Wikipedia:Village pump (policy)/Archive 190

Continuing talk from Wikipedia talk:Criteria for speedy deletion:

I am repeating this issue here because we appear to be deadlocked on how to exactly phrase an expansion of the CSD there is consensus for at

• Attempt 3
• Attempt 2
• Attempt 1

The question is which one of these would be preferable? Perhaps some other wording would be preferable? We can +1/-1 to express support or dissatisfaction for the various wordings. Awesome Aasim 22:34, 28 February 2024 (UTC)

ECR restriction does not constitute a

WP:BAN

. Therefore G5 doesn’t apply.

ECR violations, a non extended confirmation editor writing on a forbidden topic, is not necessarily object, and therefore does not fit with CSD. ArbCom of course May deleting things, but to authorise any admin to use their discretion to delete things in any namespace on the subjective judgement that it touches a forbidden topic, is incompatible with the notion that Wikipedia is a community managed project.

ECR violating material in mainspace can be speedily userfied or draftified. ECR violating material in other namespaces can be responded to with a sticky PROD, seven days of consideration as to whether it really is a violation, and seven days for an extended confirmation editor to take ownership.

There is no need for ECR enforcement to be written into CSD, let alone in a way that trashes the limited scope of CSD, for details read

WP:NEWCSD

.

Only one non-article went to MfD, and it was not deleted. Obviously, CSD is not suitable for non articles. SmokeyJoe (talk) 22:45, 28 February 2024 (UTC)

WP:ARBECR. Is that genuinely so? ~ ToBeFree (talk

) 00:17, 29 February 2024 (UTC)

A formal prohibition is not a WP:BAN. A formal prohibition applied to an entire class of editors, who receive no message about it, is a chilling barrier, and enforcement is WP:BITING of newcomers. I get it that in most cases the account is an ABF non-new editor, but that is not a good reason to treat any genuine new user so badly. Userfication and a message is appropriate for AGF mainspace violations, Sticky-PROD for any other namespace. Speedy Deletion is unnecessary, and fails WP:NEWCSD. ArbCom should be ashamed if they intended their words to override CSD policy.

The word "Ban" is important. In CSD, of all policies, accuracy and precise word use is important. It needs to be accurate and uncontestable. SmokeyJoe (talk) 02:01, 29 February 2024 (UTC)

Technically speaking an expansion of G5 is the simplest to put into effect, as it doesn't require any changes to tools like Twinkle that would be necessary after the creation of G15. An expansion of G5 also represents how admins are already in practice handling deletion of ARBECR and the GS equivalents. A quarry on this shows that there have been at least 16 deletions using G5 in this manner for ARBECR, GS/RUSUKR, and GS/AA violations since August 2023. I say at least, as there's no standardised log entries for this type of deletion yet, and I may have missed some in my quarry search terms.

Ultimately though I don't really care whether it's G5 or G15. All I care about is seeing the consensus from the recently closed RfC, that we either expand G5 or create a new CSD, is implemented. It's been over a month since that RfC closed, and I found it quite bizarre that it hadn't yet been implemented in some way. Sideswipe9th (talk) 22:51, 28 February 2024 (UTC)

We do not do what is easiest, we do what is right. What is right is discussing matters until there is consensus, not ignoring the ongoing discussion because your preferred outcome doesn't have consensus (yet). Thryduulf (talk) 22:55, 28 February 2024 (UTC)

We already have admins enforcing this through G5, amending the wording of G5 to match its use in practice seems like the most straightforward option based on current editorial practice. Again though, I don't really have a preferred outcome other than implementing the consensus from the recent RfC. If it's G5, great, let's do that. If it's G15, great, let's do that. Sideswipe9th (talk) 23:00, 28 February 2024 (UTC)

Yes it would be the most straightforward option, but that doesn't automatically make it the right option. Just because admins are currently abusing the G5 criterion does not mean that the right way forward is to legitimise that abuse, the right way forwards is to teach those admins what the correct course of action is. Thryduulf (talk) 23:06, 28 February 2024 (UTC)

Just because admins are currently abusing the G5 criterion does not mean that the right way forward is to legitimise that abuse Really? Because

WP:NOTBURO states that Although some rules may be enforced, the written rules themselves do not set accepted practice. Rather, they document already-existing community consensus regarding what should be accepted and what should be rejected. (emphasis mine) If the currently accepted practice goes against the written letter of the guideline or policy, then the policy or guideline likely needs updating to reflect the consensus in practice. Sideswipe9th (talk

) 23:53, 28 February 2024 (UTC)

I agree that it should not apply to non-articles like talk pages, drafts, or userspace. But I think we can extend G5 to cover article creations. An admin should be able to speedy delete it if it is inappropriate and created by non-ECP person. The main situations I see here are very biased pages being created. If someone (an admin or ECP person) does want to keep the page then remove speedy delete tagging and do not delete, and take responsibility. Such pages should not be retagged with G5. Also unilateral deletes by admins with not G5 tag should be avoided. THat will give others a chance to consider and take responsibility. On the topic of wording, the second proposal is too long and confusing. The first is too strict. Only an ECP user should be tagging these kinds of articles for deletion, otherwise they too are violating the restriction. Graeme Bartlett (talk) 22:58, 28 February 2024 (UTC)

If you don't want it to apply to every namespace then G5 is the wrong criterion for this. If you want it to be articles only it needs to be an A series criterion (A12), if you want some non-articles then it needs to be a new G-series criterion (G15) that lists what is and is not in scope. Thryduulf (talk) 23:03, 28 February 2024 (UTC)

Is the topic "inappropriate"-ness objective and uncontestable? If not, Sticky-Prod should be used. If yes, A12 is a better fit. In either case, userfication is a solution that doesn't violate deletion policy.

G5 should never be used where you can't link to an individual block/ban or SPI archive. SmokeyJoe (talk) 02:18, 29 February 2024 (UTC)

WP:ARBECR already specifies that only edit requests are allowed. How do you think admins would be able to enforce stuff like non-EC nominations at XfD? Also, what if a non-article topic, for example RfA, is specified as a contentious topic with ARBECR applied? Keeping it "general" solves all of these issues. It's future proofing as well. Awesome Aasim

05:31, 29 February 2024 (UTC)

WP:ARBECR is pretty clumsy. It appears to only consider mainspace and talkspace. It appears oblivious to userspace and draftspace.

Non EC nominations at XfD? By speedy closing, and warning. No need for deletion, let alone a CSD.

You don’t seem to know much about speedy deletion. Why do you not have a CSD log?

You seem to think that future proofing is done by removing rules. Do you have a model where this has worked before? SmokeyJoe (talk) 11:34, 29 February 2024 (UTC)

That's because I don't log my CSD nominations. Sure some of my speedies get declined but not all of them. And with regard to thinking You seem to think that future proofing is done by removing rules, not entirely true. Future proofing is done by making sure the rules can apply in all areas, so in case a general sanction prevents the creation of XfD pages or RfA pages or etc. by specific users, there won't need to be a whole other hassle to this. Am I still learning? Yes, but so is everyone else here, as we figure out what works and what doesn't. Awesome Aasim 14:54, 29 February 2024 (UTC)

If a future sanction prohibits the creation of those sorts of pages, and there is a consensus that deletion is the most appropriate way to handle those pages then would be the time to write a criterion that allows for deletion of those (and only those) pages. We don't allow the speedy deletion of pages on the offchance that in future there might be a consensus to delete some of them in specific circumstances. Speedy deletion is the very limited exception to the rule that pages are only deleted with consensus, and that exception is granted only to pages where discussion will almost always end in a consensus for deletion. So far, none of the discussions about non-article pages have resulted in a consensus to delete. Thryduulf (talk) 15:02, 29 February 2024 (UTC)

If you don't care about unintended consequences, going back to 2001 policy is simplest. Admins may deleted things at their discretion. SmokeyJoe (talk) 02:03, 29 February 2024 (UTC)

That was not at all the 2001 policy actually. Deletion was always heavily controversial, and in those days it was only page titles that were inappropriate for ever becoming a legitimate encyclopedia article with equally unsuitable content that could be deleted, and even then only by Jimbo, lms, and maybe one or two others; just because you had the sysop password didn't mean you could do whatever you wanted with it and yes there was just one shared sysop password prior to Phase II. For example this was the first revision of Ant. Today, completely unacceptable, but back then fine. Unsourced poorly written sub-stubs as a starting point were actually accepted for years. 184.152.68.190 (talk) 17:03, 29 February 2024 (UTC)

(edit conflict) SmokeyJoe is correct that there is very clearly no consensus to delete non-article pages, let alone speedily delete. Awesame Aasim's summary is additionally incorrect in that there are more than two options - A12 (which with some workshopping could be compatible with NEWCSD, unlike G5 and probably unlike G15), (sticky) prod (which both meets the spirit of the RFC outcome and is compatible with the deletion policy), XfD only (compatible with deletion policy but not the RFC result, although consensus can change of course) or do nothing (not ideal). Please read the discussion at WT:CSD for all the arguments for and against the various options rather than the biased summary provided in the OP here. Thryduulf (talk) 22:53, 28 February 2024 (UTC)

The text of ARBECR and its derivatives prevent non-extended-confirmed editors from doing any action other than make edit requests on already existing articles, for the topics covered by these sanctions. To quote a comment by an arb that all other arbs responding found convincing from the recent ARCA the restriction interacts with AfD: AfDs are not in talk space and AfD participation is not an edit request. Therefore non-ECR may not participate in any manner in those discussions. That same logic would clearly apply here: page creations are not in the talk space, nor are they an edit request. Therefore non-ECR may not participate in page creations in the relevant content areas. Following that logic through, a G criteria is the only one that allows for full enforcement of the restriction. Sideswipe9th (talk) 23:10, 28 February 2024 (UTC)

That assumes that speedy deletion is the correct outcome for all page creation but 100% of MfD discussions disagree. Thryduulf (talk) 23:13, 28 February 2024 (UTC)

Is speedy deletion the correct outcome for 100% of all articles created by banned editors per G5? No. Sometimes editors in good standing are quite happy to take a G5 tagged article and take responsibility for it. Hell we even see that behaviour in AfD for other contested CSD tags. That doesn't mean that tagging them with G5 or deleting them when no-one wants to take responsibility is incorrect however. Sideswipe9th (talk) 23:32, 28 February 2024 (UTC)

Please re-read

WP:NEWCSD point 2: It must be the case that almost all pages that could be deleted using the rule, should be deleted, according to consensus. (emphasis in the original). The proposed expansion of G5 would include many pages where the demonstrated consensus is that they should not be deleted. Thryduulf (talk

) 23:35, 28 February 2024 (UTC)

I think though perhaps the real question is, does NEWCSD have consensus to be used in this manner? It's a talk page banner and edit notice on

ignore the rule. It seems that logic would also hold for any other rule within it, assuming the banner has the weight of policy or guideline. Sideswipe9th (talk

) 23:51, 28 February 2024 (UTC)

On that note, that talk page header has been there forever and is primarily aimed at new users who frequently suggest new criteria that fail (see

WP:CSDNOT). So it would be wise not to refer to the header when an experienced editor with already the workings and understandings of the deletion policy proposes a new criteria or an expansion of an existing criteria. I don't get the reason for all this drama treating a talk page header like it's prescriptive when it is at best like an essay or an information page. Awesome Aasim

15:04, 29 February 2024 (UTC)

I don't know what the origin of the header is, but dismissing it as "just an essay" or similar is grossly misstating the level of consensus it has. Not a single new or changed criterion has been approved that doesn't meet the requirements listed there in many years. The reason it's such a big deal is that many editors feel that speedy deletion should be restricted only to pages that have explicit consensus to be speedily deleted because we do not like the idea of admins being able to delete things just because they personally don't like it.

See also Wikipedia:Arbitration/Requests/Case/RHaworth where the titular admin was desysopped for repeated failures to follow the speedy deletion policy as written. Thryduulf (talk) 15:12, 29 February 2024 (UTC)

I agree with this. CSD have to be as objective as possible so there aren't borderline deletions, which there are a lot of. The subjective part might be the "no significant edits by others", because what counts as significant could vary from admin to admin. There is nothing else subjective about G5, either the page was created while the user was banned (or restricted), or not. There is no in between, unless if you talk about a user adding information to sections they are topic banned from, in which case the appropriate action would be to blank those sections. Awesome Aasim 15:25, 29 February 2024 (UTC)

where was the consensus decided that all proposed CSDs must meet those criteria to the letter? I'm not aware of a formal discussion, but it's been many years since any change to the speedy deletion criteria that didn't meet those criteria has been approved. I'm very surprised you're picking at rule #4 as that's one that the proposed criteria indisputably do meet. It's #1 and esepcially #2 where the problems lie. Thryduulf (talk) 15:16, 29 February 2024 (UTC)

I'm not aware of a formal discussion Cool, so it's a talk page notice then. One that holds some weight, but one that consensus can form around ignoring.

I'm very surprised you're picking at rule #4 I'm not picking at rule 4. I'm using the logic of the editor who said that if a rule (in the case of that editor it was rule 4, but the general principle holds for all of them) prevents us from implementing a criteria, we can

ignore that rule. Sideswipe9th (talk

) 15:54, 29 February 2024 (UTC)

(sticky) prod (which both meets the spirit of the RFC outcome and is compatible with the deletion policy), XfD only (compatible with deletion policy but not the RFC result, although consensus can change of course) or do nothing (not ideal). The RFC close was The consensus falls somewhere between option 2 and option 3, and those two options are both about CSD. So I think PROD and XFD and "do nothing" are probably off the table, and that we should focus on CSD. –Novem Linguae (talk) 23:12, 28 February 2024 (UTC)

I said the spirit of the RFC close for a reason. (Sticky) prod was not brought up in that discussion, it only emerged as an idea later, so there was not consensus against it. Given the difficulty people are having in writing a CSD criterion that is compatible with NEWCSD (which is why we haven't got consensus for one yet) it's right that we don't rule out other options arbitrarily. Thryduulf (talk) 23:20, 28 February 2024 (UTC)

I respectfully disagree. We went to the trouble of having an RFC. We should follow the close of the RFC. –Novem Linguae (talk) 23:23, 28 February 2024 (UTC)

Generally yes, but where implementing the closure of an RFC is proving significantly harder than anticipated it's right to take a step back and see whether, in the light of this new information, it is still the consensus of the community. This was done with, for example, the ordering of candidates on ArbCom election pages when the consensus of the RFC proved very difficult to implement technically, and with one of the RFCs following the deletion related editing arb case when it turned out that despite getting consensus for something high level, when it came to the detail it proved impossible to get consensus. Thryduulf (talk) 23:33, 28 February 2024 (UTC)

I agree 100% @Novem Linguae. The idea of a sticky PROD like BLPPROD to replace G5 is off the table for now. I do like the idea though as it is a reasonable compromise between those who outright want to speedy ARBECR violations and those who want to send them to XfD. Banned means banned. "Not permitted" means "not permitted"; it does not mean that you can only if the creation is good or good faith. Awesome Aasim 23:57, 28 February 2024 (UTC)

To add on, to have a sticky PROD will require a whole other RfC. Not implementing the consensus there. Awesome Aasim 23:59, 28 February 2024 (UTC)

It took three RfCs to create BLPPROD to solve a more serious problem, along with an Arbcom motion, dramaboard threads, and wheel war. You really don't want a repeat of the process to create a new type of deletion. I helped create it, the process was ugly. Better to stick to the existing deletion methods for all but the most urgent crises which this isn't. Keeping it within CSD will be much easier. The Wordsmith^{Talk to me} 06:36, 29 February 2024 (UTC)

Perhaps, though the above hints otherwise. I wonder though since the committee said deletion was optional in these cases, if most of the time it would be best to just draftify + ECP a page as a standard CT/DS action. At that point if an EC editor wants to take responsibility for it and move it back they can, otherwise the 6-month draft clock acts similar to a PROD. All of that should require only a change in practice rather than policy, might bloat the CT logs a little though. 184.152.68.190 (talk) 17:11, 29 February 2024 (UTC)

Fundamentally,

WP:ARBECR specifies that just because an edit is allowed to be reverted/deleted does not mean that an edit must be reverted or deleted. If we want that to apply then we need to add a word about that in the speedy deletion policy under G5. Awesome Aasim

17:16, 29 February 2024 (UTC)

if most of the time it would be best to just draftify + ECP a page as a standard CT/DS action Would that not go against

WP:ATD-I which states that Incubation must not be used as a "backdoor to deletion"? Sideswipe9th (talk

) 17:24, 29 February 2024 (UTC)

Only if the page is older than 90 days, but I don't think that is likely to come up very often. 184.152.68.190 (talk) 17:31, 29 February 2024 (UTC)

So it's not the 90 day limit on draftifying that I'm concerned about. Most of the time when ARBECR violating articles appear, they're spotted and handled in some way pretty quickly. It's using the "6-month no editing draft auto-delete" timer as a backdoor to PROD or some other deletion reason. That to me seems like it would be contraindicated by ATD-I. Sideswipe9th (talk) 17:35, 29 February 2024 (UTC)

I'll quote the policy in full:

Incubation must not be used as a "backdoor to deletion". Because abandoned drafts are deleted after six months, moving articles to draft space should generally be done only for newly created articles (typically as part of new page review) or as the result of a deletion discussion. Older articles—as a rule of thumb those older than 90 days—should not be draftified without prior consensus at AfD

The way I read that in context is that the prohibition applies only to newly created articles.

