Wikipedia talk:Did you know/RFC DYK process improvement 2015

This is an RFC to improve and simplify the DYK review process. — Maile (talk) 14:59, 2 June 2015 (UTC)

Suggestion for overall workability

 * DYK already has a three-editor approval process: (1) Reviewer; (2) Promoter to prep; (3) Promoter to queue
 * If it's not working, one or more steps are too time-consuming to be carried out correctly.
 * Step 1 - Initial review - everybody gets to post their opinion, no limit
 * Step 2 - Promoter to prep is required to re-review each article promoted
 * Step 3 - Promoter to queue is required to re-review ALL articles in a set
 * Steps 2 and 3 are too often occurring as a last-minute emergency when nothing is out there. Not feasible as such, given the requirement to review each and every hook involved.


 * DYK has always been buried in rules creep scattered over multiple pages. It needs a one-page simplification, as in WP:GA? and WP:FA?, something so clear even a first-time editor can grasp it at a glance. No need to require the three-person vote on promotion of each nomination to make this work.
 * Form a work group that includes Admins to revise and simplify the message. Suggest DYK veterans, , , , as needed participants. Others as so motivated.
 * Take the rules revision project to its own page, devoted exclusively to the task.

The process itself is too convoluted and buried in commandments to work as is. Re-structure the whole process, rather than patchwork that adds even more rules. — Maile (talk) 13:19, 2 June 2015 (UTC)

Basic criteria checklist

 * QPQ (what is the current rule on this?)
 * Has not appeared previously on DYK
 * Has not appeared on In the News as the qualifying linked and bolded article
 * Minimum size - 1,500 minimum characters of readable prose
 * Eligibility
 * New - No more than 7 days old in article space (not AFC) at nom
 * Improved - 5X within 7 days of nom; 2X for previously unsourced BLP
 * GA - Promoted to GA within 7 days of nomination


 * Verifiable sourcing
 * Hook(s)
 * 200 characters or less
 * Must be supported by article prose and cited there in
 * Must be neutral and not violate WP:BLP
 * Must make sense and have correct spelling, punctuation, and grammar
 * No redlinks


 * Check for copyvio/close paraphrasing
 * Check for neutrality
 * Within policy for WP:BLP
 * Image - Must be in the article and properly licensed
 * Sounds - Must be in the article and properly licensed and formatted with DYK Listen
 * No dispute tags
 * Not a stub
 * Stable - no edit wars

Nominations

 * Each nomination should have a minimum of two reviewers to pass.

Promotion to prep

 * The promoter is responsible on each hook they promote for re-checking that all criteria have been met.

Promotion to queue

 * Admins should make a final check of all hooks in the set for spelling, punctuation, grammar, and whether or not it makes sense.

Discussion
I'm up for that if others are. EEng (talk) 13:44, 2 June 2015 (UTC)
 * Highlight one of step 2 or step 3 as the place to re-review. Problem is with having two places is that each might think the other place double-checked a review when neither did. So just make it one of them. Cas Liber (talk · contribs) 13:24, 2 June 2015 (UTC)
 * I absolutely think this is the heart of the problem. Right now accuracy and compliance are everyone's responsibility, and therefore it's no one's responsibility. EEng (talk) 13:44, 2 June 2015 (UTC)
 * Requiring it to be Step 2 might encourage more Admins to step in at the last minute to promote to Queue. — Maile  (talk) 13:35, 2 June 2015 (UTC)


