Jump to content

Wikipedia talk:Criteria for speedy deletion

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

This is an old revision of this page, as edited by Jc37 (talk | contribs) at 05:40, 12 February 2023 (add). The present address (URL) is a permanent link to this revision, which may differ significantly from the current revision.


RfC: Removing F10. Useless non-media files

The following discussion is an archived record of a request for comment. Please do not modify it. No further edits should be made to this discussion. A summary of the conclusions reached follows.

Result - Deprecated in favour of WP:PROD usage.

If it turns out that PROD isn't handling things as hoped, feel free to revisit this discussion in the future. - jc37 05:35, 12 February 2023 (UTC)[reply]

I commented out the F10 section. Someone is welcome to please archive as appropriate. - jc37 05:40, 12 February 2023 (UTC)[reply]


Should speedy deletion criterion F10 (Useless non-media files) be removed? --Trialpears (talk) 05:16, 9 January 2023 (UTC)[reply]

F10 was created in 2008 to deal with huge amounts of Word documents, spreadsheets and PDFs in an era before file PROD. The environment has changed a lot since then and I believe it's time to deprecate it now.

We no longer have to deal with most of the file types impacted by the criteria since we have restricted what file types can be uploaded. Today the only files that are neither image, sound, nor video files which can be uploaded are PDFs (technically DjVus as well but we currently don't have any of those). The number of PDFs uploaded has decreased as well from several 100 per month to just 76 during all of 2022.

The criteria itself is also seldom used. Ignoring deletions I made while converting our last bmp files to pngs, there have only been 137 F10 deletions since 2019. Compare this to the 193 files that are currently being proded and it's clear that we can handle these deletions through proposed deletion without any issue. This would also increase accountability by giving a week delay where any editor could oppose the deletion. --Trialpears (talk) 10:17, 7 January 2023 (UTC)[reply]

Should probably link this 2021 discussion on the subject. The PDF count today is 268 and the unuploadable files count is 1 and that one is being deleted as well. I've quickly looked through all of the pdfs at User:Trialpears/PDFs. While a bunch of them are still orphaned they are in my opinion on topic enough to be worth proding rather than just deleting immediately. The category I believe speedy deletion is justified and useful for is Category:Images in non-image formats where a F10 after uploading a converted image is reasonable imo and something I'm currently working on. For these cases I would argue that this can be done without F10 as well since the pdf page gets redirected if its history is merged into the converted image. --Trialpears (talk) 10:40, 7 January 2023 (UTC)[reply]
Reading the 2021 discussion it seems I was opposed then, in part because Trialpears' comments there, including usage statistics, indicated there was still some use. However it seems things have changed since then and other processes can handle the load more than well enough, so I'm happy to support this time around. Thryduulf (talk) 15:55, 7 January 2023 (UTC)[reply]
Seems reasonable. There's no need for it if PRODs work just as well and allows editors time for justification of keeping the content, if needed. SilverserenC 00:06, 8 January 2023 (UTC)[reply]
Agreed, sounds appropriate. Jclemens (talk) 00:44, 8 January 2023 (UTC)[reply]
Agreed with everyone else above. * Pppery * it has begun... 01:17, 8 January 2023 (UTC)[reply]
+1. Fastily and Cryptic, do you still oppose deprecating F10? If not, I think we can avoid a formal RfC. HouseBlastertalk 03:07, 9 January 2023 (UTC)[reply]
My stance from 2021 is still the same. I'll only support deprecation if the ability to upload PDFs is removed. Also, if we're talking about deprecating CSD criteria, then this really should go through a formal RfC. -FASTILY 04:07, 9 January 2023 (UTC)[reply]
I'm happy to slap an RfC tag on this since you asked.
Fastily With regards to your actual argument I just want to add an update version of Cryptic's quarry of PDF uploads per year. As you can see the last two years have only had 52 and 76 PDF uploads respectively which is a lot less than the number of 200 mentioned in the 2021 discussion. --Trialpears (talk) 05:13, 9 January 2023 (UTC)[reply]
So what? People can still upload PDFs. F10 provides an excellent avenue for dealing with inappropriate PDFs, and I see no legitimate reasons to remove this avenue. Like I said, I'll gladly support deprecating F10 when editors are prevented from uploading new PDFs. -FASTILY 21:39, 9 January 2023 (UTC)[reply]
If we were enacting CSD criteria today, would F10 still meet all four clauses? The third, frequent, is at best borderline, and I don't think we've even really addressed objective or uncontestable. Jclemens (talk) 06:36, 10 January 2023 (UTC)[reply]
Admittedly this is going to be somewhat of a sidetrack, but the question regarding F10 meeting all four clauses made me think of portals, as we have two criteria on the books: P1 (propagation of CSD of articles) and P2 (underpopulated portal). I obtained 195 and 366 hits for P1 and P2, respectively, since 2006. As for the other four clauses, P2 in practice does not meet the "objective" clause, as when it is applied, it is a weird of mix of G4 and expanding the strict requirements as to how many articles should be available to base a portal upon. Both P1 and P2 arguably meet clause 2 ("uncontestable"), but portals were the locus of an acrimonious dispute not so long ago, which, in fairness, has appears to have simmered down since. Criterion P2 is clearly nonredundant (clause 4) although of limited utility, whereas P1 is arguably simultaneously redundant and not redundant, as it propagates the CSD of articles. It may be cleaner to change the first sentence in article subsection to read "These criteria apply only to pages in the article and portal (main) namespaces." All of this to say is that F10 meets all four clauses better than the portal speedy deletion criteria. Maxim (talk) 18:26, 11 January 2023 (UTC)[reply]
Then fix them too..? WP:OTHERSTUFFEXISTS probably applies here just as validly as to article deletion. casualdejekyll 19:39, 11 January 2023 (UTC)[reply]
If the usage rate is this low, then PROD should be able to handle this instead of CSD. Furthermore, leaving it to PROD and FFD provides a longer, though not overly long, window in which interested users could ameliorate the root issue of F10 (i.e. non use in articles) if appropriate. ---- Patar knight - chat/contributions 20:05, 12 January 2023 (UTC)[reply]
  • What's the problem with this criterion? We don't want people to upload a curriculum vitaæ, or an article as a PDF. I note that the count provided by User:Trialpears is for all PDFs, not for all files that are neither image, sound, nor video files. Some of the PDFs are in fact images (and thus ineligible for deletion under this criterion), and a curriculum vitæ can be uploaded as a JPG or PNG (and would then be eligible for deletion under this criterion). --Stefan2 (talk) 18:20, 11 February 2023 (UTC)[reply]
  • For a start whether something is or is not useful to the encyclopedia is inherently rather subjective, and speedy deletion criteria are supposed to be as objective as possible. The criterion isn't actually used that much at all, as you can see from the query linked above. What's more a lot of the times it is used are actually abuses of the criterion, only 55 of the last 100 F10 deletions are actually non-media files, the rest are images which by definition don't qualify for F10. (A document uploaded as an image is an image file and therefore doesn't qualify). I was one of the people who supported introducing it back in 2008, at the time Wikipedia had enormous numbers of PDFs with no encyclopedic purpose which have largely been deleted now. Hut 8.5 18:46, 11 February 2023 (UTC)[reply]
  • Beyond what Hut 8.5 said above, I will note what it says in the CSD criterion itself: "Most non-media file formats cannot be uploaded to English Wikipedia, pdf files being the only exception."HouseBlastertalk 21:45, 11 February 2023 (UTC)[reply]
The discussion above is closed. Please do not modify it. Subsequent comments should be made on the appropriate discussion page. No further edits should be made to this discussion.

G10 Requirement that a page be unsourced

There have been a number of times where I have observed a page nominated for G10 which meets every criteria for being a G10 (attack page) except that it is sourced. Sometimes it's improperly sourced (e.g. through an external link rather than reference) and I don't think I've seen one with a high quality reliable source but a source exists. These pages normally end up deleted - and good riddance I say to them even though I'm reluctant to press the button - despite not technically qualifying under the criteria. I'm wondering if we want to soften the wording of that requirement given this practice. Best, Barkeep49 (talk) 17:16, 11 January 2023 (UTC)[reply]

I've always assumed that the unsourced bit only applies to the "biographical material about a living person that is entirely negative in tone" part, so it's fine to delete something under G10 if it's sourced if it meets any other part of G10 (e.g. "material intended purely to harass or intimidate a person"). It is entirely possible to have a legitimate encyclopedia article about a living person which is entirely negative in tone, e.g. most articles about criminals. Removing the sourcing requirement entirely would allow them to be deleted under G10. Hut 8.5 17:52, 11 January 2023 (UTC)[reply]
I agree with @Hut 8.5. Look at the grammar and punctuation of the sentence.