That said if it we were only using this with the intent the pages be deleted that would still be disingenuous, and its important to be open and honest on exactly what we are doing. The intent however is to give ECP editors a chance to take responsibility for the page, if they do, even if only to move the page back then no move-warring should happen and the page can be treated as though created by an ECP editor, if not they get deleted. Arguably this can already be done by patrolling

CAT:SD, but this is far less time sensitive. 184.152.68.190 (talk

) 17:46, 29 February 2024 (UTC)

Unfortunately, draftification does not resolve the fact that the page was created in violation of an arbitration or community remedy. The only resolve would be for a user not subject to the sanction to remove the tag or to request undeletion (and take responsibility for any problems). As soon as that happens G5 no longer would apply. Awesome Aasim 17:36, 1 March 2024 (UTC)

The thing is that the arbitration remedy specifically says the pages may be deleted, but not that they must be deleted. The idea here is to try to find compromise between the two positions since draftification is not prohibited. In a sense this is a minimalist solution that requires no change in policy only a change in practice which I thought should be on the table for that reason alone. There may be an even better option that fits within those parameters. Personally I kind of lean in favor of a new A12 with a C1 style hold as described below but that will require more work. If you have any better ideas feel free to propose them; we're still in the brainstorming phase and I don't claim that either of the two I've thrown out there are perfect. 184.152.68.190 (talk) 19:49, 1 March 2024 (UTC)

may be deleted, but not that they must be deleted. Sure, but tagging a page with a relevant CSD is not a sign that it must be deleted either. In almost all cases it's a judgement call on the part of the processing admin whether or not the page meets the nominated criteria. Sideswipe9th (talk) 23:06, 1 March 2024 (UTC)

In almost all cases a page either meets the criteria or it doesn't, because the criteria are as objective as possible. If a page does meet the criterion it should only exceptionally not be deleted. Together this means there is almost no subjective judgement required from a CSD patroller - especially as if it is not clear whether a page meets a criterion it does not ("Administrators should take care not to speedily delete pages or media except in the most obvious cases.").

This is why the speedy deletion criteria are written as they are and why they are intended to be interpreted narrowly. The point of the NEWCSD requirements 1 and 2 are to ensure that new and expanded criteria are compatible with this (3 is partly to avoid speedy deletion when there is no benefit, and partly to ensure that there is enough precent to ensure that there really is a consensus that [type of page] should always be deleted; 4 is about avoiding unnecessary redundancy, avoiding possible conflict between criteria and keeping the whole thing manageable). Thryduulf (talk) 23:48, 1 March 2024 (UTC)

I've already explained how I believe G5 or G15 meet NEWCSD. I've also made the point elsewhere in this discussion that there's no policy or guideline specific reason for why we have to follow NEWCSD to the letter. It's at best an edit notice and has all the weight of an essay behind it.

WP:NOTBURO

applies.

Respectfully, you've made it crystal clear that you oppose adding this as a G5 or G15, and why you think NEWCSD should be followed. That's fair, even if I disagree with your reasoning for both. But you don't need to keep making the same point, over and over again towards me. Your repetition of the same arguments isn't going to convince me to change my mind, any more than my repeating of my arguments is likely to convince you. Thankfully

consensus doesn't require unanimity, we don't have to agree for either of our respective positions on this to find consensus. Let's let other people talk, and see what they think. Sideswipe9th (talk

) 00:06, 2 March 2024 (UTC)

My previous comment was not in reference to the proposed criteria, but noting that your comment that Sure, but tagging a page with a relevant CSD is not a sign that it must be deleted either. is incorrect, and explaining why it is incorrect.

You keep denigrating

WP:NEWCSD

as just a talk page notice, but if it was just that then it would be trivial for you to find multiple examples in the past decade and a half where consensus has been not to apply it - I've been active at WT:CSD for much of that time, and I don't recall a single instance.

You have indeed asserted your opinion about how NEWCSD is met, but every single one of those reasons has been rebutted (usually multiple times) by multiple people. Consensus is determined by the strength of the arguments, which is why I sometimes feel it beneficial to make the same point from multiple angles to reinforce the strength of the argument. Thryduulf (talk) 00:57, 2 March 2024 (UTC)

is incorrect, and explaining why it is incorrect If that was the case, then every CSD tagged page would be deleted. However many get rejected. Sometimes they're placed in good faith but for the wrong reasons per the criteria. Sometimes they're contested by other editors who have the page on their watchlist (and in some of these cases the page is later brought to the relevant

WP:XFD

where it may or may not be deleted). Hell, sometimes disruptive editors even place them in bad faith, and they're later rejected by an admin patrolling the relevant CSD category. Like any deletion process on enwiki, CSD has an element of subjectivity behind it.

I've been active at WT:CSD for much of that time, and I don't recall a single instance The recent RfC on the status of G5, that lead to this discussion is one such example. Two editors, SmokeyJoe and Crouch, Swale, raised NEWCSD as a reason to not modify G5, but no-one seemed to support or otherwise remark on those assertions. As the closer of the RfC remarked the consensus was somewhere between option 2 (adding ARBECR/GS violations to G5) and option 3 (adding a new ARBECR/GS speedy criterion), so there is already an inherent consensus there that NEWCSD isn't the final word. Others have taken the same position I have in the post-RfC discussion, that NEWCSD shouldn't be seen as an obstacle or hindrance to implementing a consensus. Even in this discussion, the only editors seeing NEWCSD as being prohibitive are you and SmokeyJoe. No-one else other than myself, in direct replies to you, have even mentioned it.

If strict NEWCSD compliance was a bigger issue here, I'd expect more editors to be remarking on it.

but every single one of those reasons has been rebutted (usually multiple times) by multiple people Examples? I know you've disputed it, and that SmokeyJoe agrees with one of those points. But I don't recall reading any other editors disputing it, much less multiple times. Sideswipe9th (talk) 03:08, 2 March 2024 (UTC)

There is no conflict between the RFC close and NEWCSD, because the RFC close did not specify how either of the options should be worded - it is implicit that the exact wording would need to be agreed and that that wording would need to be compatible with NEWCSD (because all new and expanded CSD criteria are). It's not impossible I suppose that some of those voicing support of making ECR violations speedy-deletable were doing so with the attitude that was so vitally important that it should be shoehorned in as soon as possible without regards to the potential consequences, but it certainly isn't reasonable to regard every comment as meaning that when not a single person mentioned it. The post-RFC discussion at WT:CSD failed to come to a consensus (which is why we are here), and this discussion has not reached one yet, so it is clearly not an example of a consensus not to apply NEWCSD.

Given that your last comment was a complaint about my repeating things unnecessarily, I shall refrain from unnecessarily repeating the rebuttals to your points. Thryduulf (talk) 03:23, 2 March 2024 (UTC)

Given that your last comment was a complaint about my repeating things unnecessarily, I shall refrain from unnecessarily repeating the rebuttals to your points. Fair enough. Though will you please strike the sentence where you said but every single one of those reasons has been rebutted (usually multiple times) by multiple people? Because as far as I can tell, that has not happened. Sideswipe9th (talk) 03:54, 2 March 2024 (UTC)

Rebutted, and "rebutted in a way that will

WP:SATISFY you personally or that will make you change your own mind" are importantly different. His opinion is that every reason you've given has been rebutted. Editors should not strike their opinions just because you happen to hold a different opinion. WhatamIdoing (talk

) 03:13, 7 March 2024 (UTC)

@Sideswipe9th: Regarding tagging a page with a relevant CSD is not a sign that it must be deleted either yes I agree with that. Speedy deletions are routinely declined and they are supposed to be for uncontroversial deletion only. We even have for example {{g8-exempt}} that says yes this does meet the criteria but I believe it should be kept barring an MfD, which is honored unless placed by the page creator.

I was simple responding to Awesome Aasim's statement that it didn't resolve the issue by pointing out there's no issue that must be resolved. The pages don't have to be deleted so no violation occurs by draftifying. There are other issues, compared to hypothetical delayed CSD I believe drafts would be less likely to receive attention. Compare for example the percentage of drafts moved to main as the deletion looms vs percentage of declined PRODs, but I'll admit it's speculative.

I remain hopeful that some way will be found to balance the concerns as expressed that this will result in deletion of pages that formerly would've been kept just because they meet the specific criteria even if the pages are well-written and compliant with policy, with the concerns that without a clear formal process the pages created in violation of ECR will not receive the screening they should, but as is so often the case that's easier said than done. 184.152.68.190 (talk) 02:29, 2 March 2024 (UTC)

There is judgement needed when dealing with CSD, but almost all of it is in determining whether the page meets the criteria or not rather than determining whether a page that does meet the criteria should be deleted. {{G8-exempt}} is not saying the page meets the G8 criteria but shouldn't be deleted, it's saying it is covered by the clause which states This criterion excludes any page that is useful to Wikipedia and thus does not meet the criteria. Thryduulf (talk) 03:37, 2 March 2024 (UTC)

That's fair I should have explained that less clumsily. I guess the better way to put it would be that the judgement of any independent good-faith contributor as to the non-applicability of a CSD is honored as a matter of course. Thus for example, if you placed {{G8-exempt}} on a talk page consisting of a completely random out of place comment, I'd still be compelled to send the page to MfD even if I thought it was useless. And after all it might be something that the community needs to examine for reasons not at all obvious to someone casually inspecting the page. 184.152.68.190 (talk) 03:48, 2 March 2024 (UTC)

If keeping it within CSD is vitally important, another potential compromise might be a hold period such as the one for C1 or the old T3. So articles tagged with the hypothetical A12 could be deleted after some specified period of time has passed. Use of ECP following tagging could also prevent removal except by those authorized. As with the other ideas this too has its issues, just trying to put other options on the table. 184.152.68.190 (talk) 19:45, 29 February 2024 (UTC)

This is essentially the sticky prod idea, that some are rejecting out of hand but which I feel is the best of all worlds. Thryduulf (talk) 20:11, 29 February 2024 (UTC)

Yes that was the main objection I foresaw, sticky PROD by another name, but there is precedent for it. 184.152.68.190 (talk) 20:41, 29 February 2024 (UTC)

• Deleting pages for no other reason than this "restriction" is ridiculous. Bans are for dealing with experienced users who have caused problems and it is in the interest of the project and the user to have such bans. They aren't for new users. I'm fine though with deleting borderline cases but not for no other reason than the "ban". Crouch, Swale (talk) 08:42, 29 February 2024 (UTC)
• My preference is for Attempt 2 (link) and so keep it is part of G5. Keeping it as part of G5 keeps things simple and also reflects which I expect is current practice. Attempt 2 provides that if there are substantial edits by other (non-prohibited) edits than the criteria doesn't apply. It also includes a dot point that gives some more detail on deletions under the criterion. Callanecc (talk • contribs • logs) 08:55, 29 February 2024 (UTC)

None - pages created in violation of

WP:ECR should be draftified, not deleted. They should only be deleted if they meet one of the existing CSD criteria. G5 shouldn't be expanded to include ECR violations. Levivich (talk

) 17:49, 1 March 2024 (UTC)

Also, per

WP:LOCALCONSENSUS. Editors who think G5 should be changed need to start an RFC and widely advertise it. Levivich (talk

) 17:56, 1 March 2024 (UTC)

@Levivich Yapperbot was down and I don't think anyone put the thing on centralized discussion. It was on [1], which should have gotten people watchings' opinions. Otherwise we would have had more participation. In any case, we can have a more thorough RfC below to truly determine consensus. I'd phrase it the same way with the options being the same, but I would probably add different wordings rather than just deciding how to actually implement it. Awesome Aasim 18:14, 1 March 2024 (UTC)

I'm usually not one to question the legitimacy of an RfC, but that attracted less than 25 people from my quick count. Something as important as a new CSD criteria should be advertised at

WP:AN too. The Wordsmith^{Talk to me}

18:46, 1 March 2024 (UTC)

(edit conflict) Another RFC with options that are just vague desires for some sort of change would be a complete waste of time as we'd just be back where we are now. Any RFC options should be presented something like:

• All pages (including articles, talk pages, drafts, and pages in userspace) created in violation of ECR restrictions should be speedily deletable. This should be achieved by rewording G5 to read [proposed wording]
• All pages (including articles, talk pages, drafts, and pages in userspace) created in violation of ECR restrictions should be speedily deletable. This should be achieved by a new criterion (G15) [proposed wording]
• Articles, but not other types of page, created in violation of ECR restrictions should be speedily deletable according to the following new criterion (A12) [proposed wording]
• Pages created in violation of ECR restrictions may not be speedily deleted but should be subject to a (sticky) PROD-like process. [proposed wording]
• Pages created in violation of ECR restrictions may not be speedily deleted and should not be subject to a (sticky) PROD-like process.

Thryduulf (talk) 18:47, 1 March 2024 (UTC)

I'd suggest we draft up very specific wordings for each of those options (basically have the diff to add it to WP:CSD ready to go), then have an informal survey of those options in a subsection below (

WP:RFCBEFORE). Then whichever one ends up being the strongest in the judgment of whoever wants to open the final RFC, RFC the strongest option in a binary yes/no way. "Shall a CSD XX criteria with this exact wording be adopted?" or "Shall the following exact wording changes be made to G5?" Anything less binary and specific I suspect would just become too messy an RFC to achieve consensus, since there are so many different ideas being mentioned. –Novem Linguae (talk

) 21:05, 1 March 2024 (UTC)

I was thinking something along the lines of "ranked choice" because it allows for better evaluation of consensus. Awesome Aasim 22:45, 1 March 2024 (UTC)

Yes I think that's a good idea, historically once more than three options are on the table the chance of a no consensus close rises dramatically unless we do something like that. Might also want to try a quick informal survey to see how far apart people's positions are right now to see if additional compromise options can be found. 184.152.68.190 (talk) 02:39, 2 March 2024 (UTC)

If a ranked choice !vote were to be held now between the 5 options above, my strong preference would be sticky prod > nothing > A12. I might be able to support a G15 in a pinch but the exact wording would be critical, I cannot think of any way in which expanding G5 could be better than G15. I'd put draftification equal with sticky prod if that were added as an option. Thryduulf (talk) 03:31, 2 March 2024 (UTC)

I proposed exactly this just above, and if my analysis is right this doesn't require any changes to policy at all. Just move+ECP, as there is no requirement to delete only a discretionary ability to do so. Then EC editors have 6-months to either take responsibility for the page by either moving it directly back, or improving the starting point to a version they are comfortable taking responsibility for. 184.152.68.190 (talk) 19:56, 1 March 2024 (UTC)

pages created in violation of

WP:ECR

should be draftified. Agree. As a draft, allegedly in violation of ECR, a week at MfD is appropriate. A series of MfD will show whether a CSD is justified.

A week allows for the newcomer to learn what the problem was. SmokeyJoe (talk) 03:07, 2 March 2024 (UTC)

I understand where this is coming from, and as an experiment/trial it should in theory work to give us much more data both on what the typical ECR violating page creation looks like and what the community consensus on deletion is for those pages. But MFD is usually quite low-volume, I'd be reluctant to dump a bunch of new pages on the MfD regulars and yes I know you are one, or at least used to be one unless we have an idea of the numbers likely to be in place first.

Furthermore what's the scope? Only drafts? Some proposals extend to talk and userspace which would add for example non-EC users creating userpages that have pro-Palestine or pro-Israel userboxes on them or talk pages created with say {{WikiProject Israel Palestine Collaboration}} on top off the drafts. I'm not sure how many that would add but it would add.

The page creator will be prohibited from participating in the MfD, which doesn't mean they can't read it but we might get just as much educational value from just explaining things on their talk page.

Finally if people start objecting to the trial we will need yet another RfC. Again it's not a bad idea, but not without its pitfalls. 184.152.68.190 (talk) 03:35, 2 March 2024 (UTC)

MfD has a robust structure, and when important things go there, a lot more than the regulars turn up. SmokeyJoe (talk) 03:46, 2 March 2024 (UTC)

Fair enough, I'd still prefer to see some numbers, but I trust your judgement. 184.152.68.190 (talk) 03:52, 2 March 2024 (UTC)

Maybe a super clever user could list the top MfD discussions by number of participants? And if a test is in play, test cases can be pointed to from WP:CENT. SmokeyJoe (talk) 04:22, 2 March 2024 (UTC)

MFDing a draft created in violation of ECR would be a waste of time. MFDing drafts in general is a waste of time. If a draft isn't eligible for CSD (e.g., it's not an attack page, etc.) there is almost no reason to delete it. Levivich (talk) 03:52, 2 March 2024 (UTC)

The point is to run an experiment. Are these pages typically something that the community believes should be deleted or not? Are their any patterns that would allow for the crafting of a more refined G15 or A12? MfDs allow for a more concrete examination. And I suspect that with enough contrivance I could come up with some pathological counterexample that does not meet any of the other CSD that you would nonetheless !vote to delete at MfD. 184.152.68.190 (talk) 03:57, 2 March 2024 (UTC)

I'm sure the answer would be "not." The community will not delete a promising draft just because the editor who made the promising draft wasn't ECR.

If anybody wants to look at ECR-violating creations, at least in the IP topic area, there is edit filter 1276. What you'll notice from scanning the page is that "good" (notable, policy-compliant) ECR-violating articles are kept and the bad ones are deleted, but nobody is going around even trying to delete (or draftify) all ECR-violating articles. Examples from last week: Haim Bresheeth-Zabner, Killing of Sidra Hassouna (the creator of this one is blocked, the article is marginally notable [possible merge candidate], and yet still CSD's get challenged).

This is so for the obvious reason that all editors must always keep in mind: We are here to build an encyclopedia. We are not here to make up a set of rules and enforce them. We are not going to delete promising work just because the editor who made the promising work didn't meet some qualification requirement that we set. The purpose of the qualification requirement is to make it easy to filter out the crap -- to revert disruptive editors -- it's not to filter out everything, the good with the bad. We will filter out the bad and keep the good. That's why a CSD for ECR violations is never going to get consensus IMO.

