Wikipedia:Village pump (miscellaneous)

 Policy Technical Proposals Idea lab WMF Miscellaneous 

• Table of contents
• First discussion
• End of page
• New post

The miscellaneous section of the village pump is used to post messages that do not fit into any other category. Please post on the policy, technical, or proposals sections when appropriate, or at the help desk for assistance. For general knowledge questions, please use the reference desk.

Discussions are automatically archived after remaining inactive for a week.

« Archives, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75

can i get a download txt file of all entries in wikt:category:English lemmas

can i — Preceding unsigned comment added by 115.188.159.190 (talk) 10:35, 12 July 2023 (UTC)

You can get a list using the "Download data" here. As the query has only one column, the tsv option will produce a simple text file. Certes (talk) 13:39, 12 July 2023 (UTC)

WMF English banner fundraising campaign 2023 - community collaboration starting now

Dear all,

We would like to share with you the community collaboration page around the English fundraising banner campaign. This page is for en.wiki volunteers to learn about fundraising and share ideas for how we can improve the 2023 English fundraising campaign together. On this page you’ll find information to increase transparency and understanding of the fundraising program, background on improvements around community collaborations that have been made since the last campaign, new spaces for collaboration, and messaging examples to invite volunteers to share ideas for how we can improve the next campaign together.

The fundraising banner pre-tests phase on English Wikipedia starts on the 19th of July with a few technical tests, using messaging that was created with the community during the December 2022 campaign. We will regularly update the collaboration page with new messaging ideas and updates on testing and campaign plans as we prepare for the main campaign that will launch at the end of November.

Generally, during the pre-tests and the campaign, you can contact us:

• On the community collaboration page
• Via one the community calls (see more details on the collaboration page)
• At an in-person event (see more details on the collaboration page)
• On the talk page of the fundraising team
• If you need to report a bug or technical issue, please create a phabricator ticket
• If you see a donor on a talk page, VRT or social media having difficulties in donating, please refer them to donatewikimedia.org

Best wishes,

JBrungs (WMF) (talk) 15:57, 13 July 2023 (UTC)

Please don't be a sock proxy

For those of you who have never seen this, there are a number of LTAs who try to avoid scrutiny (or work around page protections) by asking other editors to perform edits for them. Sometimes these are requests on user talk pages. Sometimes they're via email. If you get a request to perform an edit from somebody you've never heard of (especially if they're a new account), just don't do it. RoySmith (talk) 01:14, 14 July 2023 (UTC)

For example, 27.65.26.100 (talk). --Redrose64 🌹 (talk) 22:18, 14 July 2023 (UTC)

And 2402:800:6344:6610:E9CA:3119:3C88:9387 (talk), who is probably the same person. --Redrose64 🌹 (talk) 07:28, 15 July 2023 (UTC)

2402:800:6305:1035:E9CA:3119:3C88:9387 (talk). --Redrose64 🌹 (talk) 09:50, 16 July 2023 (UTC)

Wikipedia:Sockpuppet investigations/Haiyenslna is directly relevant. --Redrose64 🌹 (talk) 09:57, 16 July 2023 (UTC)

Andrzej Sapkowski and The Witcher Saga

Hello!

In ENG Wikipedia article about Sapkowski, the beginning of the Witcher series is stated as The Blood of Elves book. But, hm... as a Polish fantasy fan, there are 2 short stories collections that truly - in my view - began the saga and without which the Witcher saga wouldn't be created? And the very last short story is about Ciri and the attack of Nilfgaard. And many things characteristics of the Witcher series (moral ambiguity, Dandelion, Yennefer etc.) originated in the short stories.

Thus, my question is - shouldn't the article about Sapkowski be changed to include the short stories as the beginnings of the Witcher saga?

(I mean, I dunno, maybe you already had this kinda talk before and I am spamming? But from my Polish perspective, the article kinda... does not sound right?).

Best regards -- Kaworu1992 (talk) 15:04, 14 July 2023 (UTC)

@Kaworu1992 The best place to discuss things like this is on the talk page for the article in question. You're talking about Andrzej Sapkowski, so the associated talk page is Talk:Andrzej Sapkowski. On the other hand, what you're suggesting doesn't sound controversial, so what I'd really suggest is be WP:BOLD and make the change. If anybody objects, they'll (politely) raise an objection and then you can discuss it further, per WP:BRD. RoySmith (talk) 15:12, 14 July 2023 (UTC)

Is there a tool for auto-archiving citations?