 * Thinking aloud, we possibly just need a "GA-lite", which focuses on criteria #1 (well written) and #2 (factually correct) but not the others. Would take an experienced reviewer 10-20 minutes to spin through, depending on the article size. The nominator works through the article raising issues exactly like a GAN, and there's dialogue and immediate feedback. At the end, a pass goes straight onto queue for the main page. It's quite different for what we do now, but having a far more in-depth review and less prep / queue checks should give us the balance we're looking for. We'd put the wrath of God on someone saying "long enough, has sources, good 2 go" for a GA review, so why don't we do the same here? Ritchie333 (talk) (cont)  14:03, 2 June 2015 (UTC)
 * Checking for copyvio is a biggie, also, when putting something on the main page. On the plus side, GA has review templates. On the minus side, GAs nominated here have proven that GA reviewers don't always really check what they're supposed to in order to pass it. — Maile  (talk) 14:11, 2 June 2015 (UTC)
 * You'd want to create WP:WIADYK which covers everything, size and newness being other criteria. However, an experienced reviewer can spot a duff GA a mile off - Talk:Queen (band)/GA1 for example. It should be easy for somebody assembling the queue to spot a review hasn't been done properly and send it back. Ritchie333 (talk) (cont)  14:14, 2 June 2015 (UTC)
 * How about re-reviewing at step 2? Then, the reviewer at step 3 doesn't have to check an entire queue for a single error in one article. Epic Genius (talk) 14:15, 2 June 2015 (UTC)
 * How about, if we find that a review was inadequate, we put to death the reviewer, the nominator, the promoter, and the article's author, all at once on live Skype hookup. That should eliminate shoddy reviews and cut down on the surfeit of nominations, all at once. EEng (talk) 19:49, 14 June 2015 (UTC)


 * I'm uncomfortable with Ritchie's suggestion of "straight onto queue for the main page" after a single person's review. But I think we're not facing some unfortunate facts:
 * Doing a better job of reviewing will require more work per article, and not everyone is good at doing it
 * Not all nominators are capable of bringing their article up to standard without someone else essentially doing it for them
 * (therefore) We may need to dump the current practice of passing almost everything nominated.
 * It also seems to me it's a useful division to separate responsibility for checking the article from responsibility (at a later phase) for rechecking the hook (verifiability, using the article; and all the little technical details that often get overlooked, like dabs). EEng (talk) 14:31, 2 June 2015 (UTC)


 * The idea is that it should take a longer amount of time for each individual reviewer, with a greater possibility of simply failing (as with GAs), with the trade-off that the time to check the queue is much smaller. Bear in mind it's possible for two newbies to review each other's GAs, resulting in a bunch of half-assed articles with a big "this is a good article" stamp on them - at least DYKs get a wide audience to trap this sort of thing immediately. Stuff like dabs should re-use the tools available, same for checking deadlinks. I don't want to name names, but I recently reviewed a DYK that I thought "nope, not up to snuff". At a GA review, the conversation and responsibility could end at that point. Ritchie333 (talk) (cont)  14:47, 2 June 2015 (UTC)


 * Hmm... maybe we can have something like !votes and comments for DYKs, kind of like in an RFC, except the DYK nom only needs to run for a week. Epic Genius (talk) 15:10, 2 June 2015 (UTC)
 * Wouldn't that be duplicating what already should have happened on the nomination template (and sometimes runs for months before an approval happens)? — Maile (talk) 15:22, 2 June 2015 (UTC)
 * Yeah, but the things is, many people think that there need be only one reviewer per DYK. Epic Genius (talk) 17:11, 2 June 2015 (UTC)
 * I second this. There is usually just one person per review level, at varying degrees of effectiveness. No one person is going to notice every single problem and regulars who scroll through noms/preps/queues are not going to catch everything either. Make it more collaborative and encourage more comments at the nomination stage, à la ITN and FA, so that more editors can highlight potential problems at the beginning rather than complaining when it hits Main Page/Errors. If a nomination is not worked on within that week, close it and move on. Fuebaey (talk) 18:56, 2 June 2015 (UTC)
 * My worry is that we improve the process until it creates no errors on the main page ... and forget what we are here for. The purpose of the project is to create an encyclopaedia. We are doing very well at that - if we break a few eggs then ,"sorry, I'll clear it up and who wants some omelette?") There are areas of UK/US life which are 9X% done, but there are thousands of women, Chinese heroes, Vietnamese landmarks etc. that are unfinished and poorly covered. Many of these would be happy to get a "start" article. If we want to create a process that creates no errors then its too easy. (Just don't approve anything) If we are trying to keep our most ranty and uncivil critics happy then that's impossible. (They enjoy complaining or they'd stop doing it with such emotion). The project still needs new (imperfect) articles and even more than that it needs new (and imperfect editors). If the objective is still to deliver new (and imperfect!) articles and new (and imperfect!) editors then count me in. If you want to create perfection or to please the insatiable then I suspect we/you have forgotten what we are meant to be doing. (My positive suggestion would be to require reviewers to edit the article that they review and to add one extra source. That would mean that the reviewer would need to understand the article and it would require then to check the main substance of the subject. Is this a magic bullet?.... no. Will the same people rant about some error they were too lazy to fix in an uncivil way - of course). Victuallers (talk) 16:06, 2 June 2015 (UTC)