I'll tell you as an editor who works on this topic area: I am not going to write every damn article in this topic area and neither is any other editor. If somebody comes by and writes something good, even if they're not ECR, I'm going to preserve it (even if it's in draftspace), and if somebody else tries to delete good work, it's going to piss me off (and lots of other editors, too) and I'm going to fight it because it just adds more work for the rest of us.

Unless I'm mistaken, there are only two topic areas with ARBECR in effect: Antisemitism-in-Poland and Israel/Palestine. Those are two topic areas where governments (Poland and Israel) have made efforts to rewrite history on Wikipedia, and where there are (perhaps as a consequence) some of the largest and most active sock-farms. Sometimes the dragnet we use to keep out this volume of disruption also catches some good-faith but new editors. In those cases, we let them through. There is no point in stopping it.

Folks, I don't usually say this as everyone is welcome to edit everywhere and nobody owns articles, but I'm looking at the list of pings at the top of this thread and what I don't see is the names of people who actually edit in the Holocaust or Israel/Palestine topic areas. And I'll say: if you don't have a lot of experience editing in ARBECR-protected areas, you maybe shouldn't involve yourself in CSDing ECR-violating articles. If ECR-violating articles were such a problem that we needed to CSD them, the regulars in the topic areas would be calling for that. I don't speak for them, but I don't think they'd be in favor of CSD or PROD just because the article creator wasn't ECR. Levivich (talk) 04:22, 2 March 2024 (UTC)

I still reckon discussing CSD for ECR violations is premature while ECR violations are not listed at

WP:DEL-REASON. It’s as wrong as ArbCom taking a leading role in deletion policy applied to content, as opposed to being a last resort. SmokeyJoe (talk

) 04:25, 2 March 2024 (UTC)

I appreciate you writing such a long a thoughtful comment; summarized then, you aver the following:

• No experiment is needed because we already know the community's views and they are strongly against deletion
• A CSD for ECR violations is never going to get consensus

 please do correct me if I'm wrong on either point

However the following counterpoints must be contended with:

• There are editors who dispute that claim and believe the community is in favor of deletions as clearly evidenced since
• There are editors in this very thread who believe there is already a community consensus to add a CSD for or modify G5s wording to explicitly cover ECR violations

An experiment would allow a better assessment of the merits of these competing appraisals. That is not to say there aren't other reasons to refrain from starting it. 184.152.68.190 (talk) 04:45, 2 March 2024 (UTC)

Yeah you got me. And the other counterpoint is, according to the preceding discussion, there apparently is "extensive" use of G5 to enforce ARBECR already. As a non-admin, I can't see how many "good" articles we've been deleting, but for all I know, I have it all ass backwards.

To your other two counterpoints, my explanation is that the editors in this discussion and the preceding one by and large are not very involved in the affected topic areas. Once this "issue" is advertised and the "regulars" in the affected topic areas see it, I predict most of those editors will be most concerned with

WP:ATD, far more than anything else, far more than they care about "banned means banned" or what Arbcom's jurisdiction is. Levivich (talk

) 04:52, 2 March 2024 (UTC)

To be sure there is more than one way to skin this cat, and an experiment isn't needed to run a follow-on RfC so the main concern is to balance the desire for additional data with the need to limit the draw on volunteer time. And that later point will carry a lot of weight if it turns out an RfC is needed just to run the experiment.

I nonetheless remain hopeful there is a solution out there that everyone can accept, at least for a time anyway. Let's face it, this is Wikipedia, problems here never get solved, they just metamorphosize into new and even more chronic and intractable problems. 184.152.68.190 (talk) 05:03, 2 March 2024 (UTC)

I was heavily involved in the NEWCSD G13 and U5. For both, the driving rationale was the history of SNOW deletions at MfD. Arguments and theory were not driving. MfD is well capable of handling test cases. If it’s A12 that’s wanted, send half a dozen or so to AfD and see if they get SNOW deleted. If it’s a new G, or a G5 expansion, Draftify and send to mfd. It will generate real data. SmokeyJoe (talk) 06:41, 2 March 2024 (UTC)

The other places that

WP:ARBECR has been authorized by either ArbCom or the community. Why should it be controversial for community authorized areas? Awesome Aasim

16:28, 2 March 2024 (UTC)

I stand corrected, I hadn't thought of the GS areas. I don't think we should treat the GS and CT areas differently; everything I wrote above about IP and APL applies equally to AA, RUSUKR and KURDS. Levivich (talk) 16:40, 2 March 2024 (UTC)

@Levivich "normal" G5's (for sockpuppets and the like) are also declined all the time. Indeed, with the exception of copyvio's (and probably attack pages), there's no punishment or even social expectation against an uninvolved user declining the speedy, even if they say they believe the criteria are met (but want to keep the page for other reasons). Mach61 03:21, 3 March 2024 (UTC)

I think we are wrapping ourselves too much in rules that we forget that the spirit of the rules are more important than their actual wording. The whole purpose of

Hence the idea for the experiment to see how the purpose of avoiding counterproductive discussions can best be achieved. In the short term it may lead to excess discussions, the existence of which as SmokeyJoe pointed out was the driving force in CSD expansion. In the long-term though it should help to tailor the specifics in a way that meets both the desire to add well-written content and to avoid wasting any more time on dealing with disruption.

I understand what you are attempting with the proposed text, but the issue is that some may find it insufficiently precise and objective. This is why I think the earlier discussion focused on keeping it simple with verbiage that said any EC editor could remove the tag and take responsibility. 184.152.68.190 (talk) 21:29, 2 March 2024 (UTC)

How about this wording for G5:

G5: Creations in violation of bans, blocks or general sanctions
This criteria applies to any page created in violation of a ban or block, or in violation of a general sanction, with no substantial edits by others not subject to the sanction, and that is unlikely to survive an XfD. If an editor not subject to the sanction removes the speedy deletion tag or requests undeletion at

WP:RFUD

, it should imply that the editor is willing to take responsibility for the violating edits; thus G5 would no longer apply.

To address the elephant in the room regarding the proposing guidelines:

1. Objective: This is about as objective as we can make it. The "substantial" automatically excludes minor edits, whether flagged as such or not, and the third part is in the spirit of the use of arbitration and community sanctions to remove problematic material, leaving it at administrator discretion.
2. Uncontestable: This is about as uncontestable as we can make it. The community has authorized the Arbitration Committee to assist in the addressing of disruptive conduct in specific topic areas by setting specific administrative policies. The letter and spirit of the banning policy states A ban is not merely a request to avoid editing "unless they behave". The measure of a ban is that even if the editor were to make good or good-faith edits, permitting them to edit in those areas is perceived to pose enough risk of disruption, issues, or harm, to the page or to the project, that they may not edit at all, even if the edits seem good.. Also, This does not mean that edits must be reverted just because they were made by a banned editor (changes that are obviously helpful, such as fixing typos or undoing vandalism, can be allowed to stand), but the presumption in ambiguous cases should be to revert.. In other words, it can be agreed a high quality page that follows all of our policies and guidelines should not be deleted simply because it was created by a banned/blocked/restricted/sanctioned editor.
3. Frequent: We frequently (and semi-frequently) deal with disruptive page creations in violations of sanctions. Unfortunately there is no other reasonable enforcement measure that can address this. Blanking the page is unacceptable, and draftification does not fix the "created in violation of" part. At the best, draftification should only be used if a non-sanctioned editor is willing to help edit the content to make it comply with Wikipedia policies.
4. Nonredundant: G5 is already de facto being used by enforcing administrators (if you see the
   WP:GS/KURD
   ). And even then, some are citing "arbitration enforcement".

I don't think there is anything that can be more objective than this. Awesome Aasim 23:54, 2 March 2024 (UTC)

Thanks for taking the time to write a lengthy explanation. For a minimalist change, that is not bad. I forsee a few objections

• The portion that states unlikely to survive an XfD is going to draw complaints because there is siginficant room for interpretation This is also true for existing CSD but people want to avoid writing stuff like that ever again
• It is distinct enough that it would be better spun out as a separate CSD (G15) for the same reason that we don't merge A7 and A9 yes G6 is already a grab bag but there is strong opposition to ever doing that again.
• Consensus only exists to delete articles so it must be spun out as A12 with corresponding rewording
• There is not in fact any consensus for this CSD see comments by other participants in this thread above and so any addition is inappropriate.
• Draftification is a fix for "created in violation of" since if an EC editor is willing to remove the CSD request they surely have the ability to subsequently draftify it as a normal editing action and request ECP at RFP under CT or DS as a normal editor action
• It is contestable because the from the data we've gathered from the filter on ARBPIA ECR violations shows such pages are often retained

Nonetheless I think its a reasonable starting point for discussion on a proposed wording regardless of whether it ends up being a new criteria, which once workshopped could be an option in the ranked choice RFC. 184.152.68.190 (talk) 00:20, 3 March 2024 (UTC)

The

Israel and Judah

will also be triggered by the filter.

That said, I looked at some of the pages related to the topic and there is a lot of ban and block evasion and more than several articles tagged with notability problems that I can see why ArbCom imposed ECR in the topic area. Only a couple of pages were actually expanded on by other editors and then ECP'd. One example is

WP:GNG. I found a third page creation (redirect) that I tagged with G5, Muslim–Israeli conflict, and it is clear that ban and block evasion as well as POV pushing, etc. are rampant in these topic areas. People want to push their cause onto Wikipedia but this is not the place for it. Yes, deletion is not cleanup, but having very low quality pages on very complex and contentious issues is worse than having nothing at all. Awesome Aasim

02:03, 3 March 2024 (UTC)

@

Wikipedia:If a rule prevents you from improving or maintaining Wikipedia, ignore it. The spirit of that rule certainly doesn't align with there being no exceptions to "not permitted". WhatamIdoing (talk

) 02:34, 6 March 2024 (UTC)

I said "banned means banned" lol

But yeah. This. If an edit is detrimental to the encyclopedia we should obviously revert it, but if it is obviously good we should keep it, regardless on whether the user is banned or not. That does not mean the banned user gets a free pass though. Awesome Aasim 03:39, 6 March 2024 (UTC)

And following that logic to the obvious conclusion:

• If an article created on a subject that normally requires extended-confirmed to edit is detrimental to Wikipedia, then we should obviously delete it, but
• if that same article is obviously good, we should keep it, regardless of whether the user who is extended-confirmed or not.

Given that, why would you try to expand G5 to include articles that ought to be kept? WhatamIdoing (talk) 03:28, 7 March 2024 (UTC)

A bigger question to be asked is why do we have G5 to include articles created by banned or blocked users? Which is why I think we need an expansion about not being able to survive XfD. It should capture the letter and spirit of the banning policy. Awesome Aasim 20:16, 7 March 2024 (UTC)

Draft RfC started

Please feel free to add more options before we actually start the RfC. Awesome Aasim 03:14, 3 March 2024 (UTC)

I moved the draft RfC into Project: space under

Wikipedia:Requests_for_comment/Deletion_policy_expansion_2024. However, it is not open for comments yet. I want to see if there is any other changes that need to be made. If there are no objections to the current format I will just start the RfC. Awesome Aasim

19:14, 7 March 2024 (UTC)

Better link:

WT:MOSIBOX#MOS:INFOBOXPURPOSE and links to related articles RoySmith (talk)

02:27, 8 March 2024 (UTC)

As I've worked on a draft, I've noticed that Wikipedia doesn't have a policy covering those as sources. Should obituaries be used as sources? Ominateu (talk) 19:20, 28 February 2024 (UTC)

Obituaries are fine as long as they are reliable, published secondary sources, which excludes the vast majority of published obituaries, but includes obituaries written by major publications about notable people. SportingFlyer T·C 19:27, 28 February 2024 (UTC)

@Ominateu There's an essay at Wikipedia:Obituaries as sources. Some obits are pretty much ordinary newspaper articles, some are written by family, and so not independent, but can still be considered for simple facts like "had siblings", "grew up in" etc. Gråbergs Gråa Sång (talk) 19:46, 28 February 2024 (UTC)

Why not? Obituaries are among the few things published by newspapers that are genuine secondary sources, describing a person's whole life rather than just a news event. Of course they need to be distinguished from death notices (classified ads taken out by the family of the deceased) and news reports of a person's death.

Phil Bridger (talk

) 20:26, 28 February 2024 (UTC)

At least you're working on a draft. My concern with obituaries is using them on an existing page. Too often, obits repeat unsourced content from Wikipedia, causing a

citogenesis.—Bagumba (talk

) 04:19, 4 March 2024 (UTC)

Per Sporting Flyer above, whether an obituary should be considered reliable or not depends very much on the publisher. A well-researched obituary from a top-tier newspaper etc can often be one of the best biographical sources available, when looking for the sort of background information (upbringing, schooling etc) that a well-rounded biography may need. Obituaries from lesser sources should always be treated with a great deal of caution though. As Begumba notes, it isn't that unusual to come across obituaries that appear to have been lifted almost wholesale from Wikipedia, and both neutrality and independence are very often compromised. Sadly, as a consequence, obituaries of less-notable individuals, who tend not to get covered by the major media sources we'd most trust, are often the least useful to us. I'd be inclined to treat a typical local newspaper obituary as worth citing for date of death, and little more. Certainly not for anything potentially controversial without further verification. AndyTheGrump (talk) 13:59, 5 March 2024 (UTC)

I'd also be fine using them like other primary sources, as a supplementary source to support or give context something that already has a secondary source. Definitely not as the sole source for controversial information, except like you mentioned with the case of a top tier major newspaper. For the smaller or regional ones, a lot of them fail to make clear distinctions between paid obits and original journalism. The Wordsmith^{Talk to me} 17:57, 5 March 2024 (UTC)

Then the rule should be "you need to be able to tell whether it's a real obit or a paid death notice", not "don't use anything from regional papers". (Regional newspapers are often quite large. See some examples at Wikipedia:Notability (organizations and companies)/Audience requirement. You probably don't want to exclude them.) WhatamIdoing (talk) 02:23, 6 March 2024 (UTC)

I might have been unclear in my comment, but I mostly agree with you. I'm not suggesting we should ban regional newspapers, some of which are quite good. Maybe something like "If it isn't obvious that the obit is original journalism rather than paid content, it should probably be treated like a

non-independent source". They can be used to support other things that already have secondary sources, or fill in certain non-controversial details like birth/death dates, but not anything controversial or speaking to notability. The Wordsmith^{Talk to me}

18:21, 6 March 2024 (UTC)

I'm not sure I'd specify "original" journalism, because we don't need any more fights about editors claiming that they magically know that the biggest newspaper in that country got the death date from the family, which would make it "not original", but otherwise I agree with you. WhatamIdoing (talk) 03:20, 7 March 2024 (UTC)

Citogenesis from WP is not a concern if the obit was published before WP had an article about the person, although other caveats would still apply. Donald Albury 15:36, 5 March 2024 (UTC)

Likewise, I also try to source uncited statements with sources from before the specific content in question was added to WP. —Bagumba (talk) 15:43, 5 March 2024 (UTC)

Hmm, it appears that the OP has won a CU block. WhatamIdoing (talk) 02:14, 6 March 2024 (UTC)

This is a worthwhile topic, nonetheless. —Bagumba (talk) 02:35, 6 March 2024 (UTC)

One problem that we have with such discussions is that we have no commonly accepted definition of what a "regional newspaper" is.

Regional newspaper goes to an unreferenced, poorly written section of Newspaper, and the term is used in various places but never explained. As an example, I believe that the San Francisco Chronicle is a regional newspaper, but other editors call it a local newspaper and I have actually heard the argument that each US state can only have one regional newspaper, applying the same standard to California with 39,000,000 people as to Wyoming with less than 600,000 people. Cullen328 (talk

) 03:19, 7 March 2024 (UTC)

See Wikipedia:Notability (organizations and companies)/Audience requirement, which I wrote last year because I got tired of typing it all out for each new person trying to disqualify a source (and in my experience, they're always trying to disqualify a source, not to identify the best ones). There's a longer version at User:WhatamIdoing/Audience requirement if that's not enough to bore you to tears. WhatamIdoing (talk) 03:24, 7 March 2024 (UTC)

San Diego Union-Tribune are not regional newspapers. Only the Los Angeles Times qualifies. That cannot possibly be correct. All the California newspapers I mentioned have large circulation areas, great credibility and journalistic accolades. I am sure that the same can be said of several newspapers in Texas and Florida, though I am less intimately familiar with them. Cullen328 (talk

) 04:10, 7 March 2024 (UTC)

Do you think the part "Examples of US regions that are bigger than local but smaller than a US state include:" from the longer version would be helpful? WhatamIdoing (talk) 04:20, 7 March 2024 (UTC)

I think we're going off on a tangent. The main issue is being able to distinguish between paid and non-paid. In the US at least, I think an obit attributed to a writer can assumed to not be paid (otherwise there would be an advertorial-like disclosure). I think someone skilled might be able to tell a paid obit by the content and style (which we could try to capture with objective criteria). —Bagumba (talk) 04:16, 7 March 2024 (UTC)

I'm doubtful about being able to identify all paid obits from the writing style, though some of them are obvious enough.

