Cenarium (Talk) 14:24, 8 January 2009 (UTC)

I think it is slightly unhelpful to consider FlaggedRevisions as working on anything other than pure versions: talking about "moving edits" or "number of revisions inducing...". FlagedRevs works on versions, changing that would be a large and fundamental shift. That said, I think your general idea here is a very good suggestion: that we can adopt more than one position between "obvious vandalism" (ClueBot reverts or abuse filter disallows) and "obviously positive" (autoreviewed). Integration between the activities of the AbuseFilter and the actions of FlaggedRevisions can only be a positive development, because AbuseFilter is the exact opposite: it works on edits and nothing else. Let each do that which it is best at, and we'll all benefit. Happy‑melon 17:40, 8 January 2009 (UTC)

That was essentially expressions, the two combined would have this kind of effects. It would be positive to have more liberty for versioning, and using the abuse filter to check edits and then base versioning on it could be worthwhile. That would require much development and tests on other wikis. Is the abuse filter available on en.labs ? Cenarium (Talk) 19:58, 8 January 2009 (UTC)

Sorry, perhaps that was a little harsh; I do agree that the result of a FLR/AbFil combination would be very powerful. I don't think AbuseFilter is available on any wikimedia sites as yet, but Werdna has recently been contracted by the Foundation to polish it off, so I'm sure they'd be amenable to firing it up either there or on test.wiki. While we're at it, an interwiki to en.labs wouldn't go amiss :D Happy‑melon 22:10, 8 January 2009 (UTC)

I think it is a very interesting proposal! It should be seriously considered. Ruslik (talk) 22:22, 8 January 2009 (UTC)

See also Wikipedia_talk:Flagged_revisions/Trial#Example_where_Flagged_revisons_would_have_helped, why this proposal isn't working. Mion (talk) 10:43, 12 January 2009 (UTC)

You have not shown that this proposal isn't working. The abuse filter has two choices when confronted to an edit: warn user or disallow, deferring the edit to a trusted user is another, intermediary response. Cenarium (Talk) 18:42, 15 January 2009 (UTC)

Multi-level sighting options

Summary: Let me be an Editor; just don't force me to sight when I edit.

I tried the test wiki a couple of days ago, and I was surprised to not find a checkbox that I could check or uncheck to not sight an edit I made. There was a third checkbox on the edit page; but it didn't seem to do anything.

So I suggest a "Sight this edit" checkbox on all edit pages presented to those with Editor status.

Furthermore, to combat the "objectors" problem in Sebastian's experience at the German Wikipedia, I suggest two radio buttons on the Editing preferences tab. The controls would look like:

By default, sight:

• All edits
• Edits to sighted versions

The "All edits: option would make "Sight this edit" checked by default in all cases. The "Edits to sighted versions" option would instead make "Sight this edit" checked only when editing an article that was already sighted. If editing an unsighted article, the checkbox would be unchecked by default. This would prevent the "forced labor" problem, as an editor could elect to only sight his/her own changes by default.

-- Ken g6 (talk) 19:32, 9 January 2009 (UTC)

Worth noticing

[4]. Seems like it is going ahead. DuncanHill (talk) 00:33, 17 January 2009 (UTC)

Is this appropriate?

Is it really appropriate to have that quote from Jimbo Wales in an alert box at the top of this page? Especially the point about everything other than his opinion being fear, uncertainty and doubt? How is this not pov-pushing? Kolindigo (talk) 19:44, 17 January 2009 (UTC)

Not in the least appropriate. DuncanHill (talk) 19:49, 17 January 2009 (UTC)

Why not? Each of the three words is applicable; the negative connotation of their combination is used by both parties. Don't overestimate the weight of this yellow box. NVO (talk) 19:54, 17 January 2009 (UTC)

There is no counter to it from any of Wikipedia's other sole co-founders or godkings. DuncanHill (talk) 19:56, 17 January 2009 (UTC)

