Template talk:Expand language

improving template

@Mathglot: pinging you at your request. My earlier post: " Right now the link for corresponding article has a switch so that it calls Wikidata information where the article is not specified. I would like the same behavior to happen for the creation of the machine translation link at translate.google.com." Basically, if there is not an article name specified, I want there to be a Google translate link generated based on the title in the foreign language at wikidata. (Currently, if no title is specified, no machine translation link is added.) This behavior already exists elsewhere in the template, just trying to add it for the machine translation as well. Calliopejen1 (talk) 13:34, 16 September 2020 (UTC)

@Mathglot: thoughts? Calliopejen1 (talk) 23:00, 25 September 2020 (UTC)

Calliopejen1, this might need a feature that is not yet available in order to implement it. I got a response at Module talk:Wd#Titles from other wikipedias to a question I posed there, regarding a subtask that I believe would be necessary in order to implement your request. I'm still looking around for other methods that might work; perhaps Thayts will have an idea. Mathglot (talk) 05:45, 27 April 2021 (UTC)

That one went stale, but I re-upped with this discussion. Adding User:Calliopejen1. Mathglot (talk) 04:50, 11 September 2022 (UTC)

N'Ko language

After a speedy move to Category:Articles needing translation from N'Ko Wikipedia, the article Madina-Oula using {{expand language}} with parameter |langcode=nqo is still looking for the old category name with diacritic Category:Articles needing translation from N’Ko Wikipedia, see [1].

{{Expand N'Ko}} also exists but is not currently used.

Code nqo or N'Ko language are not listed at meta:Table_of_Wikimedia_projects#Projects_per_language_codes.

@Pppery: are you able to assist, please? – Fayenatic London 12:46, 5 February 2024 (UTC)

 Fixed * Pppery * _{it has begun...} 15:39, 5 February 2024 (UTC)

Thank you, that resolves the immediate need. But if somebody uses {{expand language}} again (with |langcode=nqo) rather than {{Expand N'Ko}}, will it still be looking for the old category? – Fayenatic London 15:57, 5 February 2024 (UTC)

Yes, in theory, and Slovenian has the same problem. I think it's better that that bridge be crossed when it happens, though. * Pppery * _{it has begun...} 16:19, 5 February 2024 (UTC)

Multiple topic categories

A lot of articles fit into multiple topic categories. For example:

• Articles about athletes could fall into "bio" and "sport".
• Articles about elections could fall into "gov" and "hist".
• Articles about natural disasters could fall into "hist" and "sci".

Editors have different preferences for which articles belong where, and so it is very confusing to people trying to go through the categories. Numberguy6 (talk) 20:42, 13 February 2024 (UTC)

Template-protected edit request on 13 September 2024

When there are multiple languages, the template says "You can help expand this article with text translated from the corresponding article in...". This is wrong, because it refers to multiple articles. In the section of the template code for when more than one language is specified, please replace "the corresponding article" with "the corresponding articles". TTWIDEE (talk) 13:16, 13 September 2024 (UTC)

 Done @TTWIDEE: --Ahecht (TALK
PAGE) 14:50, 13 September 2024 (UTC)

Guidance for when to remove this template

I would like to use the "Expand" templates (in various languages) to help identify articles that could benefit from the help of a translator, in the context of our work at the non-profit OKA. However, I find it challenging because a large share of articles tagged with this template do not really need to be expanded.

To make this template truly useful, I think it would be better if it was only used when the corresponding other-language article is substantially better than the English article. When the corresponding article is only marginally better, or only better in very selective parts (e.g., contains one section that the EN article misses, despite the EN article being better overall), using this template does more harm than good in my opinion as it dilutes the signal (since almost all Wikipedia articles will have at least one other language for which the article is at least marginally better).

I would suggest putting the bar high. For example, this EN article has less details than this corresponding AR article. However, the EN article is already good enough so it is not clear that there would really be much value in expanding it, the AR article is under-sourced thus would probably not meet the quality requirements from EN Wiki, and the template has been here since 2008 so it is extremely unlikely that someone will work on this since it hasn't happened in the last 16 years. I think that, for articles such as this one, the template should be removed.