Do editors generally struggle with this? I have never had trouble telling which was which, but my experience is only with US newspapers. WhatamIdoing (talk) 04:34, 7 March 2024 (UTC)

Frankly, it's generally a huge improvement if a page was just cited, even if to a paid obit. It's a higher layer of refinement to then debate if it's verifiable by a better source, or just plain unreliable.—Bagumba (talk) 04:43, 7 March 2024 (UTC)

Hi Cullen328. I indented by comment under yours, but I wasn't intending to make a personal criticism. What I was trying to say is that it seems reliable to me if the San Francisco Chronicle has an obituary attibuted to staff, regardless if one considers it a regional or local paper.—Bagumba (talk) 08:05, 7 March 2024 (UTC)

Thanks for agreeing on the narrowest of my points, Bagumba. But your response does not address my main point, which is, how does an editor or a group of editors at an AfD for example, determine whether a given newspaper is local or regional? Back to the question of reporter written obituaries: I think that we both may agree that a detailed, well researched reporter written obituary of a local businessman in a local newspaper contributes less to notability than a similar reporter written obituary in a truly regional newspaper. If you disagree with that, please let me know. If you agree, then please explain how a good faith, thoughtful but inexperienced editor can determine whether a newspaper like the Fresno Bee, as just one of many examples, is local or regional? Cullen328 (talk) 08:47, 7 March 2024 (UTC)

If the Sydney Morning Herald or Washington Post write an article about someone or something from Sydney or Washington, respectively, I always work from the presumption that it's a local paper. SportingFlyer T·C 09:34, 7 March 2024 (UTC)

@Cullen328: I've admitedly just been semi-skimming responses here, so I missed that the issue of establishing notability popped up here as well. I've only been speaking to reliability to this point. Apologies for any confusion. —Bagumba (talk) 09:46, 7 March 2024 (UTC)

I think the difference is that smaller publications or online-first "new media" type places have been making the distinction between paid content and original reporting less obvious in recent years. It isn't just limited to small local papers, some of the biggest offenders are Forbes and Times of India. This doesn't just apply to Obituaries, but that's one area where paid content is most common.

It might be better to distinguish between Obits and newspaper articles about a person's death. I remember learning how to write an obituary in Journalism school; the details were often given by the family/funeral director to a reporter who assembled it. They're very formulaic, which can be good for recognizing them. They generally start with "John Smith, (occupation) died on (date) after a long illness. He was 62", describe their education, career, hobbies, include phrases like "He is survived by" with a description of family, and finish with giving details on the wake/funeral. Articles about the death of someone usually go into much more depth about the circumstances of the death, why they are famous, and what a reader would want to know about them. It might be better to say that articles about a death can be used as secondary sources independent of the subject and contribute to notability, while obits cannot. The Wordsmith^{Talk to me} 16:28, 7 March 2024 (UTC)

Just a note: I write a lot of articles on 19th-century and early 20th-century political figures, and obituaries are priceless for this. Be careful to distinguish obituaries from earlier periods from those published following the great democratization of knowledge of the Internet age. Cheers! BD2412 T 20:55, 9 March 2024 (UTC)

I think there is a problem with the way the word "obituary" is understood in different places and among different age groups. As a 65-year-old Brit the word has always meant to me the description of a notable person's life that was kept on file and regularly updated by broadsheet newspapers and published after a person has died. It was always understood that it might have been written years beforehand. It was nothing like what The Wordsmith describes - that was a death notice, something completely different.

Phil Bridger (talk

) 21:46, 9 March 2024 (UTC)

• Yes. Obituaries in (generally) reliable newspapers and periodicals are (generally) reliable sources. Paid death notices are not obituaries, and should not be described as obituaries. I should point out that historians recommend the use of local newspapers for nineteenth century history. See, for example, Historical Interpretation by J J Bagley; Sources for English Local History by W B Stephens; and Local History in England by W G Hoskins. In fact, they say that the use of local newspapers is absolutely necessary. 19th century newspapers are not 21st century internet publications and should not be viewed through a 21st century lens. James500 (talk) 11:01, 10 March 2024 (UTC)
• I wrote the essay
  WP:RSOPINION by default; the nature of an obituary, even in an RS, is often to eulogize the subject, with no attempt to treat them neutrally. This means you'd often want to attribute them via something like In an obituary in SOURCE, X was described as... or the like. Also, the formulaic nature of obituaries means that they tend to mention things that might not get mentioned elsewhere - which on one hand might be useful if we feel we need a source for them, but on the other hand means they might not be great for establishing due weight. --Aquillion (talk
  ) 11:10, 13 March 2024 (UTC)

Hi. Can you provide me with a policy or consensus on not italicising song titles? e.g. Training Season -> Training Season. Sditor (talk) 02:46, 14 March 2024 (UTC)

MOS:MINORWORK. This sort of question would be good for the WP:Help desk. Firefangledfeathers (talk / contribs

) 02:49, 14 March 2024 (UTC)

Watchers of this page may be interested in joining the discussion at Wikipedia talk:Manual of Style/Infoboxes#RfC: Change INFOBOXUSE to recommend the use of infoboxes — Wug·a·po·des 20:41, 15 March 2024 (UTC)

Richard Wagner
Wagner in 1871
Born	(1813-05-22)22 May 1813 Leipzig
Died	13 February 1883(1883-02-13) (aged 69) Venice
Works	Stage works Other compositions
Signature

A recent discussion at the MoS talk for Infoboxes (see

From this discussion, I will be opening an RfC to modify the wording of

• Proposal A: Amend the first paragraph of
  MOS:INFOBOXPURPOSE with: "However, links to other articles are permitted but should be informative: see Links to other articles
  [linked to the section that would be created by Proposal B]."
• Proposal B: Create a subsection under "Design principles" titled "Links to other articles" with the following text:
  • "Links to other relevant articles, such as lists of works, may be used, but should be as informative as possible. For example, placing "Full list" in the works parameter of Template:Infobox person does not explain what the list is, nor does it include any information about the works within the list. The parameter's content should describe the content of the list(s) and be as informational as possible: for example, "Stage works ⋅ Other compositions" in an infobox for Richard Wagner is descriptive of the lists' content. Like other infobox content, the link must also appear in the body of the article."

MyCatIsAChonk (talk) (not me) (also not me) (still no) 18:14, 15 March 2024 (UTC)

Comments are still appreciated, but I will wait for the RfC on infobox use to be resolved before opening this one. MyCatIsAChonk (talk) (not me) (also not me) (still no) 22:08, 15 March 2024 (UTC)

The due weight policy received criticism for favoring mainstream points of view over alternative ones, which is especially hamrful for creative works where the reception between media outlets and the franchise's fandom differ. Here's an overview of the change I want:

As long as the niche view is not a contested theory (e.g. Flat Earth), it can have as much coverage as the mainstream views, provided that they are marked as niche views 179.105.130.76 (talk) 19:47, 15 March 2024 (UTC)

Although I do think due weight can prove a problem in terms of source quality (for example when there is a general ignorance or bias in media sources compared with academic ones on a particular topic, but the sheer number of media means those perspectives enjoy prominence) (I actually think scholarship should be favoured over media sources in all possible instances), there is a huge danger in misrepresenting the balance of consensus with the way you propose. Which articles are causing you bother on this? Yr Enw (talk) 19:58, 15 March 2024 (UTC)

Giving as much coverage to a niche or fringe view as to the mainstream is entirely antithetical to the purpose of the due weight policy. And where do you get that Flat Earth is not a contested theory? Donald Albury 20:01, 15 March 2024 (UTC)

I think they meant that’s an example where it obviously is contested and so it wouldn’t applicable to Yr Enw (talk) 21:12, 15 March 2024 (UTC)

• Oppose "Fandom" sources of any kind (not just the website of that name) are not
  reliable sources. Cullen328 (talk
  ) 20:05, 15 March 2024 (UTC)
• Oppose I fail to see how this would be an improvement to the encyclopaedia... SportingFlyer T·C 20:23, 15 March 2024 (UTC)
• Oppose Wikipedia summarizes the majority of sources, i.e. the mainstream.
  🌺 Cremastra (talk
  ) 22:27, 15 March 2024 (UTC)
• I would fear that those who support the Flat Earth theory would consider it true and uncontested, as would any supporters of similar ideas. This would simply give them a reason to demand that their particular fringe theory be given equal weight to mainstream ideas. -- LCU ActivelyDisinterested «@» °∆t° 01:45, 16 March 2024 (UTC)
• Oppose proposal by 2-edit spa. Wikipedia is intended to give a balanced view. Xxanthippe (talk) 02:14, 16 March 2024 (UTC).
  • I believe the OP's intent with this proposal is to adjust what "balanced" mean by adjusting the weight given to different types of sources rather than to make it unbalanced. Thryduulf (talk) 11:02, 16 March 2024 (UTC)
• Oppose per Cullen etc. Gråbergs Gråa Sång (talk) 11:41, 16 March 2024 (UTC)

Poking around Wikipedia, I came across an article in the Template:Alternative medicine sidebar. The right side of the article is entirely blanketed with the template, "Fringe medicine and science" when expanded by default is literally as tall as my Chromebook screen.

It got me wondering, is there a point to this? Does someone in one of these articles need to know about the other? Acupuncture, COVID conspiracies, scientific racism, and camel urine seem so divergent as to be irrelevant to each other. Article series are for continuous things (the history of [country]) or discreet things explored in great detail (the life of a US president). -- Zanimum (talk) 13:11, 16 March 2024 (UTC)

For a long time there's been a consensus against including service information on pages about airports and airlines. But I can't see any policy-based reason why such a policy shouldn't be applied to articles about other transport. Now don't get me wrong - this would mean losing a lot of valuable information that has no obvious place to go, and I would be strongly against such an action. But I can't actually see a policy argument for keeping it. Eldomtom2 (talk) 23:24, 6 March 2024 (UTC)

WP:PRESERVE policy. -- GreenC

23:53, 6 March 2024 (UTC)

WP:NOTHOW. — Masem (t

) 01:19, 7 March 2024 (UTC)

@Eldomtom2, could you give an example of "service information" that is excluded from airports, but is included for, say, trains? WhatamIdoing (talk) 03:50, 7 March 2024 (UTC)

See, for instance Avanti West Coast/Services and Euston railway station/National Rail services. My general impression is that while many airport and airline articles have similar sections, when such sections have been challenged and the attention of the wider Wikipedia community brought to them the consensus has been for removal.--Eldomtom2 (talk) 12:20, 7 March 2024 (UTC)

I don't see any problem with those myself, and there's not really consensus around things like airline destination tables. Bus routes yes. SportingFlyer T·C 14:47, 7 March 2024 (UTC)

I don't see a problem with that either. Yes, the section for very major stations such as Euston do appear somewhat excessive. But given that railway lines are fixed, it is useful (and not trivial) to have information about which services run through there and what the previous and next stations are. It's effectively geographical data. This differs from airports; a plane can fly anywhere (subject to fuel, obviously) from an airport, and airline schedules change very quickly. Conversely the fact that "Station X is served by Service 123, the previous and next stations are Y and Z" is pretty much fixed for most cases. Black Kite (talk) 15:16, 7 March 2024 (UTC)

But it isn't geographical data. It is not listing the physical lines from Euston but rather the services which operate on them, which can and do change at least every year or so.--Eldomtom2 (talk) 15:36, 7 March 2024 (UTC)

While the timetables change twice yearly, the routes are much longer-lasting than that with most tending to change only on the order of decades (although there are obviously some exceptions to that), and precise stopping patterns are between the extremes, the information as a whole is sufficiently static that there are no reliability concerns. Thryduulf (talk) 15:44, 7 March 2024 (UTC)

It seems discriminate enough to me. jp×g🗯️ 10:15, 7 March 2024 (UTC)

• WP:NOTTIMETABLE to define exactly what is okay and what's not okay. For airports, it's airlines and destinations tables, for now at least - this argument is evergreen, and stupid (it's clearly encyclopaedic.) For trains, it's either a list of stations served by the line, or a list of lines served by the station. I think even including trains per hour is probably fine. Ports may be able to list current or former destinations. Buses are harder because bus lines are more fly by night. But at the end of the day, I'm not really sure what information the OP suggests to remove here... SportingFlyer T·C
  10:37, 7 March 2024 (UTC)

I'm not suggesting we remove anything. I'm bringing up a potential issue I see and asking what other people think. --Eldomtom2 (talk) 12:20, 7 March 2024 (UTC)

Er, Wikipedia:Village pump (policy)/Archive 187#RfC on the "Airlines and destinations" tables in airport articles, we just had this to say that that airlines and destinations are not appropriate unless DUE (backed by third-party sourcing) Masem (t) 13:23, 7 March 2024 (UTC)

That was an incredibly contentious discussion, with no consensus on whether it was correctly decided after a review. SportingFlyer T·C 14:45, 7 March 2024 (UTC)

I've seen no one challenging the close, so that RFC stands. Masem (t) 14:52, 7 March 2024 (UTC)

Wikipedia:Administrators' noticeboard/Archive359#Closure review request for the RfC on the "Airlines and destinations" tables in airport articles found "no consensus to overturn" with the closer (theleekycauldron) stating the arguments for and against were roughly equal in number and strength, so it stands but the consensus is a lot weaker than just reading the bold words in the close would imply. Whether you have personally seen someone challenging the close is always irrelevant. Thryduulf (talk) 15:35, 7 March 2024 (UTC)

Oppose: While we are not a travel guide (we are an encyclopedia), service patterns at a station are very encyclopedic. What I would be fine with is removing those pseudo-tables on station layout as they are an HTML mess to navigate, etc. and replacing them with SVGs. For example, the

BART is not much different: daily service between 6am-9pm, and daily service 9pm-12am. Awesome Aasim

20:11, 7 March 2024 (UTC)

If by pseudo-tables on station layout you mean those showing the physical layout of the station platforms, then there have been multiple consensuses at Wikipedia talk:WikiProject Trains to remove them. I'll hunt down some links if that is what you are talking about (and would find them useful). Thryduulf (talk) 20:56, 7 March 2024 (UTC)

oppose routes and next stations are really useful (to me anyway)Secretlondon (talk) 13:09, 17 March 2024 (UTC)

Please see subject discussion. Cinderella157 (talk) 12:37, 18 March 2024 (UTC)

(here). As this was my suggestion to bring the conversation over here regarding Tennis statistics used on wikipedia. With the conversation breaking down into various personal attacks and becoming unconstructive, I am seeking to bring this discussion over here to avoid the current line of dialogue and open us up to more opinions regarding the value of the tables being discussed.

The topic has previously been discussed in the project, but I believe specifically "Records against top 10 players" to be mostly

1. Record Against Top 10 players. We had decided by consensus to remove this chart per original research and trivial. A few now feel it's a good chart worthy of inclusion. Note that these players were not necessarily in the top 10 when they actually played each other, just at some point in their careers. Do we bring this chart back?
2. Wins over top 10 players and Wins/Losses over top 10 players. These have been deemed ok by Tennis Project since they show only players in the top 10 when the two met. Each individual match should be sourced but have not been in this example. We are having trouble deciding what is a better choice without too much detail for an encyclopedia. There is some debate about whether "wins only" is consensus here.
3. Looking at an article that has most things, like Novak Djokovic Career Statistics, starting about ATP ranking and downward, is there any advice others can give about what charts could be trivial or original research? We don't want to keep coming back here. Thanks. Fyunck(click) (talk) 00:11, 1 March 2024 (UTC)

With regards to point 2, claiming that the project consumers these ok is a bit of an oversimplification. There are people in the discussion questioning those too. You can include myself in that group too as notability has not been demonstrated. Tvx1 01:03, 3 March 2024 (UTC)

That could be true, but both wins and win/losses have had discussions where they both came out as ok. Swiatek's and a couple other articles came out out as win/loss at the time, and the project was latest at only win at the time. So I hated to say either was the majority choice. This is a village pump on "policy" only (as the title says) and No. 2 is a content issue so I doubt anyone will say anything about that here. I just wanted editors here to know that Tennis project isnt clear on whether to use one or the other all of the time or in specific instances. Fyunck(click) (talk) 00:21, 4 March 2024 (UTC)

As a regular sports editor, but not tennis, are these particular groupings even directly avaliable on stats sites, or require queries and procuring? Is the grouping regular discussed in prose in muliple independent reliable sources, with some mention of other members as well? That would go a ways towards ruling out

They probably are on stats sites, but Wikipedia isn’t such site. They certainly aren’t regularly discussed in prose in independent reliable sources.Tvx1 09:57, 4 March 2024 (UTC)

Whether information is typically found in sources is helpful to know so that we can decide whether we think that information should appear in Wikipedia. For example, if tennis were famous for having a "Top 5" system, then we'd almost certainly not want a "Top 10" here. If "Top 10" is normal, then we have to consider other factors. WhatamIdoing (talk) 03:31, 6 March 2024 (UTC)

Maybe its an undeveloped part of Wikipedia, but I feel your more likely to hear about competitive eras such as the Big Four/Big Three, Four Musketeers and rivalries such as Evert–Navratilova or Lendl–McEnroe than you are about Top 5s or 10s in a given era. But otherwise you'll have articles like this "Stat of the Day: Stan Wawrinka records 60th Top 10 win" from Tennis.com and it does list "Most Career Top 10 wins" for all players in the Open era, but it's hardly a comprehensive source for all players. YellowStahh (talk) 14:32, 6 March 2024 (UTC)

A source with a "Most Career Top 10 wins" list is quite different than having a per opponent breakdown of even more niche stats. —Bagumba (talk) 15:47, 6 March 2024 (UTC)

That's sort of what I was getting at, an achievement would also be winning after losing the first two sets. it's certainly talked about, but it's maybe best left to a statement in his own article rather than an explanatory table. YellowStahh (talk) 19:33, 6 March 2024 (UTC)

What you may see is the simple stat of 62-43 against top 10 players, and nothing more. That is very common on most stats websites including on the WTA and ATP governing bodies websites. Finding an actual chart that lists each of those wins and losses would be very rare, even at stat websites. However there is one sports site that puts the win-loss chart out, Tennis Abstract with Elena Rybakina as an example. I don't know if any others do, certainly not any tennis magazines or places like ESPN. I think it was the formation of the Tennis Abstract website that started the frenzy of including these many charts on the Wikipedia articles. I have never seen any site (stat or prose) even discuss the old "record against top 10 players" chart (#1 above). update In searching I found one other site that gives the exact same top 10 (at the time the match was played) match-by-match breakdown (so #2 above), but it only does the men, not women. So these are stats only websites. The sites are probably as reliable as you're going to find for modern players. As far as prose like magazines or news, I don't think you would find a single player ever discussed in that detail. I'd be surprised if a book on Federer shows a match by match vs top 10. However you will find some prose that says a player has a 62-43 record vs the top 10. Probably need to be fairly well-known players though. Fyunck(click) (talk) 10:30, 4 March 2024 (UTC)

We shouldn't be creating metrics that haven't already been discussed in RS. If sources don't have a detailed breakdown on X games against top X players, neither should we. JoelleJay (talk) 05:48, 6 March 2024 (UTC)

@JoelleJay: As I said and linked above, we have one reliable source that does a detailed top 10 win/loss chart for men and women (at the time the match was played). We have one other reliable source that only does the men. I assume that is how these tennis charts are made... from those two sources. We have no reliable sources that have charts for every player that was ever in the top 10. None that we could find. That is why that chart was removed, but some editors keep adding it back, and is mainly why we are here. To make sure we got it right in removing them. The wins and/or wins-losses top 10 (at the time) charts we also wanted advice as to whether they are appropriate. They are not original research since one reliable source has them (two if it's the men). But there could be some other reason not to include them that you could give advise on? Are they too trivial with only two sources that print them? Would wins only be better? Would none be better? Are they 100% breaking some wikipedia rule we don't know of? Fyunck(click) (talk) 10:39, 8 March 2024 (UTC)

I think it's against the spirit of

WP:OR to come up with the bulk of this from stats databases.—Bagumba (talk

) 12:51, 8 March 2024 (UTC)

@JoelleJay, @Bagumba We can see all head-to-heads vs Top 10 players readily available at live-tennis [2] and [3]

Records against top 10 players are discussed on official sites like WTATennis.com [4] or Tennis.com [5]. Sashona (talk) 18:10, 8 March 2024 (UTC)

Stats databases dont establish notability of the list topic. WTATennis is not an independent source, so should not count towards notability. The Tennis.com list is a list of players w/ most top-10 wins, which can count towards establishing a list of that type per

WP:LISTN, but not for a list with a detailed per opponent breakdown. —Bagumba (talk

) 03:53, 9 March 2024 (UTC)

WTATennis.com and ATPTour.com are the websites of the Tennis Governing bodies, the ATP and WTA which maintain all the tennis match data and records. Those are definitely primary reliable sources. Sashona (talk) 08:15, 10 March 2024 (UTC)

I don't think anyones denying their reliability, it's the fact that they are primary sources.