It is also untrue. Jimbo is talking about his personal preference for flagged revs, not necessarily the way it would be implemented. A lot of the backers of flagged revs seem to want it on the whole encyclopedia, not just as a replacement for (lower level of) protection. Most of the proposed trials don't match jimbo's description. xschm (talk) 18:26, 18 January 2009 (UTC)

Why flagged revisions is essential to fixing the problems of scale and reliability

Wikipedia is and will remain "the encyclopedia" that anyone can edit. However, this feature creates a layering, a layering that has for a long time been necessary but not achievable without this feature. Without flagged revisions, yes, unless an article is protected,which stops nearly all editing or severely inhibits it, there must be constant defense not only against vandalism, but also against the insertion of unreliable information or the removal of text accepted by consensus. Immediate editing by anyone is in conflict with reliability, unless we could guarantee that all edits were immediately reviewed. That happens with highly controversial articles, but not with much of article space. And with controversial articles, if they are in frequent flux, we don't know what version a casual reader will see.

So how can we have immediate editing by anyone and reliability? It's simple: immediate editing is confined to a layer below the top-level default layer. Anyone may propose an edit. In standard deliberative process, it was realized long ago that if a motion couldn't find a second, it was a waste of time to debate it. If we have a publisher which automatically, without review, publishes what someone writes, we'd consider that tantamount to "self-publishing." We have been that publisher, which is one reason why we are so attractive to linkspammers and POV-pushers and vandals.

With a bit more sophistication in the structure, we could actually become a truly reliable source, ourselves; but we, of course, still would not want to rely on that except in uncontroversial ways. As a reader, though, I expect an encyclopedia to be reliable. Wikipedia has guidelines and policies that require reliable sources, but the enforcement of this and the unevenness of our consensus process for applying and interpreting these has made reliability an unrealized ideal.

If there is substantial delay in sighting revisions, this means that the revisions haven't been accepted by anyone with the easy-to-obtain privilege. Obviously, such edits can't be considered to be reliable, no matter who made them.

There is an issue of responsibility that has been raised with BLPs and libel. If I sight a revision that creates libel, I have joined the original editor in personal liability, whereas with the present system, only the original editor is liable, there is no requirement on anyone that the libel be removed, until and unless Wikipedia is contacted by the potential plaintiff. What happens with flagged revisions is that someone else must be willing to assume personal responsibility for that edit, as if they had made it themselves. With "POV-pushing edits" that involve policy violations, we now must have two editors who risk sanctions, not one. Is this a restriction on editorial freedom? I think not; in fact, damage from "POV-pushing" edits is greatly lessened and is confined to the labor of sighting them, instead of creating a problem with readers. My sense is that Wikipedia can become, to some degree, more tolerant and conflict will be more focused on disputes between those with sighting privileges; if we see "tag-team" sighting, if you'll approve mine I'll approve yours, we will have more focused debates and, if necessary, sanctions that don't bite newcomers who stumble into a minefield, just because they take us as our advertising claims: "anyone can edit."

Sighting a revision should not merely be a confirmation that it isn't vandalism. It should be the sighter taking responsibility for the edit as if it were his or her own, which would require verifying sources, etc, unless the sighter is so convinced that an editor is reliable (and sources are difficult to verify) that the sighter is, indeed, willing to assume reliability. Everyone makes mistakes, but if I think that an editor is reliable in this way, it means that I consider the probability of error to be so low that it matches, more or less, my own error rate or is even better. Nobody is obligated to sight a revision, but there is obviously a very good reason why we confine sighting to experienced editors. It's risky. And it is valuable and essential, so doing "sighting patrrol," done well, will be far more of a service to the project than is current vandalism patrol, which is like falling off a log, an easy video game to play in odd moments. Sighting would ordinarily still be easier than writing, because the sources should be laid out and a draft text put together. We need to work out certain details.... thus we really need a test of the concept, i.e., the present proposal. --Abd (talk) 17:21, 18 January 2009 (UTC)