Examples of "attack pages" may include libel, legal threats, material intended purely to harass or intimidate a person or biographical material about a living person that is entirely negative in tone and unsourced. (emphasis added)

I think this could be clarified thusly:

Examples of "attack pages" may include: libel, legal threats, material intended purely to harass or intimidate a person, or biographical material about a living person that is entirely negative in tone and unsourced. (addition of a colon and an oxford comma.)

Either way though, I think Hut is right that the unsourced part only applies to the last item in the list. ~ ONUnicorn(Talk|Contribs)problem solving 18:12, 11 January 2023 (UTC)[reply]
I agree with Hut8.5 and ONUnicorn and would support adding that punctuation for clarity. Thryduulf (talk) 19:07, 11 January 2023 (UTC)[reply]
This matches my understanding and I support the proposed punctuation change. Jclemens (talk) 19:41, 11 January 2023 (UTC)[reply]
If a new article reads "John Snotblot is a fucking asshole", and this sentence ends in 16 footnotes leading to online sources that state unanimously that he's a fucking asshole, I'm going to submit the article for G10 deletion, even though it's clear that he's widely noted for being a fucking asshole, meeting WP:N without question.
However, if it reads "John Snotblot, the president of Marvinia since January 2, 2023, is a fucking asshole", then I'm going to change it to read "John Snotblot has been the president of Marvinia since January 2, 2023", possibly keeping one of the existing sources that corrobate his presidency and the date his term began even if it does also call him a fucking asshole, or else replacing all the original sources with one neutral source. Then I'm going to leave an "only" warning on the creator's talk page and call it a day, leaving it to others to produce a respectable "Criticism" section. Largoplazo (talk) 23:07, 11 January 2023 (UTC)[reply]
Even if your example was not libellous and not intended to purely harass or intimidate the subject there is no need for G10 as it meets criteria A1 and A7. Thryduulf (talk) 23:36, 11 January 2023 (UTC)[reply]
Well, the president of Marvinia bit obviously disqualifies it from both. I think Largoplazo was referring to how they read G10, and I think their assessment is purely right. Nobody is suggesting we scrap G10. casualdejekyll 02:55, 12 January 2023 (UTC)[reply]
My reading is in line with that of ONUnicorn et al: the 'unsourced' only applies to the 'biographical material that is entirely negative in tone' bit. I'd add that this part of the text is listing examples that may qualify - it's not an exhaustive list, and it doesn't define the criterion. Any page, regardless of sourcing, that disparages, threatens, intimidates, or harasses their subject or some other entity, and serves no other purpose, can be deleted under G10. Having said all that, I would have no problem with a tweak in the wording to avoid any possible confusion on this point. Girth Summit (blether) 18:57, 12 January 2023 (UTC)[reply]
Hut8.5 is correct. If I remember correctly, the "negative unsourced BLP" part of G10 was introduced after Wikipedia:Requests for arbitration/Badlydrawnjeff, principle 4 ("summary deletion of BLPs"). – filelakeshoe (t / c) 🐱 19:57, 13 January 2023 (UTC)[reply]

Portal CSDs

Maxim, casualdejekyll: moving the conversation about CSDP to its own section
Yikes. This coming Monday, it will have been a full year since P2 was last used. Besides those five deletions on January 16, it has been over two years since P2 was used.

P1 was used three times in 2022, not at all in 2021, and once in 2020. Let's take a look at those four deletions in the past three years. Portal:Ethiopia was deleted twice, once last August and once last October. The first time looks either like G4 (which does not need P1 to be applicable) or a misuse of CSD. The second time appears to have been a blank page, which would also be G2 as noted in the log entry (no idea why G13 was also listed). Portal:Content/Kenneth Nwanze was deleted under P1/G11, which again does not need P1 to be applicable. Finally, the one in 2020 appears to have been a blanked G11-eligible portal, which was also tagged with P1/A3 and P1/A7 for good measure. In sum, it appears that it has not been needed since before COVID times. Would someone with admin goggles be able to confirm?

For reference, A5 was recently depreciated for disuse when it was used 8 times in the past year, which is the exact number of times P1 and P2 combined have been used in the past year. I fully support deprecating both P1 and P2. MfD can handle an additional 8 cases over three years. I do not see the need to make CSDAs apply to pages in mainspace or portalspace, given that P1 was likely it was necessary zero times in the past three years. HouseBlastertalk 03:47, 12 January 2023 (UTC)[reply]

Was looking at making this proposal as well earlier today. Here you have queries for all deletions under both criteria in portal space: P1 P2. Looking at all somewhat recent deletions I would either categorize them as either something you could have deleted anyway using a G criteria or a single sentence (examples include "Miami is a city in the United States.", "Someone please create a proper portal for this page. This is a country over a 100 million people, it needs a portal.") which I definitely can see why you would want to be able to unambiguously be able to delete, but given the frequency this is needed I feel it's reasonable to remove both criteria. --Trialpears (talk) 06:36, 12 January 2023 (UTC)[reply]
If we want to remove the portal criteria (and I support that given the lack of use) but still allow deletion of those examples, we could just amend A1 and A3 to say that these criteria also apply to portals. However given the frequency that is probably more hassle that its worth and sending them to MfD will not be any sort of burden. Thryduulf (talk) 09:52, 12 January 2023 (UTC)[reply]
I support deprecating both criteria. (Granted, any attempt to officially acknowledge that the entire Portal space is near-useless in its current state is unlikely to gain any sort of community consensus, so I'm just going to quietly grumble about it.) casualdejekyll 14:29, 12 January 2023 (UTC)[reply]
"Fixing Portals" is outside the scope of this discussion and page but I might take it to VP/Ideas at some point. casualdejekyll 14:48, 12 January 2023 (UTC)[reply]
If you do, be careful how you word things and avoid suggesting anything that could be interpreted as either mass (especially automatic) creation or mass deletion. I recommend familiarising yourself with at least a summary of the dispute if you aren't already - it really wasn't pretty. Thryduulf (talk) 15:29, 12 January 2023 (UTC)[reply]
  • Question is G8 not applicable in some cases for portals on non-notable (etc.) subjects? For example if someone creates articles and a portal for their non-notable band and albums, and the articles are deleted as A7/A9, is the portal not dependent on the deleted articles? I honestly never understood the use case for P1. – filelakeshoe (t / c) 🐱 09:43, 13 January 2023 (UTC)[reply]
    I think the idea of P1 was that if people create their garage band/company advertisement article in portal space instead of article space, it should still be eligible for speedy deletion. Think Portal:Music/My garage band. (Back in the day, portals commonly had some original content not just derived from other articles). —Kusma (talk) 09:54, 13 January 2023 (UTC)[reply]

RfC: Should P1 and P2 be repealed as CSDs?

Should P1 (all article CSDs) and P2 (<3 links in a portal) be repealed as CSDs? HouseBlastertalk 16:49, 12 January 2023 (UTC)[reply]

  • Option 1: Repeal P1 only
  • Option 2: Repeal P2 only
    • Option 2.5: Repeal both P1 and P2, but make article CSDs inherently apply to both mainspace and portalspace without the need for P1 (added 15 January)
  • Option 3: Repeal both P1 and P2
  • Option 4: Status quo

(see also discussion section above)

Survey (repeal P1, P2)

WP:BITE and overruse of U5 (WP:NOTWEBHOST)

U5 is a very common criteria for newly created pages. While there are some pages that should certainly be deleted under it, like advertisements, fake articles (excluding drafts), and resources created for projects unrelated to Wikipedia, a lot of the pages that get deleted are just short biographies, many of which would be fine if the user was a contributor.

It's perfectly plausible that someone could create an account with the intent to contribute, and then create their user page before anything else. Immediately rushing to tag the page just because the user hasn't used the account to edit yet is quite WP:BITE-ey.