WP:PSTS is whats being applied here as Wikipedia should be based on reliable secondary and to some extent tertiary sources. YellowStahh (talk

) 14:19, 10 March 2024 (UTC)

This is true... they are reliable sources. But they don't show the chart in full for all these players and that's an issue here. So they are reliable for the actual record of 32-22. They don't show the chart with all the details we have, except in a few instances. I do disagree with editor "Bagumba" about the WTA or ATP not counting towards notability. They do count... certainly not as heavily as a secondary source, but they are a valuable notable source. Fyunck(click) (talk) 03:34, 11 March 2024 (UTC)

WTA and ATP are governing sports orgs and so are not independent sources. JoelleJay (talk) 18:44, 11 March 2024 (UTC)

But it's an iffy term. We can use sources from a player's autobiography but that is not an independent source either by that definition. Secondary sources are best, but not exclusive. The Olympics put out all kinds of details that we use, and that simply gets parroted by the news verbatim. We use tertiary sources as well... they simply are not the best source. Fyunck(click) (talk) 17:47, 12 March 2024 (UTC)

We can use non-independent and non-secondary sources in articles, they just absolutely do not contribute to notability, per GNG. JoelleJay (talk) 00:03, 13 March 2024 (UTC)

I would not personally consider "independent sources" an iffy term as we have our definitions to work with (

WP:IS). However I would ask the difference between this list and say listing every single match a person ever played, and this goes for all players and I understand Federer has articles for all of the seasons he played, but those matches receive more coverage than any breakdown of wins against the Top 10. YellowStahh (talk

) 19:01, 13 March 2024 (UTC)

What is "iffy" is what is considered independent. GNG says "excludes works produced by the article's subject or someone affiliated with it. For example, advertising, press releases, autobiographies, and the subject's website are not considered independent." That does not include the governing bodies of tennis. They do not work for the subject, and they write independent notable articles and stats on many players. Fyunck(click) (talk) 20:08, 18 March 2024 (UTC)

How are governing bodies not "affiliated with" the members they govern? And anyway we have a guideline that explicitly says "Team sites and governing sports bodies are not considered independent of their players" so there is absolutely no iffiness regarding this topic. JoelleJay (talk) 20:32, 18 March 2024 (UTC)

Just because they enforce rules doesn't mean they are affiliated with the player. There is certainly iffiness on that issue, no question about it. But it's probably off topic as we are simply determining original research on point 1... which has been done, and whether there is an issue with charts 2 and 3... no OR but likely trivial since the vast majority of sources don't go into that detail. Fyunck(click) (talk) 20:57, 18 March 2024 (UTC)

The question of governing body affiliation/independence is long settled in our P&Gs and is reflected at AfD. JoelleJay (talk) 01:23, 19 March 2024 (UTC)

No it is not, and I see no policy on it. Goodness, a newspaper writer who is a member of the democratic party wouldn't be able to print on politics and have it sourced here with that narrow of view. And the wording of GNG doesn't agree with you either. We take all of these sources together to come up with whether something is notable. Not one thing only, and we don't throw others out because they are rule-makers of independent tournaments. Fyunck(click) (talk) 01:39, 19 March 2024 (UTC)

I've quoted NSPORT guideline two times in this thread: Team sites and governing sports bodies are not considered independent of their players. Although statistics sites may be reliable sources, they are not sufficient by themselves to establish notability. This has been affirmed multiple times. Sources that have a vested interest in the subject are not independent, so governing bodies/teams/tournaments do not count towards GNG. And a source has to be all of independent, secondary, and SIGCOV to count towards GNG, as should be evident from thousands of AfDs. JoelleJay (talk) 01:45, 19 March 2024 (UTC)

Nor is an independent source like a newspaper alone enough to establish notability. That's what I said. It requires multiple instances to prove notability.... but you don't throw it away like a used napkin. It's part of the whole. We do not exclude it, we use it with other sources to show notability. You are reading into the "guideline" something that was never intended. Secondary sources are the best, but we also use tertiary and other sources to come to an agreement as editors as to whether something is notable. I have no idea how we even got on the tangent of a roller coaster. We aren't even talking about notability but rather OR and trivia. They have nothing to do with one another. Your quote is specifically used for showing notability of a person and whether an article should be created.... not on content. Fyunck(click) (talk) 03:00, 19 March 2024 (UTC)

You do throw away sources that are non-independent, non-secondary, or non-SIGCOV when evaluating GNG. Those sources contribute nothing to notability. There's no other way to interpret multiple published[3] non-trivial[4] secondary sources which are reliable, intellectually independent,[5] and independent of the subject than "multiple sources, each of which is non-trivial, secondary, reliable, and independent". JoelleJay (talk) 03:11, 19 March 2024 (UTC)

Your NSPORT quote does not say to throw them away by any stretch of reading it. You do not throw it away. But again it does not matter at all here since we are not talking about notability of a person but rather the content of a notable person's article. Fyunck(click) (talk) 03:14, 19 March 2024 (UTC)

You throw it away when it comes to determining notability. That's the whole point those sentences are in the guideline. JoelleJay (talk) 06:06, 19 March 2024 (UTC)

We have to agree to disagree with what the wording says, but it's a non sequitur in this discussion anyways. We aren't talking about a person's notability here. Fyunck(click) (talk) 06:13, 19 March 2024 (UTC)

I wrote the part about governing body non-independence, after two well-attended discussions at NSPORT arrived at a near-unanimous consensus that such sources are non-independent and therefore do not count towards notability. Does that really need to be spelled out like we do for trivial sources Trivial coverage of a subject by secondary sources may be used to support content in an article, but it is not sufficient to establish notability and primary sources Primary sources may be used to support content in an article, but they do not contribute toward proving the notability of a subject? JoelleJay (talk) 16:12, 19 March 2024 (UTC)

Not all governing bodies are the same.... you can't lump them together. With the NFL and MLB the players are employees, as with many governing bodies. Tennis players are independent contractors simply playing under some set rules. Their contracts aren't owned by the governing bodies. Apples and oranges here. Plus your link to NSPORT says it can't be the only source for notability.... which sounds right. But again it doesn't matter since we aren't talking about notability here. Fyunck(click) (talk) 18:35, 19 March 2024 (UTC)

If you want to argue we change our guidelines specifically for tennis player governing bodies then you can do so at NSPORT, but the current guidance prohibits any governing org from contributing to notability of the topics it governs and hopefully you have been following that in your article creations. JoelleJay (talk) 19:05, 19 March 2024 (UTC)

Nope, nor does it seem that others do either. All published articles get looked at when it's decided by consensus, like it always has been. Different weight may be given depending on the type of publication, but nothing gets thrown away. And again, this discussion has nothing to do with GNG so it doesn't matter. Fyunck(click) (talk) 20:33, 19 March 2024 (UTC)

...Have you never participated at AfD? Of course non-independent and non-secondary sources are discarded when considering notability. Why do you think assessing source independence and rejecting sources that fail is such a major part of AfD (here's 12,500+ hits, and another 2,500, and 4500+ more) if non-independent sources can contribute?

Not counting governing sports bodies has also been the standard for the last decade, e.g. [6] JoelleJay (talk) 21:03, 19 March 2024 (UTC)

Hundreds of times in the years I've been here. And consensus always wins regardless of a guideline... since it is a guideline, not a policy. Same with RfC's. You could say the sun is orange all you want and have the sources to prove it, but if consensus says it's blue we use blue in the article. Guidelines can't cover every single instance. You make it sound like guideline, and the way you personally write them, are set in concrete... and that has never ever been the case here at Wikipedia. Again a non sequitur here where the topic heading isn't about notability, it's on original research and trivialities and that's what we should stick to. Fyunck(click) (talk) 21:16, 19 March 2024 (UTC)

We're not talking about the rare times where consensus to IAR can override a guideline (though I'd like to see any where a source that was determined to be non-independent was still explicitly counted towards notability), which anyway have no impact on the actual guideline per LOCALCON. We're talking about a key principle required for establishing GNG-based notability--something you brought up, which up until now you've been arguing is not supported in our PAGs. Your new argument seems to be that guidelines are irrelevant because they're not policies and you can ignore them whenever you want while still expecting your opinion to be respected in discussions. That stance is anti-consensus and is not borne out empirically. JoelleJay (talk) 22:53, 19 March 2024 (UTC)

Goodness gracious... like talking to a wall with my words being twisted and stomped on. No thanks to that type of conversation. Fyunck(click) (talk) 23:11, 19 March 2024 (UTC)

With regards to GNG-based WP:notability, they simply don't count. You might call this "throwing away" when evaluating wp:notabiltiy, but that's not a metaphor I would use. North8000 (talk) 19:30, 19 March 2024 (UTC)

The sports-specific notability guideline provides guidance about when a sports figure is likely to meet English Wikipedia's standard for having an article. What is included in an article does not have to meet this standard, though. As Bagumba stated, it's a question of whether or not including this information gives it appropriate due weight. There are a lot of generated stats that some people have discussed somewhere. Articles can't cover all of them, so editors need to weigh which ones are most suitable, based on what has been given the most prominence in appropriate sources. isaacl (talk) 21:44, 19 March 2024 (UTC)

Which was why, I assume, this was brought here. One, to determine if #1 was OR (as the project had determined earlier). That was pretty much confirmed. Then, on #2 & #3, though it is essentially a content/due weight/trivia issue, whether they were breaking any other policy or guideline rules. In reading the responses it looks like no rules are being broken on #2/#3 but due weight and trivia are likely being infringed upon given that only a couple super data sites carry the charts. Thanks. Fyunck(click) (talk) 22:24, 19 March 2024 (UTC)

The fact is that chart has turned into

"AI-generated images aren't copyrightable" Is that true enough? Also, I think I saw a discussion somewhere on a picture made with some app, and the app-terms said something like "you can not release pics made with this app under a commercial free license." I don't know how common that is. Gråbergs Gråa Sång (talk) 07:58, 28 February 2024 (UTC)

I don't know enough about these diffusion engines or about copyright law to answer the question, which I share. I will point out that the Commons upload process currently has toggles for saying I generated this work using an AI tool prompting a dialogue box that asks, Enter the name of the AI engine used. If someone else’s original work was used to generate this work, list them as well.—so I guess that's sort of the "guidance" that exists, not that that's supposed to be a guidance, it's just supposed to be an upload portal, but in the absence of other information... what does an editor learn from that?

I agree with InfiniteNexus's sense that there should at least be guidance to prefer human-made images over images created by diffusion engine-type programs.

talk

) 08:26, 28 February 2024 (UTC)

Fwiw, I started this discussion awhile back: Wikipedia_talk:Manual_of_Style/Images#WP:USERG_portraits Gråbergs Gråa Sång (talk) 12:12, 28 February 2024 (UTC)

Ping to @SMcCandlish if you have an opinion to share on the topic of discussion. Gråbergs Gråa Sång (talk) 12:18, 28 February 2024 (UTC)

I am not a lawyer, but my understanding of the copyright question is that it's unclear. Copyright laws (unsurprisingly) do not address the question and, to my knowledge, it has never been tested in court. I believe the possible options are

1. The images are uncopyrightable, in which case they are probably legally OK to use here as a Free image (but see also below)
2. Copyright belongs to the copyright holders of the training data. Copyright status of the output image would thus depend on the copyright status of the training data used for a model:
   • Exclusively on public domain works then the output images should be public domain too
   • Exclusively on freely-licensed works that don't require attribution (or a combination of these and public domain works). The output images would be free if a license compatible with all the licenses used on the training data is used
   • Exclusively on freely-licensed (and public domain) works. As above, but only if attribution is given to all the relevant copyright holders - I'd be surprised if this was possible in practical terms.
3. Copyright belongs to the developer of the AI, in which case they can choose what license(s) the output may be used under.
4. Copyright belongs to the person who gave the AI the prompt, in which case it would be identical to a human-created image from a copyright perspective.
   • It is not impossible that this will depend on the nature of the prompt, e.g. "flowers in a field" may not be eligible for copyright but "red and blue flowers, grassy upland meadow with Friesian cows, summer day with overcast sky, Alpine mountain background, style of Vincent van Gogh" might be.
5. Copyright is shared between two or more of the above parties. If it's shared between the developers and the prompter, then its possible (at least in theory) for some AI-generated images to be Free. Any other combination would require the training data and its copyright status to be known, and the bullets under option 2 would seem to apply in addition to license used by the other party/parties.

Whether terms like you can not release pics made with this app under a commercial free license. are enforceable (and relevant) has also, afaik, never been determined. If the developers of an AI image generator have any copyright stake in the works then almost certainly they can impose non-commercial or other restrictions as they choose. If they have no claim on the copyright of the image, then I see three possibilities:

1. They are deemed similar to restrictions claimed by art galleries on digitised public domain works (which vary in scope and validity by jurisdiction) and they would not be relevant for our purposes (which doesn't guarantee they would be free of course).
2. They are treated as valid terms of service independent of copyright, in which case whether we regard such terms as relevant to us is for us to decide, either
   • We would treat it like we do restrictions on photography at events - i.e. it's a contractual matter between the prompter and the AI service provider that we are not party to and which does not impact copyright.
   • We would regard such restrictions as binding on whether the image was free or not.
3. They are deemed invalid and thus unenforceable, in which case they are completely irrelevant to the nature of the work.

It is of course possible (in fact almost certain) that different jurisdictions will have different answers to the relevant questions!

Any modifications done by a human to an AI generated image will affect copyright in the same way that the same modifications would impact the copyright of a human-generated image. Thryduulf (talk) 12:06, 28 February 2024 (UTC)

Good overview. I'll add that IIRC it has been demonstrated that image AIs are able to spit out exact or near-exact replicas of images in their training corpus, a twist on your first #2. This isn't realistically detectable by either us or the (human) prompter (we'd have to be familiar with the entire training corpus to recognise the similarity). Would recommend we stay away from them until they become more established - DFlhb (talk) 14:23, 28 February 2024 (UTC)

Appending an article on this. Quote: "any memorization that exists in the model is small, rare, and very difficult to accidentally extract". The Commons essay linked by InfiniteNexus (c:Template:PD-algorithm) mentions the same issue and frames it as something to be tackled when discovered, i.e. the same way we would treat normal uploads like photographs, rather than preemptively banning a whole technology to address this risk. More reasonable than my suggestion just above. DFlhb (talk) 00:30, 29 February 2024 (UTC)

Tend to agree with the broader caution, at least for the short term. Also agree with Thryduulf's excellent summary, which neatly ecapsulates the legal arguments out there so far, or at least agrees with that I've been reading on the issue).  — SMcCandlish ☏ ¢ 😼  00:49, 29 February 2024 (UTC)

AI-generated images aren't images aren't real images, because they have no true subject. They can contain errors that may be misleading. We cannot use them in articles without wholly compromising any claim to reliability.