Is there any general guidance as to when it is considered ok to delete this template? If we can align on clear criteria, I could task the translators in my non-profit to go through the full backlog and remove it when required.

@Mathglot @Piotrus FYI as I suspect you may have an opinion on this 7804j (talk) 19:11, 15 September 2024 (UTC)

Yes, I would support this in principle, but there is also WP:CREEP to be considered. The "guidance" offered in most templates is often nothing more than the opinion of the editor who wrote the documentation for the template, subject to the changes to the doc made by other editors, based on their opinions, so you could say that the guidance is based on consensus, sometimes of a pretty low level, and often most based on WP:SILENCE. (That said, a few templates get a lot of attention, like {{Talk header}}, {{Infobox}}, the structure of the User warning or Welcome templates for example, and the design and documentation of those templates have correspondingly higher levels of consensus.) I would say that {{Expand language}} (and its child templates {{Expand German}}, {{Expand Catalan}}) have a reasonably high level of visibility, and are not very contentious (this Tfd ended with a strong Keep); where there is some contention is about whether to place the template at the top, or bottom of articles (e.g., here, and also at the linked Tfd) but this has not resulted in any change in present behavior.

As far as placing the bar high, I would agree with that as well, but it is still quite subjective, and difficult to codify. Some templates are scrutinized more than others; for example, some editors will place a {{POV}} or {{Disputed}} template on top of an article purely because they disagree with the content, and sometimes when they have raised objections on the Talk page that have been roundly rejected, but that is not a legitimate use of those templates, and they may be removed per WP:WTRMT; having a template that impugns the content of an article based on one person's opinion is generally seen as disruptive. I don't think {{Expand language}} has that level of visibility, and it doesn't argue that the article is deficient in any way, so has a lower bar for inclusion: there is not so much objection to "one editor thinks this article could be improved by expanding it with translations from the Spanish article" as there is to "one editor thinks the content of this article is in dispute".

One possible problem with the proposal is WP:CREEP. Often there is objection to adding more guidance in writing, where common sense or consensus should prevail. So, while I agree with your proposal of how it *ought* to be used, I'm not sure where I stand on the question of whether that should be codified by additional language in the template, and will be interested in what other feedback you get about it. I can see pluses and minuses on both sides.

Regarding your point about the foreign article just being better in one section, see param |section= of the template; in those cases the template should be moved to the section in question, and if it is absent, just add the header by itself, and place the template under it with |section=yes. Finally, there is the issue of whether you want paid OKA editors to be spending time on this, rather than translating articles; I think the goal of removing unneeded templates is a good one, so this is mostly a question of prioritization of editor time, and in this case, OKA funds as well; not something that I can or should have any input on. Mathglot (talk) 20:19, 15 September 2024 (UTC)

One other solution for a similar situation: when you have an empty article section for which you wish to highlight articles in other languages, you could place a {{Further ill}} template instead, as in Principle of legality in criminal law#Italy, Armed Forces of Belarus#CSTO, or Steam railcar#Sweden. Mathglot (talk) 00:30, 16 September 2024 (UTC)

I mostly agree; the issue is that articles are expanded sometimes but the template is not removed. And yes, sometimes it is added when the differences are not major. But like with all other templates, there is no manpower to verify them. If you want to use this or anything other for any serious project... sadly, you have to verify each one yourself (or through some bot or metric, accepting some errors). For example I use some templates to select articles for my students for various activities, and sigh, we have a lot of false positives (ex. the 'requested image' template, used on talk, is very often left in place in articles that have images - this needs some major cleanup, sigh). _{Piotr Konieczny aka Prokonsul Piotrus| reply here} 05:36, 16 September 2024 (UTC)

Very true. Mathglot (talk) 06:43, 16 September 2024 (UTC)

Thanks both for the response!

Then I will start removing this template on a few pages such as the example I shared, adding a link to that discussion in the edit descriptions, and see if anyone objects to this approach.