I think pages should only be tagged right away if the user is blatantly WP:NOTHERE. Otherwise, they should at least get a warning first. Zerbu Talk 13:41, 13 January 2023 (UTC)[reply]

​I wholly agree. A number of clear-cut NOTHERE user pages are often eligible for G3/G10/G11/G12, and I prefer to give benefit of the doubt unless there's a clear pattern to their contributions suggesting otherwise. Short biographies as a first edit are not harmful, and a number of established users probably made some of their first edits to their user page. However, I observe that not all users share these sentiments, and discretion varies among admins patrolling pages tagged for U5; this would run afoul of CSD being objective and uncontestable.
Moreover, per WP:UP#DELETE: A user's contributions that consist solely of a lone edit to their user page should not normally be speedy deleted unless it consists solely of spam or other speedy deletable material. Yet it appears that a number of pages tagged for U5 and subsequently deleted are exactly this.
I believe it may be a good idea to spell out what is and isn't eligible for deletion under U5, similar to how A7 is only appropriate for articles in a few narrowly defined categories, and accordingly explicitly codify in U5 what is written in WP:NOTWEBHOST and other egregious violations of WP:UPNOT as appropriate. For instance, this could include CVs, lengthy autobiographies, blog- or social media-like pages, "fantasy" content that would meet G3/A11 if in mainspace (such as fantasy football, original fanfiction), and pages intended for socializing or recreation unrelated to Wikipedia, along with (trivially) inappropriate pages that meet another CSD.
Also noting that this was (briefly) raised at WT:New pages patrol#Reminder (courtesy ping Liz), though no extensive discussion followed. Complex/Rational 18:51, 13 January 2023 (UTC)[reply]
According to that rather unhelpful discussion, the people who patrol new user pages and tag them for U5 are not New Page patrollers. So who are they? —Kusma (talk) 19:06, 13 January 2023 (UTC)[reply]
Me, I tag a lot of pages with U5. There's probably at least a good 10 or so other users that I see regularly do so as well during my editing window (between 7am and 4pm EST). It really depends on which admin reviews a borderline U5 tag, some pages will be deleted and some won't. I try to always give users the benefit of the doubt, and I've started to err more towards not tagging pages that I know some admins would delete because I want to give people more of a chance on those borderline cases. As such, I think it could be beneficial to more thoroughly define what is and isn't eligible under U5 because there's different interpretations of what's currently written. It's not like users can review what was deleted from previous tags to understand situations when tagging is appropriate and when it's not. Hey man im josh (talk) 19:31, 13 January 2023 (UTC)[reply]
I've been thinking about this some more, and thought: perhaps, as an alternative to speedy deletion, such user pages could be blanked and replaced with {{userpage blanked}} (currently used mainly for drafts) or a similar template, which is less hostile than straight-up deletion. Zerbu Talk 23:00, 13 January 2023 (UTC)[reply]
I am inclined to think that U5 is the most overused speedy deletion criterium, and certainly the most subjective. In particular, I often see draft-like pages like User:Larsro/sandbox or short userpages being tagged as U5. Jo-Jo Eumerus (talk) 09:18, 11 February 2023 (UTC)[reply]

G4 and salted articles

A recent AfD, Wikipedia:Articles for deletion/Gregory Marchand, appears to raise the following issue: the same article was previously deleted and salted under different variations of the subject's name, but may not be close-enough to the deleted version for G4 to apply. G5 could potentially apply, as the previous versions also had sockpuppet issues, but the 2020 creator of the new version is as of now unblocked. Should there be a clause of G4 that applies to creation of the same topic as a deleted and delete-protected article under a different title, even when the content may differ? Or is that too much of a stretch for CSD? (Marchand appears to be headed for sure deletion regardless, but it would be helpful to be able to shortcut such discussions without having to use excuses like WP:SNOW.) —David Eppstein (talk) 08:30, 16 January 2023 (UTC)[reply]

I have observed, and even noted it once at DRV, that G4 is in practice given wider and stronger applicability when the topic’s earlier title is WP:SALTED. That is, if someone sidesteps the salting by recreating under a variant title, admins do G4 and WP:SALT the variant title, and do not get criticised for it at DRV. I think this is good practice, even if undocumented.
SALTing should be read as a prohibition on the creation of the same topic under any title. Challenges to the SALTing should be made at WP:RfPP or WP:AfC. WP:THREE is very good advice on how to make a good case. Challenging a SALTing at DRV should be reserved for complaints of denial of unsalting at RfPP. SmokeyJoe (talk) 08:47, 16 January 2023 (UTC)[reply]
G4 is only for substantially similar pages. If there is a case where content may differ then it is not eligible for G4, regardless of the location of the page, because in theory there could be new information or writing that overcame the original deletion rationale (e.g. GNG is now met). Primefac (talk) 09:41, 16 January 2023 (UTC)[reply]
To nitpick, the GNG can become met without any changes to the page, because the GNG refers to sources that exist, not sources used on the page.
For this sort of reason, I think G4s should readily slip into draftifications. If the page looks unimproved, put it in draft for the new source to be added. SmokeyJoe (talk) 03:00, 27 January 2023 (UTC)[reply]
G4 already applies to the page "having any title", so people circumventing SALT by using different titles for the same content is already covered. But, as Primefac notes, like all speedy criteria, there needs to be an objective standard. If the content is no longer fundamentally the same, it shouldn't be a single admin's decision whether the original reasons for deletion still apply (especially when the page contains information that was not available in the first XFD). Regards SoWhy 10:32, 16 January 2023 (UTC)[reply]
I support more liberal use of G4 in mainspace in the six months following an AfD consensus to delete. This would include a re-creation using different but similarly failing sources to those that didn’t satisfy the participants in the AfD. New sources, new since the AfD, not merely different sources, would amount to a sufficient difference to make G4 proper. If a proponent for re-creation of a recently deleted article disagrees, the answer is AfC, not a title variation, even if it includes a different bad source.
How long should a consensus at AfD to delete be respected, prohibiting bold re-recreation without overcoming the reasons for deletion? I think the answer is six months.
Some admins err in SALTing for too long. I think a SALTing over six months should be dependent on an AfD consensus to SALT. SmokeyJoe (talk) 11:09, 16 January 2023 (UTC)[reply]
I understand where you are coming from and I don't think G4 is misapplied if the changes were only made to try and game the system, e.g. by replacing one bad source with another bad source (call it IAR or applying the spirit of the policy). But that requires a blatant attempt to circumvent G4. If the reviewing admin has to start judging the merits of the new article, the applicability for G4 ends. And I would also oppose any kind of recreation ban without going through AFC, considering that AFC is not a requirement for new articles and AFC is completely overwhelmed already anyway. However, that is not a question for CSD but instead would require a change to the deletion policy (after a widely advertised RFC). Regards SoWhy 11:50, 16 January 2023 (UTC)[reply]
I'll take a more conservative tack than SoWhy: Substantially altering the article in any way, such as the addition of a reference that had been mentioned in the AfD discussion but not put into the article, makes G4 inapplicable. There is no time limit on G4, so theoretically if such a change were made 15 minutes after the discussion were closed, G4 would remain inapplicable. That's what G4 says.
Now, would the act of doing so be disruptive editing? Possibly, but that's a judgment call requiring a determination of user conduct. Since it's not pure vandalism, one admin acting unilaterally would be inappropriate. In such cases, the user conduct becomes the discussion: the article content, having been sufficiently changed to avoid G4, might well be deleted again later. Of course, if the user can demonstrate that the addition of content solved the problem that led to the deletion, then it's obviously no longer a conduct issue: we don't punish editors who improve the encyclopedia simply for editing with timing that looks bad. Jclemens (talk) 08:39, 17 January 2023 (UTC)[reply]
To expound a bit more on why SmokeyJoe's take on G4 is wrong, let us look no farther than the text itself: It excludes pages that are not substantially identical to the deleted version, and pages to which the reason for the deletion no longer applies. Since the sentence contains two clauses it is clear that there can by copies that are substantially identical to the deleted version but to which the reason for the deletion clearly still applies, as well as pages to which the reason for the deletion no longer applies. that are substantially identical to the deleted version. Now, I think the latter may be a null set (except, perhaps, if policy changed to make previously unacceptable content acceptable), but CSD policy is written to ensure G4'able content must be both substantially identical and the reason for the deletion still applies. Jclemens (talk) 08:47, 17 January 2023 (UTC)[reply]
I agree completely with Jclemens - both parts of G4 must be met for it to be applicable. I can also think of another example of where the page is substantially identical but the reason for deletion no longer applies - where the deletion reason was based on a premise which has since been shown to be incorrect. In every such case I can think of, and indeed every case I can think of where Jclemens' scenario would apply, going via the deleting admin and/or DRV would be better than just unilaterally recreating the article, but I'm not going to say doing so would never be correct. Thryduulf (talk) 15:56, 17 January 2023 (UTC)[reply]
I believe there may be space between policy as written and policy in practice. G4s happen where the admin thinks it would be a SNOW deletion per the recent AfD. The criteria for SALTing is lower than for G4. Also, in NPReview, a re-creation after a recent AfD is likely to be WP:DRAFTIFY-ed (fitting criterion 2a-ii). SmokeyJoe (talk) 00:17, 18 January 2023 (UTC)[reply]
I believe there may be space between policy as written and policy in practice. When it comes to CSD there isn't - the consensus going back years that
  • The criteria for speedy deletion (CSD) specify the only cases in which administrators have broad consensus to bypass deletion discussion, at their discretion, and immediately delete Wikipedia pages or media. and
  • Administrators should take care not to speedily delete pages or media except in the most obvious cases.
combined with the nature of speedy deletion and the deletion policy mean that pages must meet the letter and spirit of the criterion to be eligible. Thryduulf (talk) 01:37, 18 January 2023 (UTC)[reply]
Ideally, yes. Sometimes, stuff happens without “broad consensus”, despite the care that should be taken.
My reading of DRV and WT:CSD over the years is that liberal G4 deletions are negligible compared to liberal G6 deletions. SmokeyJoe (talk) 02:59, 18 January 2023 (UTC)[reply]
G6 is indeed by far the most misused speedy deletion criterion (which is why I've been proposing reform of it for years), but misuse is still misuse and should never be endorsed however frequently it does or does not happen. Basically if a page does not meet both the letter and spirit of a CSD criterion then it cannot be correctly speedily deleted. If there are pages you think should be deleted but cannot be then propose a new or modified criterion and get that consensus - c.f. WP:IARUNCOMMON. Thryduulf (talk) 10:17, 18 January 2023 (UTC)[reply]
  • In order for G4 deletions to meets its letter, not just it’s spirit, I propose that it’s wording be modified. There should be mention of thresholds. The soft words are “sufficiently” and “substantially”. These should be connected to sources. Things that affect the threshold of applicability of sufficiently/substantially include: the age of the AfD (six months default threshold); and whether the title has been SALTed. At the risk of bloat, consider advice to use REFUND to draftspace and improve the draft to make the case that the reasons for deletion have been overcome. —SmokeyJoe (talk) 00:54, 27 January 2023 (UTC)[reply]