🌺 Cremastra (talk

) 12:37, 28 February 2024 (UTC)

they have no true subject. They can contain errors that may be misleading Some AI generated images have no true subject, others do; some AI images contain errors, others don't; some AI images are misleading, others aren't. The same statements are equally true about human generated images. These are reasons to carefully evaluate individual images, of whatever authorship, before using them in an article - especially because pretty much every image can be perfectly appropriate in one context and completely inappropriate in a different context. It is also completely irrelevant to copyright. Thryduulf (talk) 12:48, 28 February 2024 (UTC)

Leaving copyright to one side (it's an important issue, but I don't think we can comment reasonably on it until there is a body of case law or changes to statutes in major jurisdictions), what is better about original drawings and diagrams produced by a human using computer tools than those produced by AI? And what is so different about an AI-produced image from an image taken on a camera where a human chooses the lighting, the shutter speed, the focus and many other variables? There are many questions relevant to Wikipedia that we could be working on until the copyright issue is resolved. I am not trying to answer any of them yet.

Phil Bridger (talk

) 13:40, 28 February 2024 (UTC)

A difference is that the human's intent is likely to be closer to what we seek than what the computer's intent is. A human painting a portrait of Baron von Vonvontilherdaddytakesthetbirdaway is presumably primarily to capture the likeness of the Baron, although they may have added intent (to make the Baron look handsomer than he was, or an evil countenance, whatever). To the degree that an AI engine has an intent, it would be to create something that would not be at odds with that descriptor, which is different. And the human can indicate to us beyond the images what limitations it had -- that it had to be drawn from memory, that the Baron asked her to make the hair fluffier, whatever. I've yet to see an AI output include process concerns in its output; if it has no reference images of the Baron, it's going to produce some kind of picture of a human and not tell us of its limitations in process. (Having said that, I don't have much AI experience, and for all I know, that may be an extant feature of some engines for all I know.) This does not mean that AI could not produce something that is sufficiently verifiable by a human-- if we ask AI to create a graph based on data set X, the creator can then verify the data on the graph before uploading it. -- Nat Gertler (talk) 14:01, 28 February 2024 (UTC)

An issue yet resolved by courts is that even if AI images can't be copyrighted nor be considered as derivative works of the training material, is whether various AI generation systems violate copyrights by unauthorized use of material in their training sets which I know are claims still alive in a couple early cases. For us, while that is a level past the AI art, are we going to be comfortable in using AI images that knowing misused copyright (assuming those charges are found true in court). Until we have a clear picture on the full nature of copyright of AI art, we should tear far clear in using it save for examples of AI art on pages directly about creation of AI art. Masem (t) 00:16, 1 March 2024 (UTC)

I think the main issue with using AI images is that there doesn't seem to be much of a consensus among authorities on whether they are copyrightable - and if so by who. And some of them might be

derivative works of other copyrighted works, without it being obvious that they are derivative. Human drawn vs AI on the other hand seems more like an aesthetic preference. Jo-Jo Eumerus (talk

) 12:04, 2 March 2024 (UTC)

Most images are in Wikipedia because they purport to be an image of a particular thing or person. This inherently does not exist for AI generated images. Most are also photographs which are in because they are a photographic record. Again, this does not exist for AI generated images. North8000 (talk) 13:54, 28 February 2024 (UTC)

Whether the thing depicted is what is claimed to be depicted is something that can only be determined at the level of the individual image. Whether we prefer a photograph or other image (and there are lots of non-photogrpahic images) is again something that can only be determined in the context of the intended use. As just one example, the

tin opener we're illustrating the concept of a tin opener not a specific tin opener, which is different to illustrating say Tim Berners-Lee). Thryduulf (talk

) 14:19, 28 February 2024 (UTC)

Yes, all true and good points. I did not intend to indicate otherwise when I made my points above about "most" Wikipedia images. Sincerely, North8000 (talk) 22:24, 28 February 2024 (UTC)

A human-rendered version of the Mars habitat is far better than an AI because the human (likely a NASA employed person) is going to be able to use other knowledge not limited to imagery to establish how the base will be placed, and because these are generated by NASA, while they might look nothing like the final image, their release would implicitly be confirmation this is what NASA thinks it will be, and which may not be explained in other material published by NASA. An AI generated image may use all that art but cannot be able to connect other info to that and thus may generate something that looks like the habitat, but completely in an unfeasible location. For that reason alone, I cannot see any case where an AI image would be preferable to a human-rendered one. Masem (t) 01:13, 1 March 2024 (UTC)

AI-generated art is not copyrightable under U.S. law and belong in the public domain. A federal judge ruled this six months ago [7] [8] [9] [10]. This was a pretty big deal, so I assumed this was common knowledge; apparently not. See also [11]. I don't know what the copyright status is in other countries, but the WMF is headquartered in San Francisco and Wikipedia is thus bound by U.S. copyright law (

Few things "settled" by a district judge are actually considered settled or remain so for long. This is apt to turn into a circuit decision (for or against the district's ruling) and eventually one at the Supreme Court, though this could take years. And that'll just be one (albeit major/influential) national juridiction.

Anyway, from my PoV, "art" ceated by "AI" isn't art at all, it's pixels arranged in a way to attempt to suit the expectations of a textual prompt. It has no (and cannot presently have any intent beyond that. Such output is notriously inaccurate as a general matter (even about really basic things, like how many fingers humans have, what facial muscles do to produce expressions like smiles, etc.), even if a few individual images from that sector might accidentally turn out not to be badly hosed. As such output is not an image (a captured photo, or an intentional attempt to represent in the case of a painting) of the subject, just an attempt to generate CGI that suits a text prompt that mentions the subject, there is no real rationale to use any such image on WP, except perhaps to illustrate an article about "AI" "art" in and of itself. An LLM image illustrating "Margaret Thatcher" or "Botswana" should be removed as pseudo-depiction, or simply falsification to put it more bluntly. Perhaps some kind of argument can be made that LLM image s could be appropriate for depicting things that can't be photographed but which can be mathematically modeled (i.e. with predicable results than can be compared acrossed models), like black holes, and molecular bond arrangements between atoms in complex polymers at the fine microscopic level. But I would expect concerns and caution in this area because of the very, very frequent accuracy problems (even about basic matters) exhibited by LLMs. Such an image would need to be reviewed by subject-matter experts. Another concern is that WP generally doesn't want human editors' own artworks (graphs and maps are often exceptions), even for bio subjects for whom we have no good-quality free-use images if any at all. There is no reason for us to reject those yet accept editors' LLM-generated output, which involves even less human judgement.  — SMcCandlish ☏ ¢ 😼  00:49, 29 February 2024 (UTC)

As you point out, Wikipedia is based in San Francisco, which is in federal district 9, while the ruling you're pointing to is in the DC district. To the best of my not-a-lawyerly understanding, the DC district cannot make binding precedent on the ninth, nor vice vera. -- Nat Gertler (talk) 01:06, 29 February 2024 (UTC)

The U.S. Copyright Office has repeatedly refused to register works generated by AI. One of the applicants sued, which led us to the ruling where the district court affirmed the Copyright Office's decision. The Supreme Court declined to hear the case. Commons has more about this at c:Commons:AI-generated media#Can AI-generated media be lawfully hosted on Commons?: In the United States and most other jurisdictions, only works by human authors qualify for copyright protection. In 2022 and 2023, the US Copyright Office repeatedly confirmed that this means that AI-created artworks that lack human authorship are ineligible for copyright. The Commons community has rejected deletion requests that relied on such copyright claims, and tagged images generated by models such as DALL-E as {{PD-algorithm}}. Our policies are based on the current understanding and interpretation of the law, which could of course change in the future, but the current consensus within the U.S. government and court system is that AI-generated art is ineligible for copyright. InfiniteNexus (talk) 01:57, 29 February 2024 (UTC)

There is no consensus in the court system or government. The DC District Court's ruling does not constitute the current consensus of anyone. It's not even binding in DC, nevermind the rest of the country. District Courts are trial courts and trial courts do not create binding precedent. The District Court's rulings are limited to the single case the court was ruling on and those particular parties. (Barring an injunction, which is not at issue here.)

SCOTUS declining to hear a copyright appeal also is not an endorsement of the copyright office's decision.

There won't be any binding legal precedent on this issue until appeals courts and/or the Supreme Court issues a decision. The appeals court decisions will only be binding within their circuits. Any District Court decisions only apply to their individual cases and will have no binding precedential effect.

The copyright office's interpretations are binding on applicants until/unless they're overturned by a court, but don't reflect the consensus of the entire government, just the copyright office. At most that's one branch of the three branches.

The copyright of AI is a very open question currently being addressed by multiple US courts. No binding or consensus answer yet. Levivich (talk) 16:03, 1 March 2024 (UTC)

There's also two different questions. I don't see the general ruling that computer generated art is not copyrightable necessarily changing, since it's long established that copyright requires human authorship. My sense is the bigger question is copyrighted material in the training data. SportingFlyer T·C 20:46, 9 March 2024 (UTC)

As far as I'm aware, it is "currently being addressed" in the sense that courts are currently addressing these cases by dismissing them on the basis that they do not explain how it's physically possible for their claims to be true. Like I said at Commons a few months ago:

If you want a reason why these models don't contain "fragments of source images", I can give one: it is physically impossible. I can't speak to what goes on with closed-source models, but checkpoints for publicly available models are a few billion bytes of neuron weights (e.g. Stable Diffusion XL 1.0 is 6.94 GB). The datasets these models are trained on constitute a few billion entire images (LAION-5B is 5.85 billion images). I would like to see someone explain how images -- fragment or otherwise -- are being compressed to a size of one byte. It is basic arithmetic.

One byte is eight bits: it is a number between 0 and 255. The binary representation of the number 255 by itself takes one full byte (11111111).

A single colored pixel (i.e. yellow-green, #9ACD32) is a triplet of three bytes (10011010, 11001101, 00110010).

The smallest file on Wikimedia Commons, a GIF consisting of a single transparent pixel, is 26 bytes. The comment you're reading right now (check the revision history for the actual number) is 2080 bytes (minus my signature). This 186 x 200 photograph of an avocado (as a JPEG -- a highly optimized, lossily compressed file format) is eleven thousand bytes. Even if we disregard the well-documented literature concerning neural networks (and the subset of generative models that create images like these) actually work, it is not mathematically possible to achieve the compression ratios necessary to simply store training images inside the model. jp×g🗯️ 18:44, 10 March 2024 (UTC)

So ... is there support for some wording on

WP:IUP to prefer human-created images to AI-generated ones, if not to ban them entirely (which would require a formal RfC)? If editors believe these images aren't in the public domain, they should voice their concerns to the Commons community, but until they regard AI-generated images as copyrighted, Wikipedia is not going to either. InfiniteNexus (talk

) 00:01, 1 March 2024 (UTC)

I oppose banning them. If the copyright or other restrictions turn out to be problematic down the road we can deal with it then, especially as we could use such images under fair use if they can't be hosted on Commons (assuming they meet the NFCC, which can only be determined at the level of the individual image). We should always prefer the best image, and I see no reason to prohibit using an AI image if that is the best image for a particular usage (unless you want to make the encyclopaedia worse to prove a point or something?). At the current level of the technology, an AI image is pretty much guaranteed not to be inferior to a human image when representing a specific person or thing (and in many other cases too), so banning the AI image would be pointless in the same way that we don't need to ban potato paintings. The only question that arises is whether we want to prefer a human image over an AI image when all other factors (quality, relevance, license, aesthetics, etc, are equal). I don't have a reason to oppose doing that, but equally I don't have a reason to support doing that. Thryduulf (talk) 00:33, 1 March 2024 (UTC)

I support creating wording on

WP:IUP

to prefer human-created images over ones generated from diffusion engine-type programming ("AI"). As much as it's the case that there are ways human-created images introduce certain degrees of artificiality (choice of lighting in a photograph, choice of pose in a video, emphasizing or de-emphasizing certain attributes in a portrait, etc.), we are in a much better position to examine and critically consider these human made choices than the inscrutable, at times even 'black box'-esque generation process of diffusion engine-esque programs. As for a ban, for now I would tentatively support that.

And, while this is not the major thrust of my reasoning, the possibility of Wikimedia Commons becoming some kind of repository of images generated using programs that can and do, sometimes by being given such an input and sometimes just doing so without such, imitate and ape the oeuvres of living human artists, troubles me. In valuing accessibility as a project, it hasn't been our goal and it needn't become our practice to tread on the rights of others. Even now Commons' upload process asks the uploader to make sure the photograph doesn't non-incidentally include material that is the protected intellectual property of others. I find myself comparing images generated from diffusion engine-esque programs, and the way they crib from the copyrighted works of artists, kind of like photos that non-incidentally include the intellectual property of others.

talk

) 01:01, 1 March 2024 (UTC)

I think we should consider a hands-off approach, let people argue for inclusion case-by-case and give the community a chance to experiment rather than being risk-averse. Copyright issues are unlikely given the court decisions so far. Masem's argument (which boils down to educational/encyclopedic value) is a lot more interesting to me than the copyright angle, but it's one that would justify giving people leeway to experiment and see if they can find encyclopedic value in it.

And as a point of order - we're just talking about policy on what to transclude from Commons, right? Since AI pics go to Commons. So I'm not sure how copyright is relevant to us. If it's a problem it's for them to deal with. DFlhb (talk) 03:04, 1 March 2024 (UTC)

There is no yet a worldwide agreement on copyrights on AI, we only have affirmation that the US will not grant AI images copyright. So no, they should not go to commons. We probably should made a new image for PD-US-AI, reflecting the copyright office's stance.
Once there's broad worldwide agreement AI images can't be copyrighted, then commons can take them. — Masem (t) 04:49, 1 March 2024 (UTC)

Copyright aside, I am having a hard time imagining a situation in which an AI-generated image being added to a Wikipedia article complies with

WP:NOR. But I suppose in theory an AI generator could be used to create some kind of chart or something that is backed up by RS. Except I don't think AI image generators are good enough to do that yet, like they don't handle text well for example. But from a Wikipedia policy perspective, I'm not concerned about the copyright aspect -- if and when the law catches up, Wikipedia will comply with whatever that law is. Until then, I'd just apply our ordinary policies. If an editor uses AI as a tool to create an image that otherwise complies with policy (i.e., it's not OR), then I don't see why it matters what tool the editor used to create the policy-compliant image, whether it's MS Paint, Photoshop, or some AI image tool. BTW for those who are not familiar, be aware that "AI image tool" is a broad category, covering everything from an AI image generator that creates an entire image, to AI tools that alter images (e.g. by removing something from the image or adding something to it), to AI tools that restore/filter images, and so on. Levivich (talk

) 18:04, 1 March 2024 (UTC)

Don't most photographs and other images in articles fall foul of

Phil Bridger (talk

) 18:32, 1 March 2024 (UTC)

Photographs don't as long as they don't "reach or imply a conclusion not stated by the sources." So if I take a picture of Buckingham Palace and my picture shows Buckingham Palace in the same way that RSes show Buckingham Palace then it's fine. Other images like charts, graphs, maps, tables, infographics, etc., may or may not fall afoul of NOR depending on their content and whether the content is novel or is just summarizing RSes. Levivich (talk) 18:50, 1 March 2024 (UTC)

That's all very well for ultra-notable things like Buckingham Palace, but most photographs on Wikipedia are of things or people that have no published (elsewhere) photographs, or, if they do, they are not in reliable sources. And people are encouraged to take such photographs.

Phil Bridger (talk

) 19:27, 1 March 2024 (UTC)

They are in reliable sources, just a primary source: the thing itself. Imagine I take a picture of a document, like the US Constitution, and imagine I'm the first person to ever take a picture of this document. The document is a source for what the document looks like, a primary source, but a source nonetheless. Anyone could, in theory, verify that my photograph of the document accurately depicted the document by comparing my photograph to the document itself. Same goes with a building, or any other object one can photograph. One way of looking at "OR" is that "OR" is anything that can't be verified in the

WP:V sense. A photograph can be verified, therefore it's not necessarily OR or not always OR. Levivich (talk

) 20:20, 1 March 2024 (UTC)

It's certainly possible for a human-generated photo or other image to violate NOR, but that depends on the image and how it is being used, and to a certain extent how strictly one interprets NOR.

WP:NPOV is another policy that is easier to violate with an image (choice of lighting, angle, etC), however these not things that can be determined without examining the individual image. Thryduulf (talk

) 19:09, 1 March 2024 (UTC)

@

Phil Bridger, see Wikipedia:No original research#Original images. WhatamIdoing (talk

) 01:49, 6 March 2024 (UTC)

• On preliminary thought, assuming copyright issues are worked out: 1) Ban all images of persons, and events in the past century (within the life of anyone alive). We are not here to make, or make-up, modern history or stories about people. 2) I would require any other images to state in detail what sources were relied on to create the image. 3) The rule would be, we would have to disclose, the image is AI; and 4) We always prefer traditional photographs or paintings, such that we have a no use policy if such comparables exist (some kind of NFCC - like policy) (that's at least to start). I think there might be more issues. Alanscottwalker (talk) 19:49, 1 March 2024 (UTC)

  Why? 1) If the images are not reliable depictions of whatever they purport to depict then we shouldn't be using them regardless of whether they are AI generated or not. If they are reliable then I don't understand why we would want to ban them? Why would the subject being more or less than ~100 years old make any difference to the reliability or otherwise of an image? 2) This is a combination of copyright issues (attribution) and point 1: we just need enough information to know whether it is reliable or not - sometimes that might mean a list of all (major) sources, sometimes it will just be the AI engine and prompt. 3) Yes, but again that's a copyright thing. 4) Why would we prefer an inferior image just because it's human-generated (if it is superior there is no reason to ban alternatives)? Thryduulf (talk) 20:01, 1 March 2024 (UTC)

  They are fictions is why. Welcome to reality. In no way should we have the product of, 'Computer: draw a picture of the Palestinian hospital being bombed', anywhere in our encyclopedia. Alanscottwalker (talk) 00:37, 2 March 2024 (UTC)

  That's absolutely terrible reasoning. If you want to make a valid argument, find an example where "Non-famous human: draw a picture of X" would be something we should have while "Computer: draw a picture of X" (and the AI generates good output, no copping out by pointing to "but AI sometimes makes errors like wrong numbers of fingers") would not be, based on reasoning other than "oh noes AI is bad".