 * , as long as human beings are involved, it will never be perfect. The idea is to get the process streamlined.  We have too many directives for interpretation on too many pages.  Just get the directives down to something simple.  And decide exactly how many re-reviews have to be done by whom before a nomination is on the main page. — Maile  (talk) 16:18, 2 June 2015 (UTC)


 * I support the above proposal, especially needing 2 reviewers to pass. (just covering all the bases...) Epic Genius (talk) 02:58, 3 June 2015 (UTC)


 * I oppose any process that requires more review. We already too few reviewers, how is demanding more supposed to help? I'd still be careful at the next zebra crossing. Maury Markowitz (talk) 11:57, 8 June 2015 (UTC)

I'm all for fewer rules. Just remember the only thing worse than endless rules is endless unwritten rules, applied by reviewers but unknown to article writers until after they have gone through preparing everything. The way to simplify the readable prose rule, for instance, is to eliminate it (and adjust the 1500 limit to compensate), not to "simplify" by keeping the rule and eliminating the explanation; that should cause more frustration, not less. Art LaPella (talk) 04:06, 7 June 2015 (UTC)

On second thought, maybe eliminating explanations is the only practical way to get such rules eliminated. After all, if everybody is furious because they thought readable prose simply means article length, until after they write their short article and it's too late, then maybe the rule won't get enforced any more. There never has been consensus for a more rational process. Art LaPella (talk) 05:13, 7 June 2015 (UTC)

DYK is supposed to be about new articles. New articles are commonly 1) imperfect, 2) short, and 3) written by newbies and seasoned editors alike. So... 1) A GAN-LIGHT approach is ok, but requiring perfection is not. 2) The longer the article, the more likely that an issue will slip through the reviewing cracks. I'm unclear how we would deal with an article cap limit, but let's consider it. 3) Whatever changes we make in the rules process, they should not be so egregious as to inadvertently deter participation by newbies. Rather, they should be newbie-friendly. --Rosiestep (talk) 15:56, 7 June 2015 (UTC)

I second Rosiestep's point about DYK's not having to be perfect. The entire point of having them is to showcase the wide variety of new and improved content and to encourage contributions. The current DYK requirements are a good baseline for article quality, and I do not see a reason to make them more stringent. Also, I'm worried that increasing the number of equired reviewers or making DYK like FA is just going to result in a massive backlog. Spirit of Eagle (talk) 00:04, 8 June 2015 (UTC)

Comments: This proposal starts by noting that feature creep has made this process too difficult for new users. While that is certainly true, it is also true that a number of technical features of DYK make it overly difficult. The main issue is that it is a manual multi-step process, for reasons that elude me. Another is that the instructions for the "next step" of the process are always on the page you just left ("The preceding program..."). One of these steps requires a careful edit to a very long page, where a single-character error will break everything. Now let's say you get through the process and then find there's an error. You click edit and end up on a page that looks nothing whatsoever like the page you originally edited. I find it practically impossible to figure out how to add an image to the nom, for instance, without opening another window, navigating to the nom page, editing a nom that has an image, and copy and pasting the text. This is abysmal. Perhaps we might consider fixing these actual problems before starting an existential debate on what DYK "is"? Maury Markowitz (talk) 11:53, 8 June 2015 (UTC)
 * And I should add that the review process could be greatly improved as well. We have review tools that either do nothing, literally, or produce useless output. Tools that aren't useful, like the duplicate link checker, aren't included (and perhaps can't be). If most of what the review comes down to can be stated as a rule, then it can probably be stated as an algorithm. Doing so would allow the human part of the chain to focus on what the computer's can't do, which, IMHO, is the major issue with DYK reviews, that people simply ensure every rule has been followed and don't actually care if the article itself is absolute crap. Maury Markowitz (talk) 12:02, 8 June 2015 (UTC)