Per title, I seem to recall something like this but not sure how to find it, but I have like 250 citations to archive so it'd be handy. Darkwarriorblake (talk) 14:27, 16 July 2023 (UTC)

https://iabot.toolforge.org?page=runbotsingle * Pppery * _{it has begun...} 15:56, 16 July 2023 (UTC)

Thank you!!Darkwarriorblake (talk) 17:19, 16 July 2023 (UTC)

Movement Charter drafts invitation for feedback

Hello everyone,

The Movement Charter Drafting Committee is happy to announce that new draft chapters of the Movement Charter are ready for review and feedback. The Global Council draft is available now and Hubs will be published by the end of July.

How can you engage with the Charter content? To create a Charter for our Movement means we need to hear from as many of you as possible. Everyone in the Wikimedia community is invited to actively engage with the content by sharing their feedback on wiki or attending upcoming virtual and in-person events.

We encourage individuals or groups, especially those from under-resourced Wikimedia communities, to apply for grants by July 30. These grants can be used to organize conversations, such as informational sessions to familiarize fellow community members with the draft chapters of the Movement Charter ahead of regional and thematic events from September to November, 2023.  The Regional Specialists of the Movement Communications team are available to support community organizers.

Posting on behalf of the Movement Charter Drafting Committee, RAdimer-WMF (talk) 20:03, 16 July 2023 (UTC)

How do I permanently delete my Wikipedia account?

Please, I would like to permanently delete my Wikipedia account as well as all my data if possible. MarceloLanda6 (talk) 18:59, 18 July 2023 (UTC)

MarceloLanda6, accounts cannot be deleted due to legal requirements related to attribution of authorship. Since you have made so few edits, the easiest solution is to just abandon the account and never log in again. A more comprehensive solution is Wikipedia:Courtesy vanishing. Cullen328 (talk) 19:14, 18 July 2023 (UTC)

Thank you so much MarceloLanda6 (talk) 19:39, 18 July 2023 (UTC)

Feedback on user script chatbot

After selecting text, the control panel on the right is used to give instructions. The responses by the AI model are presented in the chat panel on the left.

I wrote a user script called WikiChatbot to assist editors. It can summarize, reformulate, copyedit, and provide suggestions on additional topics, images, and wikilinks as well as assess factual accuracy and bias. It is used by selecting text in an article and then clicking one of the buttons on the right to enquire about the selected text. The chatbot can also be used by typing specific questions about the selected text or the topic in general in the chat panel.

The script uses the AI model GPT 3.5. It requires an API key from OpenAI. New OpenAI accounts can use it freely for the first 3 months with certain limitations. The AI model was not designed to assess or improve encyclopedic articles and has many serious shortcomings. Editors should always question its responses and rely on their own judgment and reliable sources instead. For a more detailed description of all these issues and examples of how the script can be used, see the documentation at User:Phlsph7/WikiChatbot.

I was hoping to get some feedback on the script in general and how it may be improved. I'm not sure how difficult it is to follow the instructions so it would be great if someone could try to set up the script, use it, and explain which steps were confusing. My OpenAI account is already older than 3 months so I was not able to verify the claims about the free period and how severe the limitations are. If someone has a younger account or is willing to open a new account to try it, that would be helpful. Other feedback on the idea in general, its problems, new features to implement, or the documentation is also welcome.

(side note: this text was already posted at Wikipedia:User_scripts/Requests#Feedback_on_user_script_chatbot but it was suggested that here might be the better place to bring up the issue.)

Phlsph7 (talk) 09:06, 19 July 2023 (UTC)

Not to be a downer, but I think this is a fundamentally bad idea. Or, rather, it could be a good idea for certain "mature" editors, but will be a honeypot trap for good-faith but naive and chatbot-trusting editors to inflict a lot of damage. I see that you're including a lot of warnings like "Please scrutinize this before changing the article" but unfortunately, there clearly exists a group of people who are extremely, extremely trusting of all LLM output and won't see the need (e.g. the famous lawyer who ignored the voluminous warnings about "may invent false data" to trust it as if it was a legal database). And it's precisely these editors who will be most excited to install such a Chatbot script. If nothing else, I would highly, highly recommend removing the "Is this true?" / "Is this biased?" / "Expand this article" options, due to the risk of AI hallucinations, or to mark it as "Does the AI think this is true (use only for comedy purposes)" or the like.