  Re "we always prefer traditional photographs or paintings", anyone want to prepare an RFC calling for digital photographs to be banned in favor of traditional photographs or paintings, if that's what we prefer? We have about a month. 😀 Anomie⚔ 06:49, 2 March 2024 (UTC)

  Oh, LOL, how droll. Your comment is the one with absolutely terrible reasoning. "Non-famous human: draw a picture of yesterday's bombing" is not what we want on our pedia, either, nor is a digital photograph not traditional photography in 2024. Most people need no rfc or dumb emojis to know to know what digital photography is in 2024. As for April Fools Day in a month, there is no reason for you to start early with such nonsense. Alanscottwalker (talk) 07:10, 2 March 2024 (UTC)

  I never said "Non-famous human: draw a picture of yesterday's bombing" was something we'd want, nor does anyone I've seen in this discussion. Which is exactly why I called out your comment using that as an example of an AI image we wouldn't want as bogus. Similarly, my attempt at humor was to point out that your appeal to "traditional" photography was similarly irrelevant. I'm sad to see you entirely missed both points and instead resorted to ad hominem attacks. Anomie⚔ 07:39, 2 March 2024 (UTC)

  So, we don't want that computer drawn picture, just I said. So not only was that good reasoning, I was correct. The only point that's being missed is by you, your not making any cogent statement in favor AI, when you agree we don't want that AI picture. Indeed, you are conceding my overarching point, controls are what's needed. And now your not making any cogent argument when you resort to false claims of ad hominem, in defense of your poor humor. Alanscottwalker (talk) 07:58, 2 March 2024 (UTC)

  We don't want bad pictures, regardless of source. If AI pictures are always as bad as you claim then there is no point in banning them, just like we don't ban bad drawings by humans. We do want good pictures. My argument is that we want good pictures regardless of source, your argument seems to be we only want good pictures produced by humans but you have not given any explanation why (ad hominems and straw men are not explanations). Thryduulf (talk) 11:19, 2 March 2024 (UTC)

  Come, now. I explained in my first post why we don't want some AI Pictures, and your not making much of any real point when you say you don't want bad pictures. Your hand-waving. That's not any standard editors can abide by, when they actually address pictures. Alanscottwalker (talk) 11:43, 2 March 2024 (UTC)

  Except the only reasons you gave are (paraphrased) "AI images are bad" and "We always prefer traditional photographs or paintings". The first is not relevant because, even if all AI images are bad (and that can be debated, and in any case may change as technology develops) there is no reason to ban bad images. The second is a statement of personal preference that is not backed up by any reasoning (other than "AI images are bad").

  I'm not sure what you are trying to say with your final sentence - when selecting images editors should (hopefully obviously) always choose the best image, with "best" depending on multiple factors relating to the specific usage. If we want to change that to "editors should always choose the best image, unless the best image is AI generated when you must use an inferior one" then we need a reason why. And nobody has yet articulated any reason for doing so (assuming we can legally use the images) Thryduulf (talk) 11:59, 2 March 2024 (UTC)

  You have not read what I wrote. Your paraphrase is ludicrous. I addressed the dangers of made up history and information about living people. I addressed that we need sources, for the reasons we always need sources, so others can check, and I addressed the standard preference for types of image making that is already well known and already used to convey reality. And I addressed being upfront with the reader that the image is AI, because we should always be upfront with the reader. Your imagining of some better AI picture, is just imagining and practically meaningless - better than reality, is not reality. -- Alanscottwalker (talk) 12:14, 2 March 2024 (UTC)

  nobody has yet articulated any reason for doing so (assuming we can legally use the images) Without weighing in on Alanscottwalker's reasons, I would point out that in this overall thread, there have been articulations of reasons. For example, the possibility of Wikimedia Commons becoming some kind of repository of images generated using programs that can and do, sometimes by being given such an input and sometimes just doing so without such, imitate and ape the oeuvres of living human artists, troubles me. There are ethical dimensions to the use and creation of diffusion engine-generated images, beyond strict copyrightability. To the extent that Wikipedia exists in the world, and in the social context of humanity, wanting our project and community to be ethically responsible isn't a non-reason. It's a reason, that one can agree with or disagree with (whether on its premise, on its execution, on some other aspect of it, etc.).

  talk

  ) 14:42, 2 March 2024 (UTC)

• I would argue regardless of what happens here there should be some exceptions. Fox example, assuming there is not a change in copyright status the image used in Artificial intelligence art shouldn’t be removed due to being AI generated since that wouldn’t make any sense. I believe this should also apply to notable programs such as DALL-E where I see it making sense that the article shows images that it created and is clearly identified as AI art.--67.70.103.36 (talk) 13:39, 5 March 2024 (UTC)

Of interest to this discussion may be one that was had a few months ago at the NOR noticeboard, Cartoon portraits. It was primarily about some very ugly and amateurish caricatures being used to adorn BLP articles, but also touches upon AI generated art and the questionable output of a project called "Wiki Unseen". Zaathras (talk) 00:33, 6 March 2024 (UTC)

It was primarily about some very ugly and amateurish caricatures Zarathas, you were called out in that discussion, multiple times, for that completely inappropriate characterisation of the images being discussed. It is still inappropriate now. The conduct of several editors in that discussion, including you, means that reading most of it was a complete waste of time - if you import that sort of behaviour to this discussion expect a swift trip to ANI. In as much as there was consensus for anything it was that "some illustrations are original research, others aren't" and "there are instances when illustrations are appropriate in an article, and instances when they are not" with individual images needing individual discussion. Thryduulf (talk) 02:00, 6 March 2024 (UTC)

Is that the one in which one of the opponents kept struggling to tell the difference between an

Picasso would have been "an amateur", as well as the whole Expressionism art movement). Or was there another depressing discussion? WhatamIdoing (talk

) 02:11, 6 March 2024 (UTC)

Thryduulf, good lord, enough with the sanctimoniousness. First the micronation debate above and now this. One of the primary movers of that debate turned out to be a now-blocked sockpuppet, and their edits reversed. My position was squarely in the right. Zaathras (talk) 03:47, 6 March 2024 (UTC)

I'm not sure which "position" you are referring to, but the NOR debate certainly did not come to the consensus that the view you were espousing was the sole correct one. The micronation discussion is completely irrelevant here, regardless of whether you needlessly personify discussions or not. Thryduulf (talk) 04:23, 6 March 2024 (UTC)

regardless of whether you needlessly personify discussions or not. Pot, meet kettle. Now, stick to the topic at hand, please. Zaathras (talk) 05:27, 6 March 2024 (UTC)

That reminds me of the kind of technical drawings used for anatomy classes. Even though the word has millions of photographs of various body parts, Gray's Anatomy and similar works are still valued. They're valued precisely because they're not specific photographs of individual – and therefore unique – specimens. They're valued because they show "typical" or "average" in a way that is more typical and average than any individual's body will ever be. Once you learn "average", it's easy to adjust to the "specific". In this area, a "made-up" drawing is a lot more valuable than a "real" photograph.

I suspect that there are instances in which AI-generated images would be useful and non-misleading. I could image, e.g., someone saying "Computer, please look at the images in c:Category:Apples. Now make me a few images that show a dozen different kinds of apples on a plain white background." You'd look at the resulting images and decide which, if any of them, had the Pertinence and encyclopedic nature that you were looking for. The fact that you used AI instead of Photoshop would be immaterial. WhatamIdoing (talk) 02:03, 6 March 2024 (UTC)

Yet my sense is that photographs of actual apples would be consistently more pertinent and encyclopedic. Even granting the varieties of apple kinds, and the variegations of lighting, etc., it can at least be agreed more consistently that the human-generated image of an apple does depict, and is intended to depict, an apple.

talk

) 05:32, 6 March 2024 (UTC)

AI-generated image from the prompt "apple"

Human-generated photograph of an apple

Human-generated drawing of an apple

human-generated picture of an apple

it can at least be agreed more consistently that the human-generated image of an apple does depict, and is intended to depict, an apple I'm not sure that is true. If I showed you 100 images that were generated by AI tools with the prompt "apple", all of them will be intended to depict an apple. If I showed you 100 images generated by humans of things that look to me like apples, I cannot say that all of them were intended to depict an apple - some may have been intended as representations of a quince, pear, cake, or something completely different. Whether something does depict what it is intended to depict, and if so whether it is better or worse (by whatever metric you choose) than some other depiction of that same thing is something that cannot be judged based on whether a human or AI generated either image. In the example photos the top image is more suitable as a lead image for the "Apple" article than any of the human-generated ones. Obviously we have no shortage of images of apples, but this is not necessarily going to be the case for every subject. Thryduulf (talk) 06:25, 6 March 2024 (UTC)

I rather would have preferred you hadn't, to try to prove a point to me, uploaded an image of an apple generated using a program trained on copyrighted works and which may itself contain major copyrightable elements of works. I don't think we as a community do right by ourselves or others by propagating material that doesn't respect the copyrightable works of others.

talk

) 06:38, 6 March 2024 (UTC)

Disliking AI and/or it's methods for those reasons is not an unreasonable point of view. However, as Wikipedians we are in the business of dealing with the neutral point of view and so our personal feelings are not relevant to whether AI images are or are not suitable for an encyclopaedia. Thryduulf (talk) 12:08, 6 March 2024 (UTC)

talk

) 03:58, 7 March 2024 (UTC)

Yes, but I think you'll find it's a bit more complicated than that. Human artists learn from studying and copying other artwork. A very typical assignment in university-level art classes is to look at all of the paintings by a great master, and then make their own versions. One of the main points is to have the students mimic the original style.

Is it ethical? AIUI scholars say it's generally ethical, within some broad limits (e.g., you don't want students producing copies so exact that they could pass for the original).

Conclusion: If humans can look at artwork and produce a new version (modulo forgeries, copyvios, etc.), it's ethical.

But: If a machine "looks" at artwork and produces a new version, even within tthe same limits, it's automatically, inherently, unreservedly, obviously not ethical?

Maybe it's not so clear cut as that. Maybe it should be okay for a machine to do what's okay for a human to do.

See also Tribute act. WhatamIdoing (talk) 04:17, 7 March 2024 (UTC)

• If we are abandoning the legal issue, or the factual dimension of whether or not it's possible to copyright the idea that an apple is red (?) and dealing entirely with issues of ethics:
  Prior to the last hundred years (give or take), the human race spent somewhere around a hundred thousand years without anything resembling modern copyright law. Even when such laws were introduced, the original terms of these protections were closer to reasonable (i.e. 14 years in the US, 14-21 years in the UK).
  Here is a scenario for you: Bonzo's Montreux is a drum solo recorded by John Bonham 48 years ago. John Bonham died 44 years ago. My dad listened to Led Zeppelin a lot when I was a kid. Nonetheless, its copyright is owned by Atlantic Records, a corporation which Led Zeppelin (of whom John Bonham was a member) signed a contract with some time in the 1960s, which means it is illegal for me to make a movie that includes me hitting a trash can with a stick if it has the same rhythm as this, unless I give a large amount of money to this corporation. It will continue to be illegal until the year 2060. How is it illegal? Laws are generally enforced by agencies that are funded with taxpayer money, so this is being paid for by everyone.
  Is it ethical for taxpayer money (i.e. the same pool of money we use to feed the hungry, clothe the sick, put out fires, and the like) to be used to enforce laws prohibiting me from humming the melodies of songs I heard as a young boy? What proportion of our childhood memories are illegal to reproduce without paying up to a million dollars? For a century?
  Of course, these things aren't really relevant to Wikipedia policy, but if you want to talk about ethics per se, I think copyright law is close to the single most morally odious institution in modern society. It seems sub-optimal to defend amoral bazillion-dollar conglomerates on the basis that their interests are aligned in a vaguely separate direction to different amoral bazillion-dollar conglomerates.
  As the esteemed Doctorow says in a voice more eloquent than my own: "Under these conditions, giving a creator more copyright is like giving a bullied schoolkid extra lunch money. It doesn't matter how much lunch money you give that kid – the bullies will take it all, and the kid will still go hungry (that's still true even if the bullies spend some of that stolen lunch money on a PR campaign urging us all to think of the hungry children and give them even more lunch money)". jp×g🗯️ 18:32, 10 March 2024 (UTC)

  If we're talking purely ethically, I do think copyright terms should be substantially shorter and shouldn't be able to be held by non-human incorporated entities for so long. But I don't think rolling over to diffusion engine-powered corporate interests at the expense of independent artists is the best way to stick it to Disney et al.

  talk

  ) 19:18, 10 March 2024 (UTC)

@

P-Makoto, I don't think it matters if the image actually depicts the subject (Is It Cake?). https://www.youtube.com/watch?v=vTJKosHxYtM

looks like an apple.

The requirement is that illustrations in an article look like they depict the subject. Thryduulf's AI-generated image "looks like" a bit more Candy apple to me, so I wouldn't use it in Apple. This isn't because it's AI-generated; it's because it doesn't look like an ordinary apple. Even if it were real and came with a certificate of authenticity, I wouldn't use that image in Apple.

My point was that, depending on the needs of the article, I might prefer a drawing like File:PSM V05 D153 Section of an apple.jpg over a photo of a particular apple, or an AI-generated simulation of such a photo. Being "real" is not inherently an advantage.

(The intention is completely irrelevant; we crop images to show objects that were only incidentally present in the image all the time, and nobody minds in the least that these objects weren't intended to be in the photographs.) WhatamIdoing (talk) 02:12, 7 March 2024 (UTC)

That's no different than saying that is a terrible portrait of an apple, why because it does not look like an an apple, what's the reason for that, because of how it was produced. Alanscottwalker (talk) 19:20, 9 March 2024 (UTC)

How it was produced is completely irrelevant - either something is the best illustration for a given bit of encyclopaedic content or it isn't. A "terrible portrait" is terrible regardless of whether it was produced by an AI, a three-year-old human or a professional artist. Thryduulf (talk) 19:24, 9 March 2024 (UTC)

No it's not, it looks terrible because of how it was produced. -- Alanscottwalker (talk) 19:28, 9 March 2024 (UTC)

You are entitled to that opinion, but that still doesn't make it relevant to this discussion. There are exactly three things that matter for English Wikipedia's purposes:

1. Can we use it the encyclopaedia? (i.e. is it freely licensed or meet the NFCC)
2. If so, does it illustrate what we want to illustrate?
3. If so, is it the best illustration for what we want to illustrate?

If the answer to any of those questions is "no" we do not use it. Why the answer is "no" is not relevant. Thryduulf (talk) 20:42, 9 March 2024 (UTC)

No. Why the answer is no (or yes) is always relevant, it's how editors actually reason through and explain the decisions they make. (Also, how something is produced is invariably significant to how it appears, so no use in quibbling that.) -- Alanscottwalker (talk) 10:14, 10 March 2024 (UTC)

There is some disagreement between us then, because I think it'd be misleading to knowingly, for example, replace the images in

talk

) 19:21, 10 March 2024 (UTC)

I would be supportive of a full ban on AI-generated images outside of contexts where AI is the subject of the article. It has already become difficult to get Google to give you a real image of something because of all the AI-generated crap polluting search results. Wikipedia should be a place you can come to to get a real image of a thing that actually exists, not what a computer thinks something looks like. Just because an AI-generated image qualitatively looks accurate does not mean it actually is, and we should not be introducing factual errors to our articles just because an editor thinks one image looks nicer than another one. We run into

OR issues if we're then trusting editors to make judgements about whether an AI-generated image is accurate or not. Sam Walton (talk

) 17:19, 12 March 2024 (UTC)

Probably we could cut out a lot of the confusion if we stopped saying "AI" images are good, or bad, or whatever; this can refer to any number of hundreds of models with output ranging from random splotches of color to genuine photorealism. Some models always create slop, and some do not. Here is one from MidJourney v6, with prompt included. jp×g🗯️ 19:34, 10 March 2024 (UTC)

Probably we could cut out a lot of the confusion if we stopped saying "AI" images are good, or bad, or whatever this has been my point. Some AI images are bad. Some AI images are good. We should always use whatever the best image is, regardless of whether it's AI or human-generated. Thryduulf (talk) 19:52, 10 March 2024 (UTC)

Why am I getting flashbacks to all the pseudoscience arguments about biotechnology and genetically modified foods, where anti-science proponents consistently tried to claim that biotech methods were relevant (and somehow harmful) in the resulting organism and needed to be explicitly noted, despite that not actually changing the resulting phenotypic or, often, genotypic resultant organism? In a similar manner, the source of the images here should be irrelevant (so long as we don't have legal copyright issues involved, which includes the human-made images), since it's the outcome that matters. I.e. what the resulting image is and whether it is visually representative of the subject it is intended to be used for. Silverseren^C 19:57, 10 March 2024 (UTC)

That's exactly the problem with AI images, though. Sure, let's say they're fine on copyright, I agree that isn't a major concern. It's the outcome that matters, and AI images simply aren't guaranteed to be representative or accurate. The fact that AI images are often close makes the problem worse, not better - it means that it's easy for problems to slip through undetected, especially to an untrained eye. Suppose we use an AI to create an image of a specific cultivar of apple that doesn't have an image currently. The AI spits out something apple-looking. Is this usable? Is it representative of this cultivar you prompted it for? How would you tell? A photograph, or a painting prepared by experts (e.g. Winter Banana (apple)'s picture from the Department of Agriculture), won't have such surprises (assuming the photographer hasn't misidentified what they're taking a picture of, of course). Even if we charitably assume that 90% of our pictures of specific apple cultivars are perfect, the damage from the 10% that aren't are not worth the minor advantage from using the AI pictures in the other cases. Just... get real pictures, instead. SnowFire (talk) 18:46, 11 March 2024 (UTC)