New General Criteria for blatantly fictitious?

Hey, this talk page gets a lot of activity and proposals and questions about the nature of different speedy deletion criteria and when they should be applied but I have a new criteria to propose that perhaps hasn't been suggested yet.

I spend a lot of time with expiring drafts and also reviewing pages tagged for CSD speedy deletion and I think there should be a general criteria, to apply to both Draft and main space, for articles that are just blatantly fictitious. These are not hoaxes, where someone tries to pretend something is true when it isn't, it is kind of similiar to A11, for things that are made up or invented but what I keep seeing are what I view as "creative writing projects". They are stories new editors have written, sometimes fan fiction, sometimes fairy tales, sometimes ghost stories, it comes in many different forms. But it is all clearly fiction, not intended to be take for factual information. I see AFC reviewers offer opinions on submissions that are like this, saying things like "not enough reliable sources" but if a new editor makes up a story about a summer camp or a mermaid, there will never be a reliable source for that content. I've seen poems, song lyrics, all of this content is more suited to a personal blog than an encyclopedia. So, for Draft space, they sit for six months and are then deleted because, honestly, these editors never return to work on this content, it's just something they wanted to share or express one day. If it's in User space than CSD U5 can fit this content (see WP:UPNOT) but sometimes this material is in other namespaces. Maybe there could be a CSD D5 for webhost content in Draft space.

A different but related criteria could be called "stories about my life". So, new editors tell you about their best friend in school or how they went to this great concert when they were 18 or who their favorite actors are and why. These can be factual but they are clearly not about sharing knowledge and are not encyclopedic. I've seen content like this sometimes tagged as a G2 test edit but there are really no criteria to cover "personal writing projects". There is discussion on deletion pages about autobiographies and while we see a LOT of autobiographies in Draft space, that's not what these are. Both instances are more like essays or personal sharing but they are not hoaxes, they are not incoherent gibberish, not really vandalism or original research either, they are simply writings that are out-of-place in an encyclopedia. Aside from waiting out the clock for six months, it would be nice to have a critera that fit this content and an informative template explaining why this kind of content is not acceptable in this collaborative project. Thoughts? Liz Read! Talk! 20:44, 20 January 2023 (UTC)[reply]