If sighting an edit makes me just as responsible and liable as the original editor, I shall refuse to sight any edits. DuncanHill (talk) 17:42, 18 January 2009 (UTC)

I want "sighting" to be a lightweight check for obvious badness; it should merely be a confirmation that it isn't vandalism. The very name "sighting" implies simply that somebody has looked at ("had sight of") the version or the diff under consideration. I believe that there's also scope for a more heavyweight "reviewing" process, but please don't use the same term for both concepts. —AlanBarrett (talk) 20:27, 18 January 2009 (UTC)

Flagged revisions allows for "lightweight" sighting and "deep validation" sighting. But in either case, I'd suggest, sighting an edit does make the sighting editor responsible. Suppose an edit is a BLP violation. If you sight it, it may then escape further notice. I'd say that while "lightweight" sighting may be of some value, it's only valuable for vandalism detection; vandalism patrol currently involves a high inefficiency, as many editors may look at a few edits, and may miss others. Having a sighted flag that is used this way -- the edit isn't obvious vandalism or obvious BLP violation -- would be worthwhile, but wouldn't greatly increase reliability, outside of more quickly and reliably identifying vandalism. The fact is that any reader can also detect vandalism. "Oh, I see that the politician I was considering voting for was caught having sex with a cow. Says so on Wikipedia, must be true!" More likely, if it wasn't sourced and verifiable, the reader, as an IP editor if not registered, might remove it. I've seen a fair amount of vandalism reversion from IP editors.... It will be important exactly how flagged revisions is implemented, and there certainly could be layers of implementation.

As to being unwilling to stand behind a sighting of an edit, to be responsible for it, if this is frightening, so too should any editing be frightening. If you check an edit out, look at the source cited, looks good, you'd be pretty safe, even if it did turn out that something was awry: you'd have "deniability." I.e., it looked reasonable, and reasonable people would agree that, even if you were wrong, you weren't reckless. With a sighting, you have now two editors standing behind an edit, so if you've done due diligence (the level that's due in connnection with the type of sighting that is established), you'd be quite safe. However, I'm not concerned about editors who don't want to sight, who just want to edit articles. That's fine! Not everyone needs to do this kind of work. I'm also unconcerned about backlog. As long as anyone can see *all* the editing, not just the sighted edits, it's an improvement, not a step backward. --Abd (talk) 01:54, 19 January 2009 (UTC)

If I revert obvious vandalism, am I then "sighting" the previous version (i.e. the one I have reverted back to)? DuncanHill (talk) 02:06, 19 January 2009 (UTC)

No. --Abd (talk) 03:00, 19 January 2009 (UTC)

It's an excellent proposal. It simply calls for removal of 99% of existing content and reducing it to the body of actively edited and actively watched texts. "Substantial delay in sighting revisions" - if today nobody cares to watch/edit/reference the bulk of wiki articles, why would they go for "sighting", whatever it means?. NVO (talk) 03:35, 19 January 2009 (UTC)

Why assume such a singularly stupid method of implementing it? One of the problems is that Wikipedia creates work and rework and rework and more rework until we are sick of it. Some of the work that now goes into rolling the boulder back up the hill would go, I predict, into article validation. Who knows, we might even attract some experts, if we stop insulting and blocking them. --Abd (talk) 05:17, 19 January 2009 (UTC)

Most of the work prevented is by making it less attractive to anons to make an edit (-50%) and about attracting new users, that is going down at an abnormal speed (User:Hut 8.5/German editing stats). Mion (talk) 13:27, 19 January 2009 (UTC)

Why? The text above says "If there is substantial delay in sighting revisions, this means that the revisions haven't been accepted ... can't be considered to be reliable, no matter who made them." This means that these revisions must be deleted, doesn't it? NVO (talk) 17:46, 19 January 2009 (UTC)

To quote WP:V; "Wikipedia is not a reliable source." FR won't make it a reliable source; most of our serious problems are caused by the incompetence or malice of long-established accounts. "[Politician X] is a poopyhead" is a minor nuisance, which any sane reader will ignore for the minutes before RCP catches it, and that is what the advocates of FR claim it can do. Septentrionalis PMAnderson 14:42, 26 January 2009 (UTC)