Regarding use of translators time: I agree that it is better for them to focus on translating than fixing templates. However, I'm planning for OKA to be a very long-term project and to scale it even further, so I assume we will eventually run out of "Good content" to translate, so I want to explore other value-adding activities that translators can do on Wikipedia. Cleaning up these translation tags can help with:

1) Identifying real cases where the articles would benefit from being expanded

2) Removing noise, so that these maintenance tags can actually be used by the broader community (as Piotr mentioned, it's currently hard to use these tags as a signal, which is unfortunate). There's also the general concern that the community has of paid editors not helping with the "clean up" activities, so I'm hoping we can increasingly take a role in these maintenance tasks that volunteers don't typically like to do

(on a related note: if you're aware of important simple maintenance tasks that require a lot of manpower and that are neglected in Wikipedia, I'm also happy to look into it as a potential project for OKA, such as the "image requested" one mentioned by Piotr) 7804j (talk) 15:47, 18 September 2024 (UTC)

RfC: Reference verification

Should the sentence in the instructions to this template "If possible, verify the text with references provided in the foreign-language article" be reworded? Fangz (talk) 10:14, 4 December 2024 (UTC)

Essentially, current form of the instruction creates implication that the verification process is optional. This is not the case, as per WP:CIRC references taken from other wikis need to be confirmed before they are used, else inclusion of references that are impossible to verify violates WP:SAYWHEREYOUREADIT. While other wikis have their own policies on reliable sourcing, you cannot simply assume that the current revision of the article correctly implements such policies. Without verification essentially you are using the other wiki article as a source in itself, while disguising the fact that you are doing so, a practice that is very dangerous leading to the creation of fictious references. It would be preferable for material to be listed unsourced (possibly tagged cn) than to be listed with a reliable-looking but unverified source.

Replace with a separate bullet: "Foreign language Wikipedias are not considered reliable sources in themselves, so only include material if you can personally verify the references." Fangz (talk) 12:56, 5 December 2024 (UTC)

Include but with a maintenance tag.Benjamin (talk) 12:01, 4 December 2024 (UTC)

What maintenance tag would be appropriate here? Fangz (talk) 12:13, 4 December 2024 (UTC)

@Benjaminikuta (Summoned by bot) Do you mean this maintenance tag? I dream of horses (Hoofprints) (Neigh at me) 06:06, 5 December 2024 (UTC)

Should point to relevant policy, perhaps a separate bullet would be better: "Only translate text if you can personally verify the references; foreign language Wikipedias are not considered reliable sources."

I'd also consider pointing to WP:NOTSOURCE, is there a policy against citing essays in template instructions? Carleas (talk) 14:17, 4 December 2024 (UTC)

Good idea, I tweaked the links a bit to point to the policies I think are most appropriate. Fangz (talk) 12:57, 5 December 2024 (UTC)

I don't think we should try to raise requirements for translators. The rules for translating should be the same as the rules for a Wikipedia:Splitting or Wikipedia:Merging articles. In all three cases, editors should use their judgement to decide whether a given portion of a Wikipedia article is worth retaining, but there should not be a special rule that says translators have to do even more work than anyone else who is moving content from one page to another.

Everything Fangz says about translation applies to articles that started here and are being split or merged:

• It's possible to read WP:CIRC as meaning: Content from a Wikipedia article that you're translating, splitting, or merging is not considered reliable unless it is backed up by citing reliable sources. Confirm that these sources support the content when translating, splitting, or merging... – except that it doesn't say that and we never actually interpret it that way, because editors would revolt.
• It's possible read SAYWHEREYOUGOTIT as meaning that you have to read every single source when translating, splitting or merging – except that it doesn't say that, doesn't refer to Wikipedia's contents at all, and is focused specifically on cases of reading one source that cites a second/better source, and you actively pretend that you read the better source. If SAYWHEREYOUGOTIT were more blunt and direct, the example it gives would be this one, which is still irrelevant:

  John Smith (2009). Name of Book I Haven't Seen, Cambridge University Press, p. 99, cited in FunnyName (2024). "Funny Video I Watched", TikTok.