I know what you mean. I deleted one of those the other day (Draft:“The Waffle House has found its new host” Lore) under G3 - it was tagged as such, but it takes a great deal of forced pragmatic incompetence to see that as a "hoax" rather than just bad creative writing. Incidentally when I was a new editor here that probably would have been tagged as G1.
If not a new speedy criteria (I think modelling D1 as similar to U5 is probably the best option) maybe a PROD type thing for those drafts would be another option? For things which are clearly not meant to be encyclopedia articles and never will be. – filelakeshoe (t / c) 🐱 20:54, 20 January 2023 (UTC)[reply]
I've seen a fair amount of these, and while G3 catches some of them, that is more intended for false information masquerading as legitimate articles (especially those created in bad faith). Deleting them is certainly reasonable, and keeping never-to-be articles as "AfC drafts" seems unnecessary at best, but I'm reluctant to delete when no CSD clearly applies. As such, I believe that such a new criterion generalizing A11/U5 might work – to cover original content clearly not intended to be an encyclopedia article and that would probably qualify for U5 in userspace: things made up one day, fanfiction and original stories, fantasy sports results, fantasy maps/transportation/proposals, personal essays unrelated to Wikipedia.
However, I don't know if there are several discrete categories that would make such a criterion objective and uncontestable, and I also am under the impression that not all admins interpret U5 and G11 (and even sometimes A7, which such a new criterion would resemble most) the same. Very careful wording is required. Complex/Rational 22:26, 20 January 2023 (UTC)[reply]
It’s an old idea to push for A7 to be expanded to to cover draftspace, but it doesn’t make the justification. There is no harm in leaving weak claims lying in draft, and sometimes they are a draft for something with a claim.
The case for A11 is stronger, but, for compelling cases, like “stories about my life”, fit better to “unsourced BLP”. If it names a probably living person, it’s a BLP. No draft can be expanded on a BLP in the absence of a source, so it shouldn’t be a draft.
The answer is to expand BLPPROD to draftspace. Although not a proper CSD, PROD in unwatched places like draftspace is a pseudo CSD, and so the WP:NEWCSD tests are appropriate. BLPPROD meets these on all NEWCSD criteria but “frequent”. To provide evidence for “frequent”, please nominate all unsourced BLPs at MfD. If they are not frequent, this is still a good use of MfD.
I don’t think there are many A11 equivalents in draftspace that don’t meet any of WP:G1, WP:G3 or BLPPROD. SmokeyJoe (talk) 07:47, 21 January 2023 (UTC)[reply]
My only concern with an expansion of BLPPROD is that one can link, say, an Instagram page or TikTok video supporting a "story about their life", such that there technically is a source present, but the draft would be no more worthy of publication than without such a link. Complex/Rational 14:56, 21 January 2023 (UTC)[reply]
User:ComplexRational, it's not clear whether you understand that BLPPROD is applied to an unsourced page and then requires a reliable source before it can be removed. An Instagram page or TikTok video is not a reliable source.
If the original page had the Instagram link or TikTok video, then the author is making some effort, and they deserve a conversation about WP:RS, and I OPPOSE speedy deletion of the draft under an expanded A11. SmokeyJoe (talk) 05:38, 22 January 2023 (UTC)[reply]
I'm all for discussing WP:RS with draft creators. I understand how BLPPROD is used on articles; are you merely suggesting to expand BLPPROD and offer a more detailed explanation, even for cases that nobody (besides perhaps the creator) would agree could be a plausible article? And while sometimes social media citations have their place, they would clearly not be sufficient for such drafts.
Per below, do you think a case such as "Jack has a crush on Jill. She is the most popular girl in school[citation to Jill's social media page]." needs to go through BLPPROD as opposed to being speedied? I think it could also be helpful to survey recent deletions to gauge how frequently a case like this might occur. Complex/Rational 15:12, 22 January 2023 (UTC)[reply]
If the citation to a social media page was present before, then according to the BLPPROD method, BLPPROD can’t be used. Use XfD. A newcomer already with the concept that claims need sources is worth the time to explain things.
In draftspace, BLPPROD will be a pseudo-CSD. “Speedy” doesn’t mean “instant” or even “fast”. I don’t think there is appetite for an instant speedy NEWCSD for draftspace.
To survey frequency, search the MfD archives for “unsourced BLP”. It’s not quite frequent enough. SmokeyJoe (talk) 12:14, 23 January 2023 (UTC)[reply]
There's a lot of schoolboyish stuff in draftspace that is only borderline G3 or G10 and should still be deleted as soon as possible. From "Jack has a crush on Jill" to "Joe Schmoe is in the eighth grade and jerks off to pictures of Taylor Swift". To delete under G10 works if we assume it is there in order to bully, but sometimes the content could be autobiographical and not intended as attack. I always delete this kind of things (and they should not go through MfD or anything else that increases awareness), but I am unsure what criteria to use. Something more clearly defined to catch these ("stories about non-notable people connected to the page author"? "possible violations of privacy"?) could be helpful. —Kusma (talk) 15:37, 21 January 2023 (UTC)[reply]
@Kusma: As another example, I recently deleted Draft:Paulinka as G11 (it was tagged as G3), which was essentially a love letter, but that felt like a stretch of the criterion. I imagine this is not too uncommon and needs a clearer criterion. Complex/Rational 15:46, 21 January 2023 (UTC)[reply]
Schoolboyish stuff, borderline G3 or G10: "Jack has ..."; "Joe Schmoe is..."; autobiographical; .... This sounds VERY applicable to "unsourced BLP".
If you are not sure what criteria to use, you SHOULD NOT be speedy deleting. MfD should be used to establish a record and precedent for SNOW deletion criteria for new or expanded CSDs. You should not misuse criteria to avoid the NEWCSD process.
"stories about non-notable people connected to the page author"? "stories about ... people ..." squarely fits "unsourced BLP" in every case that I have seen that is suitable for a NEWCSD, for which BLPPROD will suffice.
Please send borderline cases to MfD. If there is a tad personal information that must be included in the MfD, the MfD can be blanked later. Pointing to a borderline G10 page from MfD does not induce a burst of pageviews beyond what is expected from the number of MfD regulars. SmokeyJoe (talk) 05:47, 22 January 2023 (UTC)[reply]
I too understand what you mean. I recently deleted Draft:Kaylin Beifong under G2, although I realise it's a stretch. That draft could probably have been deleted as a hoax, since the Beifongs are a fictional family in The Legend of Korra, and it's impossible that the user is a member of that family, but still G3 feels incorrect as well. Salvio giuliano 16:27, 21 January 2023 (UTC)[reply]
  • In all of the above, nobody has explained why leaving those pages that are not BLP violations alone until G13 claims them is a problem? Why do they need deleting rather than ignoring? If there is an actual problem and an existing CSD criterion doesn't clearly apply then nominate it at MfD (remember that if it's unclear whether a page meets a speedy deletion criterion then it doesn't). Thryduulf (talk) 21:21, 23 January 2023 (UTC)[reply]
    For the same reason that the fake game show pages that were rampant for years are a problem. It's actively abusing Wikipedia for purposes entirely unrelated to building an encyclopedia, and allowing them to go through the standard G13 process wastes the time of both reviewers and people who check for salvageable drafts. In the case of the fake game show pages it was mostly happening in user space, so U5 basically resolved it. These sorts of pages are always deleted at MfD, they're not salvageable through normal editing, giving them wider attention is actively detrimental since it begets more creations along the same lines (as the fake game show problem demonstrated, a lot of them were time sensitive and going through MfD was about as effective a deterrent as watching a horse run from a barn before locking the door), and giving them a free 6 month pass gives the misguided impression that such pages are at least tolerated. That said, I'm also not sure how to capture that in a general criterion; if I could get the wording down I would, and if anyone has a starting point I'm happy to help refine it. The Blade of the Northern Lights (話して下さい) 23:04, 23 January 2023 (UTC)[reply]
    Make as many MfD nominations as you can using the proposed new CSD criterion. For U5, it was “NOTWEBHOST violation by a non-contributor”. G13 relied on the count of tens of thousands of abandoned drafts, too many to review, and that they included copyrights and BLP violations.
    The next CSD, a pseudo CSD, BLPPROD in draftspace, can be demonstrated by many identical MfD nominations “unsourced BLP”. It is appropriate to allow any interested editor one week to add a reliable source. “Unsourced BLP” nicely covers a multitude of sins, including borderline G10 and fake information associated with a person.
    I agree with Thryduulf that there what’s been no case made that there are ongoing draftspace problems in large numbers that are not BLP violations. SmokeyJoe (talk) 02:08, 24 January 2023 (UTC)[reply]
    So, unsource BLP requires that there are absolutely zero references and external links. But if an editor makes a page about a day in their life and includes references of the places they went to, that no longer qualifies as unsourced BLP. - UtherSRG (talk) 11:46, 24 January 2023 (UTC)[reply]
    Feel free to read WP:BLPPROD. All the same principles apply. I see no reason to alter any part of it.
    New editors attempting a sourced draft are quite a step above drive-by dumpers of BLP violation. SmokeyJoe (talk) 20:58, 24 January 2023 (UTC)[reply]
    If something has a 5% chance of being a BLP violation and a 0% chance of becoming an encyclopaedia article, immediate deletion is the way to go. Draft space is not a free webhost where you can post love letters, complaints about your math homework or random graffiti. —Kusma (talk) 12:15, 24 January 2023 (UTC)[reply]
    CSDs require a match to the CSD, not your probability analysis. If in doubt, use XfD. SmokeyJoe (talk) 20:59, 24 January 2023 (UTC)[reply]
    I use XfD when I have doubts about whether a page should be deleted. —Kusma (talk) 21:57, 24 January 2023 (UTC)[reply]
    CSD is not about whether you think a page should be deleted, it requires an exact match to at least one of the criteria. The criteria list the only circumstances in which the consensus of the community is that the page should be speedily deleted. Any other page should go to XfD. Thryduulf (talk) 22:31, 24 January 2023 (UTC)[reply]
    As several admins above have stated, there are instances of pages that should be deleted and should perhaps not go to XfD, yet there is opposition here to fixing the CSD to align policy with practice with arguments like that these pages do not appear often at XfD. WP:BLPDELETE allows many such deletions even if not covered directly by WP:G10 (and of course, there is always that other policy). —Kusma (talk) 23:02, 24 January 2023 (UTC)[reply]
    If something is too sensitive for XfD, go to WP:Oversight.
    BLPDELETE is an oddity that should be fixed. Maybe CSD#10a. It does not fit nicely within G10. BLPDELETE is a very poor example of speedy deletion authorisation from a non deletion policy page. It’s use under a best-fit to G10 sets this really poor precedent that speedy deletion by best fit is ok. SmokeyJoe (talk) 07:02, 25 January 2023 (UTC)[reply]
    As repeatedly explained on this page and elsewhere, IAR is never a suitable justification for speedy deletion because IAR is only for actions that uncontroversially improve the project, deleting something out of process (e.g. speedily deleting something that is not covered by a speedy deletion criteria) is always controversial and does not improve the project.
    Regarding things not going to XfD - if something is too problematic for XfD then request WP:OVERSIGHT. If something doesn't require oversighting and also doesn't meet G10 then its not problematic enough that XfD is a problem.
    WP:NEWCSD rightly requires that new criteria be both frequent and uncontestable, the way to demonstrate this is lots of nominations at XfD which always end in deletion. Thryduulf (talk) 10:28, 25 January 2023 (UTC)[reply]
    IAR says "If a rule prevents you from improving or maintaining Wikipedia, ignore it." It does not say anything about "uncontroversial". An objection against a specific IAR delete (which I don't actually perform very often, but I do see a need for them) must explain why that specific deletion was bad, not just say "all out-of-process deletions are controversial, hence bad". As you see from this discussion, there are many types of pages in draft space that are frequently speedily deleted by admins with a slightly bad conscience, and it would make sense to either align written policy with this practice or to criticise admins for specific incorrect deletions. —Kusma (talk) 10:47, 25 January 2023 (UTC)[reply]
    Kusma: you are saying that the deletion policies need to be rewritten to fit with out-of-process deletions that admins are currently making. Smokey and Thryduulf are saying that admins need to change their workflow so that their deletions fit with the deletion policies. — Martin (MSGJ · talk) 12:32, 25 January 2023 (UTC)[reply]
    Indeed as I wrote in WP:IARUNCOMMON, if you find yourself needing (or even wanting) to ignore a rule frequently either you or the rule is wrong and needs to change. I've seen nothing in the evidence presented that convinces me the rule is wrong. Thryduulf (talk) 01:19, 26 January 2023 (UTC)[reply]