Announcement from Jimbo Wales

See [5]. DuncanHill (talk) 23:04, 21 January 2009 (UTC)

A light weight alternative

Instead of making the flagged version the default view, put an icon or tab on each page that would let a reader see the flagged version with a diff from the current version. Add an opt in preference for those who wish to see the flagged version by default. We would still be "the encyclopedia anyone can edit," but readers could quickly check and judge for themselves whether recent changes are likely to be valid. Most vandalism, in my long experience, is quite obvious in a diff. For the rare exceptions, readers would have to follow up on any refs supplied. Add a simple way for trusted editors to update the sighting after checking a diff and things should work smoothly without any 21 day limits. --agr (talk) 02:46, 22 January 2009 (UTC)

Virtual domain alternative

As another alternative, I suggest a different 'virtual domain name' for a Wikipedia site that shows the flagged revision by default. i.e. new.en.wikipedia.org would show all new edits,

     reviewed.en.wikipedia.org  would show the flagged version.

en.wikipedia.org would go to a site asking the visitor to choose which version of the article they want to see. Leave the choice up to the viewer, I suppose.

The option should only be presented for articles that have a reviewed version. --Mysidia (talk)

It should be presented for all articles but with a banner of "No reviewed version for this article" for non-reviewed articles. That way people could make it their default site and not be switching between the two. In any case... I think it'd be hard to get this to happen... gren グレン 07:18, 26 January 2009 (UTC)

Request for arbitration

I've filed a case to ask the arbcom to look into Jimbo's announcement that flagged revisions will be turned on, and whether he has the support of the community/arbitration committee to do so. Sceptre ^(talk) 15:59, 22 January 2009 (UTC)

Asking a bureaucrat to look at it might be a better idea. They at least have some mandate for judging consensus in discussions. Hut 8.5 16:13, 22 January 2009 (UTC)

Yes. In fact, it would've been better for a 'crat to declare consensus, or lack of. Consensus cannot be judged by Jimbo impartially as he is a known major proponent of flagged revisions. However, as Jimbo has declared it by fiat, the Arbitration Committee are the only people who can overturn it. Sceptre ^(talk) 16:18, 22 January 2009 (UTC)

Although, mind you, Brion also is allowed to refuse to implement it. And he may well do so, as the poll falls below his requested consensus. This may make it a moot point, but there are issues behind the closing that need to be addressed to. Like whether Jimbo enjoys immunity from our guidelines and policies or not. Sceptre ^(talk) 16:23, 22 January 2009 (UTC)

I don't think this is really in the scope of ArbCom. It may be true they can override certain rulings, but there's nothing that says they can prevent a founder decision someone disagrees with. It's a big can of worms to open, to say someone can bring to ArbmCon any one action someone disagrees with. --Mysidia (talk)

It ought to be in ArbCom's scope; if not, who reviews Jimbo's actions? The real world has had quite enough of prominent persons who act without review, prompted by idiot analyses in the Washington Post (if not the Washington Times); I don't see why Wikipedia should take up the practice. Septentrionalis PMAnderson 00:45, 26 January 2009 (UTC)

The community, perhaps? It could be a little different, I suppose, if the community overwhelmingly objected to Jimbo's actions on this matter, and Jimbo refused to budge, and further escalated matters. But Jimbo's decision is essentially a policy matter, and ArbCom's reason for existing doesn't have anything to do with setting policies.

Arbitration is supposed to be a last resort during a dispute process, when mediation, discussion, and all other possible measures have failed. There is no evidence that there has been any attempt to the mediate the matter, or for that matter, that there is even truly evidence of a dispute that needs to be settled. --Mysidia (talk)

The Kennedy test

Has anybody else actually looked at what happened at Ted Kennedy after his collapse on Tuesday? The edits which involve a claim of his death are listed at Talk:Ted Kennedy#What happened.