• Even here – especially here, since the English Wikipedia gets more WP:UPE and similarly dubious external contributions than any other wiki – you cannot simply assume that the current revision of the article correctly implements such policies, so this would apply at least as much to splitting and merging as it does to translating.
• Without verification essentially you are using the other wiki article as a source in itself, while disguising the fact that you are doing so whenever you split or merge an article. However, it won't be leading to the creation of fictious references, because a fictitious reference is one that doesn't exist ("Prof. I.M. Fake (2009), Name of Book I Made Up"). Copying a real source from one page – regardless of which wiki that's on, and regardless of whether the source WP:Directly supports the content it's associated with – to another page does not make the references fake.
• It would not be preferable for material to be listed unsourced. Actively hiding the original source interferes with identifying copyright violations, and replacing a source with {{citation needed}} is not the right way to deal with concerns like this anyway. If you genuinely believe that a specific source should be checked, then try {{verify source}}. Note that, exactly like "I WP:CHALLENGE every uncited sentence in Wikipedia" is considered block-worthy Wikipedia:Disruptive editing, indiscriminately tagging every source in a translated article will also be considered a behavioral problem.

WhatamIdoing (talk) 23:27, 14 December 2024 (UTC)

I would agree that the problem does apply to a smaller extent to the cases you mention, but that's not a reason to not try to do anything about it. I have just been through a traumatic process of removing false references and false information that have been propagated across a dozen articles and multiple wikis, and it's a hell of a pain in the arse. Editors should be encouraged to try and verify references whenever they can.

In this case the problem is worse because in the other case there's no template that recommends *not* verifying sources when splitting etc. Further, translation from foreign wikis creates additional problems, in that it often produces non verifiable references, from people who poorly understand the state of the original article (in the case that brought this up, the article in question was tagged with sourcing issues and with knowledge of the sources that were being referenced, the sources look likely to not verify, but there's no guarantee that there would be someone like me to look at the original article in future cases), and makes it difficult to track down the original page history. People also have a strong inclinations to use chatgpt style translations that can hallucinate references on their own. Meanwhile the articles we are talking about are typically quite obscure topics since otherwise usually good English language references would exist. It would be usually the case that splitting etc cannot INCREASE the number of problematic references in the English wiki. The risk reward balance is very different here.

The additional creation of fictious source in this case is by the Chinese whispers effect of retranslating text. Without reference to the original source you cannot verify if your translation accurately represents the original reference. The original untranslated article may be accurate but your translation may be entirely incorrect, and be impossible to ascertain without reference to the original, as there are many things that cannot directly translate. (For example, gender in many languages)

The whole problem here is the failure to verify sources. If that is the issue, letting people just translate without verification is giving them license to commit copyright violations. Providing information could be done with a specialised template or through html comments, but I really do not believe that known-unverified references should be given the same standing as other references. That's making them impossible to remove. Fangz (talk) 07:24, 15 December 2024 (UTC)

The problem with your proposal is that it goes well beyond Editors should be encouraged to try and verify references whenever they can to only include material if you can personally verify the references. This will get interpreted as a reason to delete good translations of accurate content because "the template says you should only include material if you personally verify it, and the translator said he didn't buy and read that peer-reviewed paper/expensive textbook/scholarly book, so he didn't personally verify the references, and therefore shouldn't have included the material".

There's a principle in policy writing that says you have to assume that whatever rules you write will be interpreted by someone who is mindlessly enforcing rules. (This is not mine; it goes back at least to James Madison in The Federalist Papers). What you have written is not hitting the right balance.

Things that might meet your actual goal, without creating a Thou Shalt Not kind of "rule" include:

• Many translators find that checking the cited sources helps them produce more accurate translations.
• Please omit any part of the source article that you think is dubious or low quality.
• Be skeptical of AI-based 'translations'. (This should have a link to a longer explanation.)

I don't think the problem you're seeing is unique to translation. A couple of years ago, I was tried to help resolve a dispute over how to describe excessive alcohol consumption. One article they fought over was about a fictional character. The article said "alcohol abuse" (an official diagnostic label from 1994 to 2013). The first editor changed it to "heavy alcohol use" (not a diagnostic label). The second editor dragged him to ANI over this and similar edits to other articles, claiming that he wasn't following the sources. However, the source cited at the end of that sentence didn't mention alcohol at all, and the only one in the article that mentioned alcohol, which was in German, said "[the] drunk". ANI stupidly supported the editor who claimed that the copyedit misrepresented the source; the end result is that we lost an editor who knew what he was talking about, and kept one who reverted thoughtful and accurate work without checking the sources. I tell this story to say that I understand and fully agree that it's useful to check the sources, but it's not a problem that it is unique to translation. It would be nice if editors checked the sources, but it's always nice, not just for translation. WhatamIdoing (talk) 19:54, 16 December 2024 (UTC)