I left a note about this discussion on WP:VPP#Discussion at WT:CSD, I figure some outside input could be helpful in fleshing this out. The Blade of the Northern Lights (話して下さい) 14:00, 26 January 2023 (UTC)[reply]

Young students put all sorts of things in draft space naming real people, classmates and teachers. Other youngsters have written autobiographies complete with their addresses. I recently CSDed a draft that appeared to be a student writing silly things about a real teacher, nominated with G10 for lack of a specific G template, but it was deleted as G3. I did realize later I could have used {{db}} with a reason (disregarding the instructions that the reason has to be one of the existing ones). Years ago I used to nominate these as A7 with comments like "too much information about a minor child" or "children writing silly things about their classmates", but that no longer works in draft space. It would be very useful to have a G template for all the stuff that just does not belong on Wikipedia. StarryGrandma (talk) 22:57, 26 January 2023 (UTC)[reply]
I think everything User:StarryGrandma suggests that could remotely suit a NEWCSD, but does not meet G10 or Oversight, will fit better under BLPPROD-in-draftspace. SmokeyJoe (talk) 01:01, 27 January 2023 (UTC)[reply]
In CSD-Draft-version-of-A11/G3, anyone can tag the page and admins can go through, check if it conforms with policy to delete blatantly false or fictional material, and delete if so. Whereas BLPPROD if I understand it correctly will only apply to BLP situations and will let anyone delete old drafts after the waiting period. Andre🚐 01:18, 27 January 2023 (UTC)[reply]
BLPPROD only applies to BLPs, but quite broadly, not just to the subject of the page must be the BLP, any person tangentially connected can make it subject to WP:BLP.
It allows anyone to tag, with a sticky tag. The page author or anyone else has seven days to add one reliable source, or there’ll be a deletionist admin who quickly clears the expired BLPPROD category. The deletion criteria on an expired BLPPROD is so simple that the process is efficient. The applicability of G3 to a short draft can be very subjective.
A11, I support as extending to draftspace, but is it overkill? If it’s not BLP, why care? No reader is going to be unexpectedly misinformed by a draft page, as the namespace in the title, in huge font is kind of clear and meaningful.
I think one week is an appropriate time for a newcomer to be given to add a source to “Mrs Smith is a great English teacher”. SmokeyJoe (talk) 03:14, 27 January 2023 (UTC)[reply]
@StarryGrandma if you see pages (written by anybody) naming non-notable individuals, especially individuals who are or may be minors, contact Oversight (the information should not be easily accessible to even admins) rather than trying to bend speedy deletion criteria that don't apply. Ideally you should redact the problematic information (or blank the page if redaction wouldn't leave anything) using a bland edit summary. Thryduulf (talk) 01:58, 27 January 2023 (UTC)[reply]
It's encouraged to temporarily revdel or speedy-delete oversightable material while waiting for it to get oversighted: see Wikipedia:Revision deletion#Hiding oversightable material prior to Oversight. If it needs a speedy rather than a revdel, G10 or G12 may be an appropriate rationale. —David Eppstein (talk) 02:14, 27 January 2023 (UTC)[reply]

G4 tweak.

Currently, the opening paragraphs include the verbiage, " If an editor other than the creator removes a speedy deletion tag in good faith, it should be taken as a sign that the deletion is controversial and another deletion process should be used." I propose that this be modified to read, " If an editor other than the creator removes a speedy deletion tag in good faith, it should be taken as a sign that the deletion is controversial and another deletion process should be used. The one exception to this would be for G4, where non-admins cannot see the original deleted article. In those cases, the tag should remain until reviewed by an admin." This is due to the fact that the current verbiage says that another deletion process should be used. And at that point, since it has already been through AfD, Prod is no longer available, sending it back to AfD does not seem like a viable next step. If a third party removes the tag, and it cannot be re-added, no admin will ever have the chance to compare to see if there is a difference. There does not appear any inherent harm to the project for the tag to remain in place until admin-reviewed. Onel5969 TT me 22:20, 25 January 2023 (UTC)[reply]

Though the person adding the tag can't see the deleted article either. If we're going to change the wording like that for that reason, then logically wouldn't that just require that only admins can use G4 anyways? Since both the person adding and the person removing the tag have no idea what the article is like before and whether G4 applies. So both actions would be making a baseless claim. SilverserenC 22:29, 25 January 2023 (UTC)[reply]
That means admins would have to do the new page patrol duties, as that's where these issues get detected. — rsjaffe 🗣️ 22:34, 25 January 2023 (UTC)[reply]
I think making placement admin-only would move in the wrong direction. Rather, we should add wording that clarifies that G4-placement is procedural and explain the process. Given that G4 isn't always a procedural shot in the dark (e.g. in cases where the nominating editor has full recollection of the previous version, or where there's a parallel version in draft space), it may make sense to create a G4.5 of sorts to allow for different text to be displayed depending on the context. signed, Rosguill talk 22:40, 25 January 2023 (UTC)[reply]
Back before I became an admin I learned of several ways of viewing deleted articles to check against G5. The most reliable loophole has been closed but yes I agree that non-admin can assess against G4 some percentage of the time. What Onel's message caused me to think about was someone removing a G10 / G12 tag but decided that ultimately if that's what they were it could still be caught and speedily deleted when it got to an XfD and was seen by an admin. Best, Barkeep49 (talk) 22:43, 25 January 2023 (UTC)[reply]
The current policy is also confusing to the person who wrote the article/other editors who see the G4 speedy deletion tag. Especially for newer editors, they might remove the tag in good faith based on some memory of the pre-deletion article and be surprised by experience editors biting them saying that the tag is "procedural" and "waiting for an admin to review". As it currently stands, any non-admin editor can remove the G4 tag based on the immediate text of the template and the policy and the article may never get reviewed by an admin. EmeraldRange (talk/contribs) 23:42, 25 January 2023 (UTC)[reply]
G4 is a widely misused criterion, for some of the reasons noted above. We just had another instance come up at DRV. I'm not inclined to add more layers to it. If anything, I would rather see things added to the AfD queue with a "may meet G4" notation, and admins patrolling AfD enact those G4s if applicable. Jclemens (talk) 22:48, 25 January 2023 (UTC)[reply]
From my viewpoint this exception should probably be codified. If a non-admin removes a G4 template, this should not preclude requesting G4 again, as that should not be considered a situation when deletion is controversial and another deletion process should be used because a valid non-admin non-creator remover of the template can not, under most circumstances, be assumed to have good knowledge that the page is not a sufficiently identical copy, and has no solid reason to contradict the suspicion that the criterion applies, so there is no substantive controversy to be had – everyone would just need to wait for an admin to look at the deleted article and make the comparison. Sometimes, as Barkeep49 said, people will in fact have good knowledge, but there's no reason to generalize from that. Other proposals in the ANI discussion were to significantly soften the template, which seems go along with the idea, and a new software feature in PageTriage which would automatically gauge similarity. —Alalch E. 22:55, 25 January 2023 (UTC)[reply]
Though that does raise the requirement that admins actually check and compare the versions properly. Rather than in cases like what Jclemens brought up above. Should there be some sort of consequence for an admin who repeatedly deletes articles under G4 that were not substantially similar (and in many cases, blatantly so)? SilverserenC 23:04, 25 January 2023 (UTC)[reply]
G4 is one of the most complicated criteria we have, and this means that pages are very often mis-tagged (I suspect due to people relying on the Twinkle dialogue instead of reading what the criterion actually says). Quite a lot of my G4 declinations could have been (and were) done as a non-administrator, such as due to the discussion having ended with a "soft delete" or "keep" or "redirect", the previous version having been speedily deleted or prodded, or that the discussion was quite clearly about a different person to the article being tagged. Looking through my G4 declinations, only two actually needed comparison with the previous deleted version (i.e. only two needed deletedtext), and one of them actually only kind of needed it, since it was clear that much had developed since the AfD. I therefore don't think this is necessary. Sdrqaz (talk) 23:33, 25 January 2023 (UTC)[reply]
If significant time has passed between delete and recreation it is also easy to see if new references or new claims of importance are added fro that time interval. So non-admins should be able to detag for that reason. If it is obvious that the AFD would no longer result in a delete because of something that changed in the meantime, there's no need to have another AFD. Graeme Bartlett (talk) 04:53, 26 January 2023 (UTC)[reply]
In addition to what others have said, if the previous page was deleted at XfD then the discussion should give an indication of the reason for deletion and in some cases it is very clear whether the reason for deletion no longer applies (e.g. if an article was unsourced and there are now sources; a redirect has a different target, etc). It would not benefit the project to prohibit non-admins from declining in these circumstances. Also, if someone other than the page creator removes a speedy deletion notice in good faith we should (in the absence of good reasons not to) treat it as controversial and require a discussion before deletion. Thryduulf (talk) 12:04, 26 January 2023 (UTC)[reply]
Additionally, bringing this from Template talk:Db-meta: I want to suggest adding language to the G4 template stating something like
If this article isn't a resubmission of previously deleted content, do not worry! An administrator will review your article.
Marking a page as reviewed and tagging it with a alarming notice that the article can be deleted at any time is non-intuitive to newer editors that this is a procedural tag which, in essence, asks for admin review. I was unsure if the tag meant anyone would swoop in and delete the article despite obvious (to me) reasons why the deletion no longer applies. EmeraldRange (talk/contribs) 12:59, 26 January 2023 (UTC)[reply]
This is wonderfully aspirational. Unfortunately, the fear you note above is neither strictly unwarranted nor irrational. Jclemens (talk) 04:17, 27 January 2023 (UTC)[reply]
  • There have been several occasions where I, as a non-admin, have removed G4s because it is obvious even without seeing the deleted article that the current article is not substantially identical. This has usually been because the current article has content about things that happened after the previous deletion discussion. Phil Bridger (talk) 08:06, 27 January 2023 (UTC)[reply]
  • Oppose There are many ways for a non-admin to determine that the deleted content is not substantially identical, e.g. Internet Archive, prior memory, new sources that did not exist at the time of deletion, etc. -- King of ♥ ♦ ♣ ♠ 00:55, 28 January 2023 (UTC)[reply]