For the hypothetical copyediting ability, I don't think this comes up that often. Maybe if you have a truly poorly written section that was somebody plopping Google Translate'd material in, perhaps it'd be an improvement, but GPT3 copyediting any halfway decent text is not going to be clearly better. Worse, there will be a class of editors who aren't very strong copyeditors who can't tell the difference, and assume that they are being a big help by throwing complicated text in GPT3 and getting something they assume is better out. We've rejected similar proposals to highlight simple "complexity of text" measurements (e.g. Flesch–Kincaid readability tests) for fear of good-faith editors running around in technical articles to rephrase them with a bunch of short sentences with common synonyms and have them think they're "helping." I suppose the one potential good use of GPT is to take an existing high-quality source and summarize it, but that wouldn't be operating on Wikipedia text then; that'd be throwing a reliable newspaper article in and hoping that the result isn't too much of a close paraphrase. But it looks like the script as is works on Wikipedia text. SnowFire (talk) 19:45, 19 July 2023 (UTC)

Hello SnowFire and thanks for having a look at the script and expressing your concerns. I think you are right that LLMs are double-edged swords that come not only with interesting opportunities but also with significant dangers. As I see it, it may not be the best response to try to avoid LLMs altogether. A different approach is to look for ways to use them productively while also making potential users aware of all their dangers and limitations. For this script, one way to do this is to clearly discuss the dangers in the documentation and include warnings inside the script.

I'm not sure that, as you say, naive editors are most likely to install the script. The great majority of script users are experienced or "mature" editors.

Your idea of renaming the button as "Does the AI think this is true (use only for comedy purposes)" would be one way to tackle the problem. But this would be a rather long button title. Maybe there is another way of addressing your concerns about the buttons "Is this true?", "Is this biased?", and "Suggest expansion". One could add an automatic warning right at the beginning of the chatbot response. For "Suggest expansion", it could be "(Please consult reliable sources to ensure that the following information is accurate before making any changes to the article)" or "(The following suggestions should be scrutinized since they may contain false information)". In my tests, I mostly got helpful results for this button: it did not write new text for the article but make suggestions like "provide more information about the historical context", "explain what this term means", or "mention the scholarly debate surrounding this claim". But this may depend on the article subject. I'll modify the script accordingly.

For copy-editing, it depends a lot on the type of article you are working with. If you prepare a GA article for FA, it's unlikely that you would need it a lot. However, if you work with stubs and drafts, this can come in handy. It can also be a problem for older articles that get very little attention from editors. In my tests so far, I didn't come accross any serious problems with this button. It usually only introduces minor changes, like fixing grammar, changing from passive to active, or splitting an overly long sentence into two parts. But of course, that doesn't mean that it is to be blindly trusted. I'm sure that there are also cases where it misunderstand an expression and changes the meaning in its attempt to copyedit it. Phlsph7 (talk) 21:36, 19 July 2023 (UTC)

"In my tests, I mostly got helpful results for this button" - ah, but that's exactly the problem. If GPT just spat straight nonsense, it'd be harmless, and if it only told the truth or "I don't know", it'd be fantastic. Instead, GPT will give you the right answer 80% of the time, I don't know 10% of the time, and a wrong but plausible-looking answer 10% of the time. We'd never accept as a reliable source something that just convincingly lied 10% of the time. If someone clicks the "Is this true?" or "Expand this section" button they should just see a preformatted static message saying "Read the sources!" Otherwise, if there's some editor who actually started aggressively using this feature without checking the sources, they will start seeding total nonsense in with the rest of their edits, which can be very bad if done on obscure topics without many other editors checking it. Which are, coincidentally, the areas where GPT is most likely to hallucinate.

Copyediting - I agree with you that it shouldn't be blindly trusted, I'm just saying that there will be editors who blindly trust it.

"The great majority of script users are experienced or "mature" editors." - When I say "naive", I mean in general, not in their familiarity with being an editor. We've had extremely hardcore editors who are astonishingly naive and clueless. This is difficult to get into in-depth without being accused of throwing shade, and even picking only banned editors can result in an accusation of gravedancing, but... we have to remember that Wikipedia is a volunteer project, which is a great thing 98% of the time, but it does mean that there's no "filter" for some people who are not very mature (sometimes for good & proper reasons, like still being 12 years old!). Maybe a "safe" one to mention is Neelix, who was undoubtedly editing in good faith, yet created enough pure nonsense to get an entire speedy deletion criterion named after him (WP:NEELIX). Or for another example of how warnings can go unheard, Wikipedia has loudly said since its beginning that both American and Commonwealth English are supported styles. Yet we still will have editors come along on a mission to "correct" articles to "their" side and feel that they're providing a very important service in doing so.