A template is not policy. There's no reason to think this would lead people to delete stuff just because the template is present. A template suggests what people *should do*, so it should be stronger than a rule stating what not to do. If you want to serve well intentioned editors you should write templates so inexperienced people following them have *no possibility* of running into conflicts with editors who use strict interpretations of guidelines, not suggest that they put a lot of effort into stuff that will get them into a revert war. People who have enough confidence to risk that would do it on their own anyway.

But more importantly I do not think the text you propose is appropriate. The goal is not "accurate translations", the goal is accurate content on Wikipedia. The content on other wikis can assist in article writing but merely reproducing a revision is not desirable in any way. Content that is reworded to suit the needs of an English reader is far superior than even a perfect translation, and as a starting point to that it is better to look at the original wiki than re-edit a translated version.

You say that there are other cases but translation is a case where the issues we are talking about especially applies, because of the inherent creative aspect of translation. Consider:

1. Given the increasing prevalence of machine translation, the vast majority of cases where editors will use this template are cases where the editor is unfamiliar with the material/state of the original article and thus unable to evaluate accurately whether the content is dubious or not. Saying "be sceptical of machine translation" isn't going to help, unless you are willing to say "only do this if you are a human translator" (unlikely) people will just apply whatever level of judgement they are able to offer (and whatever degree of scepticism they place in AI output, which is frequently *zero* amongst the people who use these tools) no matter whether you include this warning or not.

2. Material this stuff is involved in are usually borderline with respect to notability and thus rarely content Wikipedia *has* to have. Selection of only those sections that are most reliable is generally the right call in this case.

3. A simple alternative exists of simply directing users to go look at (say, Google translated versions) of the foreign language Wikipedia themselves. This would be better up to date with history and talkpage intact. So translation has to offer some kind of service that makes it superior to this option, and verification is one possibility. If we are talking about articles with niche interest and niche sourcing in English, the likelihood of it being improved on the English wiki is much lower. So having translated, even well translated content on the English wiki does not necessarily present positive value to Wikipedia and readers because they are often better off going to the other language wiki.

4. Adding incorrect content that *appears* to have good sourcing is a HUGE problem and far worse than failing to include borderline notable unverifiable content. Your example is one possible outcome here. In my experience this is a major source of conflicts that pushes away subject matter experts that aren't experienced with sourcing. Once secondary sources are written based on Wikipedia articles then you have to deal with trying to identify whether the new English language sources are reliable or circular or not.

5. Note that we cannot vouch that acceptable references on foreign language wikis are acceptable on en. Guidelines are subject to the individual wiki's community, so it's entirely possible that a reference they use is acceptable by e.g. their RS rules but not acceptable by English wiki rules. This goes further to all content, really.

Given all these things it seems extremely clear to me that the text recommendations on this template ought to favour caution and rigour. It is very easy to be lazy with this stuff and very hard to catch and correct lazy errors, and the cost of waiting for someone else who is more capable to write the article is very low. We should encourage the mindset that the other wikis should be used as basically particularly well written Geocities pages that you can use as a starting point, than basically the same as splitting/merging. Translation simply is not that kind of process. Fangz (talk) 01:30, 17 December 2024 (UTC)

I agree that a template is not supposed to be setting rules, but I think that's a reason for it to not make up rules, rather than a reason for it to overstate the rules. Editors learn Wikipedia's rules by what you call "Chinese whispers" (telephone game), and if they see a template boldly asserting that something ought to be done, they will trust that the template is not making up rules or overstating the case just to make a point.