New "G" variety for articles totally consisting of LLM text

See Wikipedia:Village_pump_(policy)#Wikipedia_response_to_chatbot-generated_content for extraordinarily extensive background. "AI"s or LLMs (better term, since those "AI"s can be dumber than dirt), are being used to generate whole Wikipedia articles or portions of articles. When a whole article is produced, there are several issues: 1) the copyright status is unclear in many cases, though some LLM text generators do hold copyright. In any case, it is impossible to go back to the generator and compare to text there, as the output changes each time. 2) LLMs are great at making up stuff and making it sound plausible--it is highly likely that there are falsehoods in the output, and it would require the same level of effort as writing a new article to validate the text. And LLMs are great at producing fake references.

There is a policy being written for proposal soon. In the meantime, we're faced with articles that are somewhere between a G3 hoax and G12 copyright, produce significant verifiability issues, and in any event don't represent original work by the author of the article.

I'm proposing we have a new "G" variety for this issue. I know this is problematic until we have an actual policy, but I believe the above-listed issues are sufficient to have a wholly-LLM-generated article deleted even absent a new policy. — rsjaffe 🗣️ 23:45, 25 January 2023 (UTC)[reply]

Are there enough of these that a CSD criterion is warranted? Elli (talk | contribs) 01:28, 26 January 2023 (UTC)[reply]
(edit conflict) New speedy deletion criteria need to meet all of the four requirements listed at WP:NEWCSD, and while there might be a need for something in this area what we have currently isn't close to meeting three of them (uncontestable seems OK, once the others are fixed) -
  • Objective - there needs to be some reliable and objective way of determining which pages are created by chatbots and which are not, and while I've not been following the discussions in detail it all feels very vague at present. Some of the detection methods seem to involve detailed comparisons with other sources which is wholly unsuitable for speedy deletion.
  • Non-redundant - hoaxes and copyright violations are obviously already speedy deletable, are there any that are not copyright violations?
  • Frequent - are there so many of these that normal processes such as AfD cannot cope? Especially when considering only those that are not copyvios and not hoaxes?
Thryduulf (talk) 01:29, 26 January 2023 (UTC)[reply]
In terms of quantity, yesterday I found 7 in about 10 minutes worth of looking. If we can by default claim copyvio then I have no problem using that criterion. There seems to be some uncertainty about the copyright status, however. As to objective, I’m only deleting those with a greater than 99% chance of being machine-generated using one of the detection websites. — rsjaffe 🗣️ 01:46, 26 January 2023 (UTC)[reply]
Wait, are you deleting them under some other criteria they're failing (since the ones I've seen have all failed multiple categories of deletion reasoning) or are you actually using this LLM claim as the deletion reason? Because I don't think there's any way to actually confirm the claim that any of these are LLM written. We can suspect, but if we get an article that actually meets our criteria and uses real sources, I don't consider "We suspect LLM text" to be a good reason at all. SilverserenC 01:53, 26 January 2023 (UTC)[reply]
I’m using a tool by ChatGPT to evaluate the probability that the text is LLM-generated: https://openai-openai-detector.hf.space/. Using a cutoff of 99% probability makes the olds of false detection very low, and all the ones I’ve found so far were above 99.8% probability of being machine-generated. — rsjaffe 🗣️ 02:32, 26 January 2023 (UTC)[reply]
That...didn't answer my question. It really doesn't matter how many 9's you go into the decimals, we can't actually confirm it's LLM text, so our deletions should be based on other deletion requirements. Trying to make a subjective other tool to claim LLM written doesn't meet the CSD four criteria. SilverserenC 02:35, 26 January 2023 (UTC)[reply]
That method is being accepted as an appropriate way to confirm it is LLM text. See Wikipedia:Large_language_models#Identification — rsjaffe 🗣️ 03:13, 26 January 2023 (UTC)[reply]
And if the editor says it's not LLM text and that they wrote it themselves, are we just going to say "No, you didn't" without actual evidence? SilverserenC 03:20, 26 January 2023 (UTC)[reply]
And if it is LLM generated text, then it does not mean it is a copyright violation. Out of 2 that the OP tagged. One was a violation of a CV on a web site, and the other did not appear to copy anything preexisting. I am going to say no to such a speedy delete criterion. As LLM may be useful to us, and the problems are on the same scale as non-English native speakers writing, or topic-clueless people writing. Graeme Bartlett (talk) 04:47, 26 January 2023 (UTC)[reply]
I would generally be okay with a G-criterion to handle such cases if NEWCSD can be satisfied but the discussion is moot at this point because there is no policy in place that actually forbids such content. If and when the community decides that such content should not be submitted, then we can discuss how we handle violations. But we cannot establish a speedy criterion for content that is currently allowed (if it does not violate any other existing policies like copyvio). Regards SoWhy 08:48, 26 January 2023 (UTC)[reply]

I am working on improving WP:LLM so that it can be proposed as a guideline. Right now, what I have kind of pieced together is the idea of nominating them as G3 if they are trash -- I mean, utter trash. Not talking about if someone uses a language model to make a rough draft and fleshes it out here, more like the "I typed 'write a Wikipedia article about blungus' into the machine and blindly pasted the output into the edit box". Most of the examples from the AN/I thread have been deleted, but Draft:Gecko's was probably the best example: it had fake references and everything. I think that if you are going through a draft, AfC submission or new page and you discover that the references are made up, it should be some kind of speedy deletion. Whether this actually fits in G3 or if some kind of purpose-made criterion is necessary, who knows. jp×g 06:14, 26 January 2023 (UTC)[reply]

Sadly, G3 only covers blatant and obvious hoaxes. The kind of faked references found in Draft:Gecko's require work to determine that they are completely falsified; they are hoaxes, but not blatant and obvious ones. —David Eppstein (talk) 07:03, 26 January 2023 (UTC)[reply]
I agree with that per WP:DWHOAX but I think it's within the spirit of G3 if anyone can determine the fakeness of the sources easily (not saying that this was the case here). Regards SoWhy 08:44, 26 January 2023 (UTC)[reply]
What about a criterion for speedily deleting anything whose references are found to be fake? After all, that would be bad if the fake references were created manually, too. XOR'easter (talk) 18:44, 26 January 2023 (UTC)[reply]
That would work best as an extension of G3. —Kusma (talk) 19:38, 26 January 2023 (UTC)[reply]
This seems like a good path forward, we could add fake content and references to G3 under existing policies without waiting for LLM policy to be decided. –dlthewave ☎ 20:23, 26 January 2023 (UTC)[reply]
I nominated a couple as G3 and the admins reviewing declined to delete. — rsjaffe 🗣️ 16:48, 26 January 2023 (UTC)[reply]
I nominated one of those at MfD, Wikipedia:Miscellany for deletion/Draft:Iraqian economic influence in 2023. I agree that this kind of good-looking rubbish should have a G-speedy criterion, if we have a way to be reasonably certain that the page is LLM-generated of course. Fram (talk) 17:28, 26 January 2023 (UTC)[reply]
That is an example of a page that was created in article space and moved to Draft world. I wonder if this is an opportune moment to revise our procedures for draftification. For example, testing positive on one or more LLM detectors doesn't point to a specific source that was copied from, so a copyvio speedy delete would likely get turned down, as happened here. But it sure seems like good enough evidence to warrant immediate draftification, with an extra proviso that it can only be moved back to article space after a thorough check by someone other than the page creator. In other words, it calls for removing criterion 3 (there's no need to wait to see if constructive edits happen), and objections from the page creator should get no weight. XOR'easter (talk) 19:00, 26 January 2023 (UTC)[reply]
If we're worried that something is a copyvio we shouldn't be keeping it in draft space either. —David Eppstein (talk) 20:00, 26 January 2023 (UTC)[reply]
That makes sense to me, but there seems to be reluctance to drop the G12 bomb with only the results of LLM-output detectors to go on. XOR'easter (talk) 21:05, 26 January 2023 (UTC)[reply]