Indeed, this quasi-demand for blind faith in what it produces is the opposite of a scientific approach, one that can be verified and if one needs to replicate it to check, how is that done. (Besides one can't imagine the scientific mind would not be interested in what and why produces the best result.) Alanscottwalker (talk) 18:58, 11 March 2024 (UTC)

This is the exactly wrong usecase for AI images. The two places they are potentially useful is generic images (e.g. "Apple" rather than a specific cultivar of apple) and situations where artists' impressions are currently useful. If say SpaceX released am image of a proposed Lunar settlement then whether we used that image or not would depend on exactly the same considerations (copyright, notability, whether it illustrates our article in a useful way) regardless of whether that image was generated by a human or an AI. Thryduulf (talk) 19:43, 11 March 2024 (UTC)

Oh, so there are definable areas of no use. You should have begun with that. Alanscottwalker (talk) 21:27, 11 March 2024 (UTC)

There are no "definable areas of no use" just areas where, with the current state of the technology, AI images are always going to be inferior to human images (it's not impossible that will change in the future, but it wont in the current generation of models). As I've repeatedly said, there is no point in banning the use of AI images where they are (currently) inferior to human images for the same reason we don't ban MS Paint drawings from illustrating BLPs (although if Jim'll Paint It ever draws a freely-licensed self-portrait we might use it on that article). Thryduulf (talk) 00:44, 12 March 2024 (UTC)

You just said the opposite. "Wrong usecase", your words, absolutely means it is definable no use. As does, "The two places they are potentially useful is generic images (e.g. "Apple" rather than a specific cultivar of apple) and situations where artists' impressions are currently useful.") as that means they are not useful in other places. You're in bad faith.-- Alanscottwalker (talk) 12:31, 12 March 2024 (UTC)

I assure you that I am acting in good faith, but if you do not believe that then discussion is pointless as you will not listen to arguments different to your own, regardless of merit. Thryduulf (talk) 13:03, 12 March 2024 (UTC)

I listen. I just listened to you say two things that are directly contradictory, demonstrating bad faith. Alanscottwalker (talk) 13:36, 12 March 2024 (UTC)

They are neither contradictory (as explained) nor bad faith. Please stop making accusations and start engaging with the arguments. Thryduulf (talk) 14:00, 12 March 2024 (UTC)

Pointing out that you are saying directly contradictory things and that is evidence of bad faith is engaging with the substance of your arguments, it demonstrates the lack of substance in your arguments. Alanscottwalker (talk) 14:15, 12 March 2024 (UTC)

I have explained how they are not contradictory: Noting that there are areas where AI images are currently always inferior to human artists is not the same thing as asserting that there are defined or definable areas where we need to ban the use of AI. You can disagree with either or both portions, but that doesn't make them contradictory nor is it evidence of bad faith. Now, please either provide actual evidence of bad faith contribution or stop casting aspersions. Thryduulf (talk) 14:28, 12 March 2024 (UTC)

You are just further demonstrating your bad faith, saying they are always inferior means no use, no use is all ban means. Alanscottwalker (talk) 14:38, 12 March 2024 (UTC)

You are either not reading, not comprehending or acting in bad faith. Regardless of which it is further engagement is a waste of everybody's time. Thryduulf (talk) 15:38, 12 March 2024 (UTC)

Oh, I read, you described large categories of AI as being no use ("wrong usecase") meaning they are functionally banned, indeed you identify only two narrow categories as usable, meaning again the rest are functionally banned. Alanscottwalker (talk) 17:50, 12 March 2024 (UTC)

I agree with Thryduulf on the use cases where AI images could be useful. There are many abstract concepts where an illustration is the best option. In situations where current AI images are categorically inferior to human-created images, there's still no use in a ban, since the decision is often between an AI image and no image at all. I would not feel comfortable using one until the copyright issue is more settled, but I would not support a ban. Firefangledfeathers (talk / contribs) 15:13, 12 March 2024 (UTC)

So, you yet have another way to say AI use is limited. It is thus bizarre not to write it down clearly.

And even to begin with your limitation

1) How do you know its AI your dealing with how do you find out so you can live within your limitation. (is there a rule on disclosure)

2) What makes you think anyone agrees with your limitation for use, which is a very narrow one, abstract concepts? (Or is that abstract like, we don´t have a picture of yesterday's event but surely AI can abstract it from the news) Alanscottwalker (talk) 16:45, 20 March 2024 (UTC)

IMO, due to the reasons discussed above, they are rarely useful/appropriate for a Wikipedia article. But "seldom useful" doesn't mean ban. Perhaps we need a couple paragraphs saying this and describing the issues which were discussed above. North8000 (talk) 19:22, 11 March 2024 (UTC)

Unfortunately, the last time we tried to put together an actual policy proposal for generative AI (

WP:LLM) it took over a year and went nowhere beyond producing an essay. My reading of what happened back then was that the policy (which was already extremely restrictive, as it should have been) was shot down by people who believed it was still too permissive and who wanted basically a total ban; I feel this position was obviously self-defeating, since they weren't able to get a consensus for the ban they wanted and shooting it down resulted in no overarching LLM-specific policies or restrictions whatsoever, even though there were plenty of restrictions almost everyone agreed on. Discussion over it as it relates to images is likely to go down the same course - just glancing at the discussions above makes it clear there's no consensus for a total ban; but there's also enough people who won't accept anything short of a total ban that we're likely to end up with nothing at all. --Aquillion (talk

) 01:11, 12 March 2024 (UTC)

I don't see that as an issue of AI vs not-AI. If the image is an accurate representation then (assuming copyright, etc. is OK) whether we use it or not should be a matter of editorial judgement at the article concerned. If the image is not an accurate representation then we obviously should not use it, but only because it's not an accurate representation not because it's AI generated. Thryduulf (talk) 03:09, 12 March 2024 (UTC)

Yes, of course, any reasonable policy would ban them from BLP's and in recreating recent human and natural events. Indeed it is asserted above that the "two places they are potentially useful is generic images (e.g. "Apple" rather than a specific cultivar of apple) and situations where artists' impressions are currently useful", so obviously that's the place to start and we should narrow it more, as need be. -- Alanscottwalker (talk) 12:44, 12 March 2024 (UTC)

Let's be pragmatic here. If machine-generated images are the best available illustrations, use them. If human-generated images are the best available illustrations, use them. If none of the available images are suitable, don't use any. If people try to add images that aren't suitable, revert and explain why they're not suitable. My view here is the same as my view on the proposed LLM policy – the issues raised are all already covered by existing policies, guidelines and common sense. – Teratix ₵ 09:05, 12 March 2024 (UTC)

• While we prefer photographs, we DO allow paintings/drawings (even user generated paintings/drawings) of our subjects when a free licensed photograph is not available. Sometimes we even have to choose between several paintings/drawings, based on which is most realistic and accurate.

We can do the same with AI generated images. We can compare the image to non-free photographs, (and to hand made paintings/drawings) to decide whether it is accurate enough for our needs… replacing it with a photograph (or a more accurate painted/drawn image) if one becomes available.

In other words… there is no need to change policy… Continue to prefer photographic images, but also continue to allow other forms of imagery (including AI) when no photo is available. Go case by case, choosing the most accurate image available, and be willing to change images if something better becomes available. Blueboar (talk) 14:36, 12 March 2024 (UTC)

Not only did you just suggest a restriction: 'only where non-free content comparisons can be made by all editors and readers, presumably on a lay basis', but there is no way to have such a restriction without putting it in our policy or guideline. Alanscottwalker (talk) 14:51, 12 March 2024 (UTC)

Eh? They literally said to do what we currently do. That means either no change to policies/guidelines are required to do this with AI or we need to change the policies/guidelines to allow it to be done for non-AI images. Thryduulf (talk) 15:37, 12 March 2024 (UTC)

They literally described a restrictive process one with the ability to check the AI image, so it would be Kafkaesque to not write the rule down (its not a secret process). Alanscottwalker (talk) 15:44, 12 March 2024 (UTC)

When someone says there is no need to change policy and that we should continue to allow other forms of imagery (including AI), we should take them at their word – they're not proposing alterations to existing practices. – Teratix ₵ 16:58, 12 March 2024 (UTC)

Their word was to describe a restrictive process for checking AI images, the suggestion of not, or resistance to, writing it down is Kafkaesque. Alanscottwalker (talk) 17:05, 12 March 2024 (UTC)

That is a ridiculous and tendentious way to interpret their point. – Teratix ₵ 17:14, 12 March 2024 (UTC)

Don't be absurd, "compare the image to non-free photographs" is as plain as day process as there is.-- Alanscottwalker (talk) 17:18, 12 March 2024 (UTC)

My point was that we already do this with human created images (paintings and drawings)… so there is no need to change the policy to extend this to AI created images. Blueboar (talk) 13:54, 20 March 2024 (UTC)

What? Having chosen many images for articles, it NOT in the least usual to try to compare photographs with paintings and drawings to pick out illustrations. Moreover to even do the unusual and very specific process you describe, you already 1) state a preference for photographs which would need to be written into policy and 2) require that comparable photographs already exist. None of these issues can be assumed as you do. Alanscottwalker (talk) 14:11, 20 March 2024 (UTC)

I have been in discussions about user drawn images as well… and it isn’t at all unusual to compare these images to photographs to see which looks more accurate. That is part of the give and take we do to reach consensus over which images to use. We don’t need rules to say it is OK to do this… we just DO it. I wasn’t outlining a process (indeed, I don’t think we NEED a process). Blueboar (talk) 15:42, 20 March 2024 (UTC)

You clearly did outline a process. And your now odd claim we don´t need one is a complete non-sequitur.

The process you outlined is:

1) Preference for photographs (when some people above say there is no such preference);

2) We have to know that we are dealing with an AI drawing, which means it has to be disclosed (when some above say it does not have to be disclosed);

3) the drawing can only depict something that can be photographed (otherwise there is nothing to compare).

4) Photographs of that exact subject have to exist. (when some above suggest, AI is ONLY useful when you CANNOT picture it)

5) the comparison has to be made (and you have a standard for your comparison, otherwise there is no discussion)

There is no way that is not a process with very specific requirements. Alanscottwalker (talk) 16:24, 20 March 2024 (UTC)

• Agreed with Blueboar, worth mentioning this use case but a change isn't needed. – SJ + 17:15, 12 March 2024 (UTC)

Meh… my comments were not intended to be a "process”. The only “process” is to hold a discussion on the article talk page and reach a consensus on whether to use the image in question.

There are many things that can influence this consensus, including (but not limited to) “does the image realistically depict the subject?”… and a great way to determine that is to compare it to photos. Call it a consideration rather than a process.

In any case… none of this needs to be spelled out in policy. An AI generated image is simply another form of art… one drawn by a computer rather than by a human. We would judge the appropriateness of using it in exactly the same way we would an artistic image created by a human. Whether you want to call that judgement a “process” or simply “things to consider while reaching consensus”… we already do this so there is no need to change the policy. Blueboar (talk) 16:57, 20 March 2024 (UTC)

You have not even said how you are to know you are dealing with AI art to begin with. So, no you are not speaking for collective consideration, you are dealing in the beginning of a mystery (and not the fun kind, the kind we know can and have been used to be actively harmful), which then extends to what is the basis for the image, if it is AI. Collective consideration, at the least, means being up-front and answering questions, and it remains Kafkaesque to not write down the collective considerations. Alanscottwalker (talk) Alanscottwalker (talk) 17:10, 20 March 2024 (UTC)

Images, Labeling, restrictions and concerns for living people and events

Youtube has just come forward with rules about AI, especially in labeling, and with a special concern for representations of living people and events.[12] There are concerns whether Youtube goes far enough (interestingly with a comparison to handling copyright). [13] As I suggested awhile ago, the least we should do is be upfront with readers about it, and have special concern for living people and event depiction. -- Alanscottwalker (talk) 12:34, 20 March 2024 (UTC)

Anyone here feel like we're moving towards a time where we require all users to post a

This is pure VP speculation. I'm just taking the temperature of the community. jps (talk) 21:48, 16 March 2024 (UTC)

No. It would be an utterly unnecessary invasion of privacy for folks to declare every possible conflict they would have if they were to edit every article. We only require a conflict statement when one is involved in conflicted editing, and that is best placed in association with the edit and the page, not the user. -- Nat Gertler (talk) 21:55, 16 March 2024 (UTC)

I completely agree with NatGertler, you should only disclose a COI if you edit or intend to edit the associated topic. The best practice as I understand it is when in doubt to avoid the topic, there are after all more pages on wikipedia than one could edit in a lifetime outside of one's significant COIs. Horse Eye's Back (talk) 22:32, 16 March 2024 (UTC)

I have wondered recently how strictly COI will be interpreted. I have declared (informally) when I have edited articles about people related to me (my first wife and a first cousin, once removed). But, should I have declared a COI when I edited articles about the high school I attended, the high school my son attended, the high school where my wife worked, the professor who was my advisor when I was a sophomore (and who later published an article I wrote in a journal he edited), the base where I was stationed when I was in the Army, the places I have lived, or the places I have visited? Donald Albury 23:40, 16 March 2024 (UTC)

I would see those as potential biases rather than conflicts of interest, and whether to disclose would basically come down to your judgement on whether they arise to actual biases. – Teratix ₵ 03:48, 17 March 2024 (UTC)

Yes to all but the last two, in my opinion. Levivich (talk) 16:23, 22 March 2024 (UTC)

Given how much we struggle to recruit and retain new editors, I’m deeply concerned about well-intentioned policies having broad chilling effects. This website is intimidating enough! Ghosts of Europa (talk) 04:39, 17 March 2024 (UTC)

You reminded me of this thread: Talk:Donald_Trump/Archive_38#conflict_of_interest. Gråbergs Gråa Sång (talk) 08:58, 17 March 2024 (UTC)

After seeing one of the recent acrimonious discussions (with opaque history due to large amounts of oversighting for "potential outing"), I'm feeling like Donald Albury and what Gråbergs Gråa Sång linked on this one. Although my feeling is more "if someone takes a disliking to you and they can dredge up any tenuous connection to use COI as a bludgeon..." Anomie⚔ 13:11, 17 March 2024 (UTC)

Come on, its not difficult to not write about your employer, etc. but if you do, the world deserves your disclosure (read carefully the guideline's sections on regulations/laws surrounding keeping your reader in the dark on your COI -- at least give the reader a chance to know, its only fair). -- Alanscottwalker (talk) 21:22, 20 March 2024 (UTC)

Hello, I wanted to know that the number of national games should only be counted by the number of games played by the player in Fifadi or all the games outside and inside Fifadi are considered for Wikipedia. Ab10sport (talk) 05:10, 23 March 2024 (UTC)

answered on your talk page. Nthep (talk) 17:25, 23 March 2024 (UTC)

Hello friends, I am writing a very detailed and in-depth article over “People’s Publishing House (India)”.

I will quote many sources, including famous authors, newspapers, journals and others. But for a two line part of my intro, I need to quote the interview of the Manager of PPH Connaught Place Showroom. I took this audio based interview in the early days of March 2024.

i know, self published work is strictly prohibited on Wikipedia. But I need to quote the part of his interview where he is saying that they used to distribute books on Mobile Distribution Vans in many parts of India.

The interview is in Hindi, and I published it on YouTube.

Please suggest whether to quote Manager of PPH from this interview or not? Pallav.journo (talk) 16:28, 22 March 2024 (UTC)

You may get a faster and more comprehensive response by inquiring at

WP:RS/N, if you'll forgive the alphabet soup. Good luck. Wehwalt (talk

) 16:48, 22 March 2024 (UTC)

WP:SELFSOURCE might cover this as well, but once again it would depend on how it's written. "PPH, has revolutionised the way India reads" I would suggest a statement like this might need to be sourced as well. YellowStahh (talk

) 20:28, 22 March 2024 (UTC)

While not all self-published sources are banned for all uses, if the manager is still living, then use of this source is prohibited, per

WP:BLPSPS. -- Nat Gertler (talk

) 00:20, 23 March 2024 (UTC)

Nat, I'm not sure that's true. I think this might fall into the category of "living persons who publish material about themselves, such as through press releases" – or interviews. RSN's the right place to ask, though. WhatamIdoing (talk) 23:06, 23 March 2024 (UTC)

• Please post at
  WP:RS/N instead. This is for policy discussion. --David Tornheim (talk
  ) 00:36, 23 March 2024 (UTC)

This image specifically [14] I feel is too bright / overexposed, so I was wondering if I could edit it to correct that.

Thanks, IOHANNVSVERVS (talk) 23:35, 24 March 2024 (UTC)

@IOHANNVSVERVS In this case probably, since the image is pd anyway. And edited already. However, there's probably do:s and don't:s, try asking for guidance at Commons:Help desk. Gråbergs Gråa Sång (talk) 07:30, 25 March 2024 (UTC)

When it comes to encyclopedic use of photos, the overall general principle is that we should try to capture reality as closely as possible. Things like correcting an exposure help with that and are fine, I'd say. Something like boosting saturation to artificially make a photo seem more colorful than it actually was would not be OK. In between those poles, I'd say it's alright to use colorized images, but only if the caption clearly notes that the image has been modified. (A related task suggestion: Someone could compile a list of articles that use photos in commons:Category:Colorized photographs or subcats and add disclosures to the captions where needed.) Sdkb ^talk 19:00, 25 March 2024 (UTC)

Thanks, I figured that as long as no substantial changes were made it would be fine. I fixed the exposure and improved the definition. IOHANNVSVERVS (talk) 19:12, 25 March 2024 (UTC)

Can be seen here [15]

IOHANNVSVERVS (talk) 19:14, 25 March 2024 (UTC)