As for your other points:

1. I think editors are pretty good at evaluating whether content is plausible or dubious. Most people edit (including translations) within a reasonably familiar area. You don't need special knowledge to know whether something's completely outlandish – a person who is ten feet/three meters tall, or a car that never needs fuel. I suspect that your concerns are about content that isn't {{dubious}}, but is instead just slightly off.
2. I don't agree that this template is largely used on subjects of borderline notability. This template can only be used for articles that already exist here, and anyone placing this tag instead of sending it to AFD is implicitly voting for the subject being notable. Here's the first 10 in Special:WhatLinksHere. I'd say they're 100% obvious notability. Do you disagree?
   • Ainu people
   • Afonso I of Portugal
   • Alfonso II of Asturias
   • Amathus
   • Ambiorix
   • Amphipolis
   • Alfonso Leng
   • Abu Nidal Organization
   • Alessandro Scarlatti
   • House of Ascania
3. Not relying on raw machine translation sounds like a "superior" approach to me.
4. I agree that bad content has many harmful effects.
5. We cannot realistically vouch for enwiki using only sources deemed acceptable here. In fact, we can guarantee that at any given point in time, some small percentage of sources here violate our ostensible rules. Some of the Wikipedias have higher sourcing standards than us. French, for example, does not have our "self-published experts" exemptions.

Given all this, it seems clear to me that the text recommendations on this template ought to accurately represent the actual rules. WhatamIdoing (talk) 02:29, 17 December 2024 (UTC)

I do not think editors are good at evaluating that. Or rather, I think that merely saying "don't include dubious" stuff has any real function. Good editors won't include that anyway. But they'll just do nothing, until a bad editor comes around and then put the dubious content in. Most people might edit within a familiar area - but this template is an anomaly specifically because it's instructing users to use an unfamiliar foreign language wiki, and assuring them that an accurate translation produces a good result regardless of their own expertise. Experienced editors with knowledge of a field would be able to research stuff and draw references from other wikis without a template so if we are talking about the "audience" of this template we *are* talking about people who are likely to make mistakes because they don't know what they don't know.

I'm not saying the article as a whole necessarily has borderline notability but that the specific aspects of the article that has no English language sources and no verifiable foreign language sources does. Let's pick the first example. This has a long standing link in the section on Ainu culture. Look at the JP article: observe that it has an original research tag and a factual inaccuracy tag. Broad sections of the article have no references at all. So yes, I think a lot of the JP article DOES have borderline notability and an accurate translation of it would indeed be very unwelcome. Let's look at the final article you listed, House of Ascania. Again, looking at the first section, no sources for the first five paragraphs. "The Ascanians are proven to have originated as Counts of Ballenstedt from an area that is now part of the Harz district of Saxony-Anhalt" Proven by whom? Dive into the archives of the talkpage and you realise the article is original research again - editor openly admits they wrote it from their own memory. I would be very very uncertain that this user used the few references that are present correctly. Now, more perniciously, this article lacks maintenance tags, so an inexperienced user might think this content is okay. So 2/2 of the articles I checked are problematic. Dubious, shall we say. But the templates encouraging people to translate them have been there for years. Perhaps this is a sign of many good editors clicking through, finding that the linked to article actually sucks, and not touching that crap, but are we just patiently waiting for the first wiki noob to make a mistake?Fangz (talk) 04:03, 17 December 2024 (UTC)

Or perhaps some of the junk has already slipped in.

I really think at this point this template might be actively making Wikipedia worse. Fangz (talk) 04:25, 17 December 2024 (UTC)

When you say something has "borderline notability", do you mean it has "borderline qualifications for having a separate article about the topic"? Because Wikipedia:Notability is what we say when we're talking about whether the topic qualifies for a separate article, and I am thinking that you're using that word to mean something else, like "doubtful value of the current content".

BTW, "I wrote it from memory" isn't necessarily WP:OR. OR is stuff that no reliable source has ever published. If your memory is accurate – for example, you remember which actor played the lead in that film, or you remember your history teacher pounding on certain facts – then adding uncited content isn't OR. NB that it might be a violation of all that's good and right, and there's a decent chance that it's a violation of WP:V, but it's not technically original research. OR requires that what you put in the article has never been published in any reliable source, including reliable source you don't cite, or even sources you didn't know about. WhatamIdoing (talk) 06:29, 17 December 2024 (UTC)