Very much support Rory's proposal. At the moment, we have no process for removing AI/LLM articles that are very obviously not suitable for an encyclopedia, but they fall through the cracks in the existing G and A CSDs: they're not patent nonsense, they're not blatant hoaxes, they're not unambiguous copyvios, they're often on subjects that appear notable, and they're not obviously invented. A CSD tag is likely to not be applied, or if applied to be turned down. With no process, these questions end up at AN, ANI or VP instead, waiting for someone to delete them out of process. This is a ludicrous state of affairs. With a forthcoming clean-up tag for drafts that might have something salvageable in them, we'll have a useful process for dealing with this bollocks and can go about our business free to pursue a life of religious fulfilment. — Trey Maturin™ 18:33, 26 January 2023 (UTC)[reply]

What objective criteria are you proposing to use to determine what content is AI/LLM-generated and which isn't? Thryduulf (talk) 19:02, 26 January 2023 (UTC)[reply]
There are an increasing number of LLM detectors. I would suggest figuring out which, if any, have consensus as reliable detectors and if at least one does setting a threshold (a very high one - I'd throw out 99%), and saying if it meets that percentage on a recognized detector it can be speedily deleted. Best, Barkeep49 (talk) 19:19, 26 January 2023 (UTC)[reply]
A field in a CSD-G15 template (and the clean-up tag for less obvious/potentially salvageable cases) to provide a link to our chosen detector and a percentage score would be an idea. It should even be possible to have a bot see the addition of the template, poll the chosen detector(s), retrieve a % figure, and add that and a backlink to the template for reviewing admins to, er, review. — Trey Maturin™ 19:58, 26 January 2023 (UTC)[reply]

Strongly support this as either a new CSD or an addition to G3. Draft:Gecko's archive was deleted under G3, however this was something of an IAR case because one could argue that it's a notable topic that could be a viable article if someone corrected all of the false facts, replaced all of the fabricated references and rewrote it in the appropriate style. If folks are worried about how to tell if it's actually AI generated, why not apply this to any and all articles that consist mainly of false/fake/fabricated content and references? –dlthewave ☎ 19:04, 26 January 2023 (UTC)[reply]

I am inclined to agree here: a criterion should probably look more like "article with completely fake references and gobbledygook text", and less like "created by this specific process". After all, these drafts would have been trash regardless of how they were made – it's just that there is a newly available technology which people are able to abuse in a way that produces this specific type of trash. jp×g 19:54, 26 January 2023 (UTC)[reply]
This is where the associated clean-up tag would come in, to catch AI articles that don’t qualify for CSD-G15 or have a glimmer of promise that a dedicated editor might be able to exploit. We certainly shouldn’t be speedying all LLM-derived articles, but we also shouldn’t be just leaving them be whilst we have debates about them in ever-more obscure places hoping that an admin will IAR delete them. A two-pronged approach, as we already have for all articles (speedy the obvious, tag the non-obvious), would cover most cases. — Trey Maturin™ 20:07, 26 January 2023 (UTC)[reply]
Thanks for making that maintenance tag. I like the idea of it, and I think it'd be easier to get consensus for than a new CSD. But I still think it's too long. I think we should shorten it to two sentences. I'd also encourage moving it out of userspace, to allow some bold editing and iteration by other folks. Hope this helps. –Novem Linguae (talk) 21:20, 26 January 2023 (UTC)[reply]
I went ahead and (with attribution) created Template:AI Generated which includes my attempt to get it to two sentences and to bring the wording more in-line with other content warnings we give (particularly about how to fix the problem). Courtesy ping @JPxG @Trey Maturin @Novem Linguae. Best, Barkeep49 (talk) 17:15, 27 January 2023 (UTC)[reply]

Okay, well, since we're on the subject, what should I tag this with? Draft:Draft:GPT-4 shows some obvious tells (i.e. is structured like a high school essay and lacks any real detail) and is scoring about a 6% on the analysis pages that have been linked earlier. jp×g 06:38, 27 January 2023 (UTC)[reply]

If you're sure it's AI-generated, I would BLAR it to the other draft at Draft:GPT-4. 6% doesn't sound like very good evidence though. Maybe just leave it for now. –Novem Linguae (talk) 06:53, 27 January 2023 (UTC)[reply]
Sorry: I meant that they returned 6% likely to be human-written (i.e. 94% likely to be GPT output). A rather important distinction I failed to make! For what it's worth, anyway (these analyzers are themselves rather seedy and seem to mostly be based on GPT-2). jp×g 09:35, 27 January 2023 (UTC)[reply]
I'm getting 99.98% GPT output for Draft:Draft:GPT-4. I'm going to try to CSD it. — rsjaffe 🗣️ 17:27, 27 January 2023 (UTC)[reply]
The difference between the two checks could be the COPYVIO material I revdelled. Best, Barkeep49 (talk) 17:35, 27 January 2023 (UTC)[reply]
And I've declined the speedy deletion because it doesn't match any current criterion valid in draftspace (it would likely be A10 in article space). Consensus for speedy deletion must come before speedy deletions. Thryduulf (talk) 17:53, 27 January 2023 (UTC)[reply]
Thanks for the info. So, until we get a changes in policy, if none of the current criteria match (bigger issue in draftspace than article space), tag it, and/or edit it to remove unreferenced questionable statements (WP:V). Right? — rsjaffe 🗣️ 18:34, 27 January 2023 (UTC)[reply]
If it's in draft space tag it and, unless there is an active problem, just ignore it. If there is an active problem take to MfD. In article space, take it AfD. Thryduulf (talk) 20:11, 27 January 2023 (UTC)[reply]
For the record, I have written an actual (crappy) article at GPT-4, so maybe this will solve the immediate problem of people trying to write extremely poor-quality versions of it with ChatGPT. jp×g 09:33, 27 January 2023 (UTC)[reply]

Remove F7d / {{Di-disputed non-free use rationale}}?

Reason:

  1. If non-free use rationale is disputed, it is not uncontestable since it is subjective what is "successfully addressed the concern".
  2. This is a 7-day process so it is redundant as files can be PRODded.

GZWDer (talk) 13:49, 31 January 2023 (UTC)[reply]

  • The mere fact something can be disputed doesn't make it subjective. In any case where a page is tagged for speedy deletion, under any criterion, you can dispute this and try to address the concern. If the reviewing admin still thinks the page qualifies for speedy deletion then it can be deleted anyway. F7 isn't redundant to PROD because PRODs can be restored on request for any reason or no reason at all, whereas an F7 deletion isn't likely to be restored unless you can show that the fair use rationale was valid. Hut 8.5 18:09, 31 January 2023 (UTC)[reply]
    • But that's not how Speedy deletion criteria are supposed to be applied. Any good faith editor can remove any speedy deletion criterion at any time, and that action should force a discussion. Why is F7 different? Jclemens (talk) 19:09, 11 February 2023 (UTC)[reply]

#A10 discussion: create exception or create sub (#A10a)

R.E.: There is a discussion about changing—or making an exception to—#A10 criteria that is now taking place regarding reverted splits. Is this situation already covered here? I couldn't find anything. As I believe that this use is a reasonable interpretation of the existing #A10 rule, your input would be appreciated before I make any needed changes to the criteria section. ThanX, GenQuest "scribble" 16:27, 8 February 2023 (UTC)[reply]

I (and at least one other person) have left comments in the linked discussion rather than here to keep it one place. Thryduulf (talk) 22:35, 8 February 2023 (UTC)[reply]
Retrieved from "https://en.wikipedia.org/w/index.php?title=Wikipedia_talk:Criteria_for_speedy_deletion&oldid=1138889978"