This is what happened, and the likely effect of FR if it had been in effect on the article at the time of an unexpected event. The first edit on the subject was at 19:44 UTC

1. An anon inserted Kennedy's death at 19:50; another anon took it out at 19:51. This would not have made to the sighted version under FR; but it would have taken longer to remove from the unsighted text, because no anon would have seen it.
2. Kennedy's death date was added to text by Newbie A (a vandal) at 19:59
3. Before this was removed, Anon B added "January 20, 2009" as a free-standing paragraph.
   • An admin reverted this to Newbie A's edit. Under FR, this would have sighted the vandalism. Anon C removed Newbie A's edit at 20:03; this would not have sufficied under FR.
4. Newbie A altered the death date in the first line at 20:02, and this was removed at 20:04 by an established account. FR would have kept this out of sighted version.
5. An established account added Kennedy's death at still another place at 20:02, but noted no confirmation in the edit summary. Under FR, this would probably have been flagged; it was removed at 20:05.
6. another pair of anons put in a claim of death, and took it out again immediately, just before the article was semi-protected at 20:03.
7. The death date was restored by an established account at 20:06, and removed at 20:08. Under FR, this would have been flagged. (The edit summary on the insertion was rvv, so it may be a confusion as to which way he was editing, or like #5, a confusion on the facts.)

Out of these seven, three would have made it into the sighted text. I think the case that FR would have prevented this, as the Washington Post praters claim, is open to much doubt. Septentrionalis PMAnderson 00:45, 26 January 2009 (UTC)

Thank you for the detailed analysis; it certainly weakens the claim that the false claims were 100% preventable. Personally I'm not averse to a big test as long as there's a sunset clause that's a short time away (like, two weeks). Tempshill (talk) 04:02, 26 January 2009 (UTC)

Thanks for the analysis, although I'm not sure I can agree with you on your bullet to point 3. When the admin reverted the edit, his edit would not have been autosighted as it was made on top of unsighted edits. In order to make the changes visible, he would have needed to review the article manually, whereupon he would have seen a diff of all the changes since the last sighted version, including the changes by Newbie A. As an unsourced and contentious claim, it would certainly have been noticed and in all likelihood removed. But there is no risk that the addition of the death-date would have been 'accidentally' sighted because that's not how the system works. Happy‑melon 08:56, 26 January 2009 (UTC)

Another analysis

I analysed the edits to the article from the beginning of 2009 up to 22 Jan. Some things I observed with respect to FR:

1. The article is only editted 7 times in 2009 until reports of the seizure trigger a flurry of activity
   1. 2 of those edits are from registered users and make unsourced and unnattributed changes to content in the infobox
   2. none of those edits were malicous, even with 2 edits from an IP (formatting changes)
2. For all the subsequent IP editor contributions until protection is affected (20 Jan, 19:44 - 20:03):
   1. 5 IP editors make 'good' edits, including 1 adding {current} and 1 removing a poor POV/peacock paragraph present since the start of the year, the remaining 3 reverting death claims
   2. The 3 IP users reverting false death claims outnumber the 2 that add it (one reverter being the first user to do so on the article)
   3. 8 other editors are attempting to update the article with info they presumably got from sources, or improve that info added by others, but none of this effort is referenced
   4. 1 IP makes a test edit
   5. No IP editor manages to add any properly sourced info before the article is protected
3. Following semi-protection, in order:
   1. Enigmaman (talk · contribs) rewords Chappaquiddick incident, changes "The Senator swam to safety, but Kopechne died in the car" to "Ted Kennedy swam to safety, leaving Kopechne to die in the car." This lasts from 20:54 to 21:13.
   2. DoenerKebabMitKohl (talk · contribs) adds that Kennedy "and urinated in his trousers" to the lede. This lasts from 01:37 to 01:42 (after a self-revert)
   3. Tatterfly (talk · contribs) adds Category:People with epilepsy to the article. This lasts from 05:53 to 05:58, and 16:01 to 16:12
4. Since the start of the year the article had included the statement "District Attorney Dinis chose not to pursue Kennedy for manslaughter", which was not removed until the last edit on 22 Jan, well after all the attention from the seizure.