I'm using rather loose language, yes, apologies. My overall point is that in my experience, quite often (perhaps even a majority of times) when you see content that is in a foreign Wikipedia and not in English, especially when a lot of time has passed, it's not included for very good reasons. The template as it is seems to imply a degree of trust in the content that is not warranted. Even if you argue that it's within the letter of the rules to just port over the content, I think the practical reality is that it really goes against the spirit of what verifiability and reliable sourcing is trying to achieve. Fangz (talk) 08:14, 17 December 2024 (UTC)

I don't think that's the case. When we already have an article here (and if this template is involved, then obviously we already have an article here; otherwise, there'd be no page to put the template on), then we should attempt to have something more than a short stub. The passage of time doesn't automatically make our article any better or the linked article(s) any worse.

As for "what verifiability and reliable sourcing is trying to achieve", I believe that they're trying to achieve content, in English, that is "not wrong". Not necessarily the truth, not necessarily cited, but not actively wrong. I wonder if you agree with me. WhatamIdoing (talk) 19:34, 17 December 2024 (UTC)

Whether it's "automatic" or not, the effect of the template when used without immense caution is frequently to make it worse, and produce potentially content that is worse than a simple stub in terms of having incorrect content. As my simple survey shows, the problems with this template are pervasive. And there's 80k+ times it's been used.

I don't agree that this is what they are trying to achieve. What those policies are trying to achieve first and foremost is to make Wikipedia trustworthy. That's the undercurrent of the verifiability policies. More important than correctness is the need to show a paper trail, a process for identifying and removing mistakes, a system that tends to improve over time (even if not necessarily monotonically), an degree of honesty and integrity and so on. It's stuff like that which I strongly believe failure to verify, especially in this instance, erodes. Fangz (talk) 05:09, 18 December 2024 (UTC)

I like the vision that you put forward about the "paper trail", but the fact is that readers don't click on any refs 99.7% of the time,[2] so the paper trail doesn't make an article seem trustworthy to them.

What makes people believe that an article is trustworthy – per research; this is not just my own opinion – tends to be mostly factors that apply to any website:

• It confirms what they already believed.
• It matches what they see in other websites. For example, people will search for COVID symptoms in their favorite web search engine, and if Wikipedia, Mayo Clinic, UK's NHS, and US CDC's websites all say the same thing, they trust the results; if one of them doesn't match, then they distrust the outlier.
• It doesn't look "broken" (e.g., no security certificate warnings when they click on the site, not filled with so many grammar errors that an average person will notice, doesn't have a significantly outdated appearance, things work as expected when you click on them).

and one factor that applies specifically to Wikipedia:

• They don't realize that anyone can edit the articles.

Understanding how Wikipedia works immediately and significantly decreases trust in Wikipedia. It's not entirely permanent, because people rationalize away their new knowledge to justify their pre-knowledge decision to trust Wikipedia, but it's still true that one of the most effective things you can do to maintain trust in Wikipedia is to never tell people that they can edit.

I'm not sure what your "simple survey" is. WhatamIdoing (talk) 06:04, 18 December 2024 (UTC)

I'm saying that click on a random selection of the articles targeted by this template and it's clear that for a large proportion a simple translation is not appropriate and would introduce issues into an article that in many cases is arguably superior to the other language version. Much more critical thought is required than what is being currently suggested.

I think the two of us have argued this enough and it's clear you aren't willing to change your opinion. I'd rather hear from someone else. Fangz (talk) 06:05, 21 December 2024 (UTC)

Template-protected edit request on 7 December 2024

Mention "large language models such as ChatGPT" as one of the machine translation tools. 67.209.130.35 (talk) 12:31, 7 December 2024 (UTC)

To warn people against using them? Fangz (talk) 09:47, 8 December 2024 (UTC)

Not really, i mean mention them as one of the tools used for machine translation among Google Translate, DeepL, etc. 67.209.130.128 (talk) 02:52, 12 December 2024 (UTC)

As the text would look like this:

Machine translation, like DeepL, Google Translate, and large language models (such as ChatGPT), is a useful starting point for translations, but translators must revise errors as necessary and confirm that the translation is accurate, rather than simply copy-pasting machine-translated text into the English Wikipedia.

67.209.130.128 (talk) 02:54, 12 December 2024 (UTC)