 * I had completely missed that all this is going on. I'll try to catch up in the next day or two. EEng (talk) 14:45, 10 June 2015 (UTC)
 * I'm sorry, I'm not able to follow the above discussion because I'm not sure what we're talking about ("putting Step 2 before Step 3"?). However, I'd like to note that the way the rules are currently structured, you just have to check the history for the start date, find the hook ref cited inline in the article, check that all paragraphs have at least one citation, and do an Earwig's Copyvio check to see there's no full-scale copying going on. YOU DON'T HAVE TO ACTUALLY READ THE ARTICLE. I think this is one reason why articles are passing with bad and no English grammar. This focusing on the rules has also tripped me up when I've promoted a hook to prep without actually reading the whole article, and then had my hand slapped on WT:DYK. Therefore, I think that the checklist for reviewers to sign off on should begin with: 1) Read the whole article. 2) Confirm that it conforms to basic English grammar. Saying it has to be "well-written" is too vague; the average Joe doesn't have a clue what good writing is. Yoninah (talk) 16:42, 14 June 2015 (UTC)
 * Yes, read the whole article. Don't just check off the tick boxes.  What we see is reviewers doing all the technical stuff (it's been expanded, the hook is cited etc) but none of the checking it's an article we want to "showcase" on the main page.   The Rambling Man (talk) 20:22, 14 June 2015 (UTC)

There are three steps to adding a DYK: So basically the entire process is a UI disaster. Why is this not one step, and completely automated? Maury Markowitz (talk) 15:42, 15 June 2015 (UTC)
 * 1) type in the name of the article in the main nom page, press the button
 * 2) a new page appears, where the user types in the details of the nom. This page includes instructions on how to complete step 3. Something like  When you click Save page, these instructions disappear.
 * 3) Step 3, adding the nom to the list, is entirely invisible. You are not taken to a page where you can add this, and there is nothing on the screen telling you what to do.
 * I agree with Maury. I've always wondered why DYK and AFD are not automated; both require you to manually add your nomination to a daily list and notify interested users yourself. In contrast, nominating a Commons image for deletion is completely automated. One click starts the whole rigmarole in motion, slapping a template on the image, opening a discussion page, and notifying interested users. Yoninah (talk) 16:04, 15 June 2015 (UTC)
 * Well it's always been one of the things that's amused me, that there's a perception that DYK is for newbies and to encourage editors to climb onboard Wikipedia but it employs the most esoteric coding to do so and turns off even seasoned regulars, let alone people who have no clue why we're working in Template space and why we need to transclude things etc etc. Simplify the process to encourage the actual newbies, not just the WikiCup editors.  The Rambling Man (talk) 16:55, 15 June 2015 (UTC)
 * Well, if you've got Twinkle there's an XfD item in the dropdown that makes deletion nomination easy as pie, so clearly such things can be automated. The difference with DYK is that there are a lot more options and gizmos to fill in. However, I really feel we should focus first on what happens after the nomination is made -- ideal or not, the process of creating nominations is clearly functioning, while the process of reviewing them is not. EEng (talk) 19:14, 15 June 2015 (UTC)
 * True. So has this thread, once again, lost its way?  The Rambling Man (talk) 19:20, 15 June 2015 (UTC)

Suggestion
So a couple weeks ago I proposed a change from 7 to 30 days for the newness criteria. See Wikipedia talk:Did you know, which seemed to have a lot of support. Meanwhile, although I haven't commented here I've been following this discussion and the concerns about the quality of articles appearing in the DYK section on the main page. I've started an attempt to revise some of the criteria and rules in my sandbox. The main changes there are making it 30 days and adding a requirement for reviewers to check for grammar, spelling, and overall readability. I'd appreciate help/feedback from anyone here in trying to draft new rules. ~  ONUnicorn (Talk&#124;Contribs) problem solving 22:27, 16 June 2015 (UTC)
 * I appreciate the effort, and a lot of what you've done looks really good, but changes to parameters like 7 vs. 30, or whether there's a grammar requirement, should not be rolled up with reshaping the workflow and improving the presentation of rules and procedures. They're independent. So let me suggest that the 7/30 issue, and the grammar/style criterion issue, continue as they have on the main TDYK. (Though if we're not careful they'll go moribund.) Here we should be talking about the reforming the workflow. EEng (talk) 23:43, 16 June 2015 (UTC)