I am open minded as to what this may or may not mean, I may or may not come back with some firmer conclusions. I note initially that all edits made by non-trusted users would not have been immediately sightable without modification under WP:BLP, even though their edits are overwhelmingly good intentioned, (and in this small set don't match some of the stats about IP vandalism I have seen) and can even excel in timely and helpfull affirmitive action, if not in content referencing. I also note the difficulties in grading/judging what constitutes a libelous/defamatory edit, or what just makes Wikipedia look bad. I also note that in 'current event' type situations, queuing and bunching together proposed changes has obvious merits in avoiding the mistakes made by all levels of user, even admins. I also note the possible difficulty of judging whether editors, even long term ones, deserve trusted status.

Possibly the most pertinent for any trial, I note that to meet Jimbo's stated claim that FR would have "100% prevented" the press attention this got for Wikipedia stating Kennedy had died for a few minutes, somebody has to make the case for why this unprotected, quiet, edit war free article that had previously laid un-molested for 20 days would have been under FR at all to make that claim true, particularly if the premise is not to be simply that all BLP's/all of Wikipedia, are to be subject to FR.

MickMacNee (talk) 05:22, 26 January 2009 (UTC)

26 January 2009 - Practicality

I would like to address the practicality of flagged revisions.

In my opinion, applying it to all articles or to all new articles would be incredibly counterproductive. I find that a huge number of articles are simply not checked over for long stretches of time. Many articles in the categories I focus on, (independent music in particular) do not receive high editing traffic. Similarly, I focus on copyediting and counter-vandalism, and there are a great many of these things that are ignored and stay in place. Many of the factual errors I have encountered have been in place for long periods of time as well. If we do not already have enough editors checking over these things, the introduction of flagged revisions for new articles would by necessity cause a backlog. I honestly don't think that we can cope with it, particularly if the number of 'trusted editors' is small.

If something really must be done to all Wikipedia pages, I would support semi-protection rather than flagged revisions. I don't like either of them, but semi-protection seems the lesser evil to me. It would cause no administrative backlog, nor would it stifle registered users. It certainly wouldn't stop all vandalism, but that is an unrealistic goal anyway.

I can understand using the flagged revision system in a pragmatic manner to target problem articles in the short term, if that really is necessary. If flagged revisions is implemented at all, we'll need to have the maximum possible number of trusted editors, so I suppose an automatic process would be necessary. A limited revision period would be nice too, so that an ignored edit will not forever be condemned to limbo. Some other people on this page have also raised other suggestions to minimize the harm.

I hope that the Wikipedia community will not be overzealous in their efforts to combat vandalism with this method, because it has the potential to cause a much bigger problem than the one it was meant to solve. BecauseWhy? (talk) 04:01, 26 January 2009 (UTC)

In my opinion, applying it to all articles or to all new articles would be incredibly counterproductive. -- good thing noone's planning that then, huh? ♫ Melodia Chaconne ♫ (talk) 04:36, 26 January 2009 (UTC)

Well, it appears to me that there are people advocating this in various places. For example, over at the BLP protection discussion, there's a whole section for it. Similarly, it appears that a report about the German Wikipedia's implementation has reported the type of problems that I mentioned regarding workload. So, I thought it important to address the issue. Granted, I have been absent from Wikipedia for a while, so I have probably missed things. BecauseWhy? (talk) 05:06, 26 January 2009 (UTC)

On the off chance nobody noticed it...

I was very surprised that searching for "BBC" on this talk page didn't find any matches...

[6] (featured prominently on the main page of the BBC News International website)

Nice going, Jimbo. heh, it's amazing that the same guy who came up with the brilliant idea for Wikipedia also has such a poor vision for how to make the project succeed... --Jaysweet (talk)