Anyway, you might find this ChatGPT manuscript an amusing read: https://chat.openai.com/share/183ff912-224c-4846-80f7-13314ce55b48 . In particular, check out the answers on the Ottoman - Ptolemaic War and the surprising new details on the life of Mamiya Rinzō. I knew the answers to these questions as I asked them, and it's very impressive how much ChatGPT got right! ...but... it's definitely still not 100%, and it's not 100% in the dangerous way of "inventing details from whole cloth to justify its fabrications." SnowFire (talk) 22:20, 19 July 2023 (UTC)

I implemented the suggestion and included automatic warning messages for the buttons in question. I think you are right that the accuracy rate of AI models is not sufficient to implement their suggestions directly into articles. So if someone were to propose a script to directly use text freely generated by AI models to expand articles, I would be on your side. But this is not what this script does. It merely gives suggestions and warns editors about the reliability of these suggestions. This can be useful to many editors since it may prompt them to research what reliable sources say on the issue before working on the article. But it could be misused by editors who ignore the warnings and do not care about what reliable sources say. However, for almost every tool, there are ways to misuse it. If it was a criterion for scripts that it shouldn't be possible to misuse them then there would be very few scripts left on our user script list.

If it turns out that you are right and that this script mainly attracts naive and clueless editors then it would be a good idea to remove functions that could be misused or to remove the script altogether. However, there are good reasons to think otherwise and I'm not sure that these steps are justified based only on the vague assumption that the script will be mainly used by naive and clueless editors.

Thanks for sharing the interesting chat on real and fake wars. Please note that the hallucinations about Mamiya Rinzō (and the Ottoman - Ptolemaic War) occurred after a lengthy chat. Hallucinations tend to become more likely if the chat history gets long. The chat history for my script is very limited. The script ignores the chat history if buttons are used. This should mitigate some of the problems but does not fully solve the issue. Other factors are the AI model used and the temperature parameter. When I tried to put the question about the Japanese invasion of Kamchatka right into the first prompt (replacing the Toledo War), it gave me the correct response. I also tried adding the response you got to our article War (only using the preview function, I didn't change the text) and used the button "Is it true?". It correctly identified the problem. Phlsph7 (talk) 08:24, 20 July 2023 (UTC)

There will be a lot of enthusiasm from people who find editing to be difficult or boring, but there will be skepticism from others who regard editing by predicting what word might go next with suspicion. At a minimum, any tool like this must tag edit summaries with some standard code for periodic checking. Johnuniq (talk) 09:52, 20 July 2023 (UTC)

That's a good point about edit summaries. In the chatbot documentation, it is stated in the second lead paragraph and later in the section "Usage and purpose" that editors have to mention in the edit summary if they used text created by the chatbot. This is also in tune with the policy draft WP:LLM. However, I'm not sure that it's possible to add new tags directly with the help of a user script. If so, it probably requires special user rights. According to WP:Tags, tags are usually added automatically by MediaWiki. Phlsph7 (talk) 10:09, 20 July 2023 (UTC)

I don't know, we'd have to ask at WP:VPT. Some assistance with setting up a tag and triggering it would certainly be needed although I don't know if it's possible here. Johnuniq (talk) 10:38, 20 July 2023 (UTC)

Having a tag for AI-assisted edits in general could be considered. One difficulty would be how to find out whether an AI was used for a particular edit. For example, it probably wouldn't be a good idea to simply tag every single edit of editors who have the chatbot script installed. A manual approach would be to add a checkbox called AI-assisted next to the minor edit checkbox and rely on editors to be honest about it. I'm not sure that this is already such a frequent phenomenon to merit its own checkbox. Until that time, the best approach may just be to tell editors to declare AI assistance in the edit summary. Phlsph7 (talk) 11:32, 20 July 2023 (UTC)

Tags can be made a part of the script in such manner that it is applied only when an edit is made using that script. For example, "AWB", "Twinkle", "AntiVandal", etc. —CX Zoom[he/him] ^{(let's talk • {C•X})} 11:45, 20 July 2023 (UTC)

Thanks for the clarification. The script does not make any edits itself. It only provides suggestions. For example, if editors want to use it for copyediting, they have to copy the copyedited text provided by the chatbot in the chat panel, open the edit page, paste the new text into the textarea with the wikitext, and manually push the button "Publish". Phlsph7 (talk) 12:03, 20 July 2023 (UTC)

Unfortunately as shown by the most recent ANI case even long-term editors can fail to understand the short comings of LLMs. They do not understand content, in the same way the Go AI was shown not to understand Go. They are only predicating the best fit text. That could still be extremely useful in giving a starting point for editing, but the litany of recent cases show it should be handle with caution. -- LCU ActivelyDisinterested ∆transmissions∆ °co-ords° 11:35, 20 July 2023 (UTC)

Notes