Talk:Organizational culture

The lack of mechanisms by which culture can be modelled has severely hampered social science research to date. Why then is there a desire to delete the section on measurement? Is there perhaps a fear amongst social scientists that modelling & measurement will be an end to the endless, and arguably pointless debate regarding what is, or what is not culture?

This article was the subject of a Wiki Education Foundation-supported course assignment, between 5 September 2024 and 13 December 2024. Further details are available on the course page. Student editor(s): Zw2306 (article contribs).

— Assignment last updated by JANJAY10 (talk) 00:41, 5 November 2024 (UTC)

Criticism of "organizational culture" began in the early 1980s. Most criticism comes from writers in critical management studies who for example express skepticism about functionalist and unitarist views.

This sounds like it was written by someone in HR trying to protect the org. Most criticism comes from workers, not writers. “Writers" as it is used here is incredibly nebulous, likely deliberately so as to obfuscate the fact that workers are dissatisfied with organizational culture, not "writers". Viriditas (talk) 04:57, 5 February 2025 (UTC)

This article was the subject of a Wiki Education Foundation-supported course assignment, between 10 June 2025 and 29 July 2025. Further details are available on the course page. Student editor(s): Ajpopereflect (article contribs). Peer reviewers: MTurner2025.

— Assignment last updated by Leading Learning (talk) 17:30, 6 July 2025 (UTC)

Summarize Timeline Top Qs Fact Check

So this article had four editors contributing between the 16 and 20 of September. They added about 42k of content between them, and never came back. I'd guess this was an educational project that ignored WP:WEP protocol and wasn't announced on the talk page.

The edits were:

• User:OmongbChar added 6k of content, then deleted it, then added it back rewritten. It reads a little robotically and ends with an AI-style Finally... summary.
• User:219.89.22.65 added 21k of content with some unusual formatting and another Finally.... Their edit summary also has an AI tone, saying at defensive length how they ensured neutrality and clarity, supported arguments with academic sources, and updated the content to meet Wikipedia’s style, accuracy, and comprehensiveness standards.
• User:Merrenchige added 8k of content: this diff has a strong AI tone with lots of list-of-three examples, and it cites a 2019 paper on "the ethical use of AI" in business, which sounds a bit premature.
• User:Emsososo added 9k of content, which doesn't seem obviously AI generated to me, at a glance

Spot checking references I haven't found any fake ones, but I may have been lucky. I don't know enough about the article subject to spot factual mistakes.

How much of this is likely to be AI generated, and how much should we revert? Pinging User:Gnomingstuff who reverted all of it, and User:Sundayclose who restored it saying that it needed discussion on the talk page. Belbury (talk) 19:15, 3 November 2025 (UTC)

This was the analysis I was going to do - thank you for getting to it! Idk if you saw my comment at Special:Diff/1320264929 but I came to the same conclusion that this was likely a group of "undeclared" students using LLMs to expand this article for a class assignment. NicheSports (talk) 20:11, 3 November 2025 (UTC)

Please don't ping me. I have said everything I have to say and if I wasn't listened to then I'm not going to be listened to now. Gnomingstuff (talk) 21:31, 3 November 2025 (UTC)

This text is definitely AI-created (which is not an issue, IMHO). However, it is also mostly devoid of information (which is very bad), and, in the only place I have checked, uses falsified source to support text that is not related to this source at all (which is disastrous). See my analysis at the User_talk:Gnomingstuff#AI edits??, and conclusion Since AI can generate bullshit very quickly, it is not fair to offload [verification] problem onto other editors and thus waste their time checking all of AI output, so it is very reasonable to just delete unverified AI contributions in their entirety without analyzing them word-by-word. Викидим (talk) 23:02, 3 November 2025 (UTC)

I restored Gnomingstuff's reversion. Belbury and I seem to have independently concluded that these edits were LLM-generated and added by a group of undisclosed students. And all three of Викидим, Gnomingstuff, and me agree that the correct approach is a complete reversion of those students' edits. Courtesy ping to @Sundayclose NicheSports (talk) 23:28, 3 November 2025 (UTC)

I said I was going to stay away from this issue for a while, but I'll just pose a question. Some of it appears to be unsourced. I'm fine with removing that content. Except for that, are there policy violations; does it just need cleaning up; or can it stay in the article because it's good enough? Sundayclose (talk) 20:39, 3 November 2025 (UTC)

The (admittedly local) consensus for cleanup at WP:AINB is that in cases of high-probability unreviewed LLM article expansions/rewrites, a complete reversion is appropriate. This is also consistent with the logic behind the WP:G15 speedy deletion criteria, which has project-wide consensus and I think has been working well. NicheSports (talk) 20:44, 3 November 2025 (UTC)

I'm no expert on the matter, but that seems contrary to WP:BATHWATER if there is good content in the mix. It makes a lot more sense to put the adequately sourced, adequately written content on the talk page to see if it can be salvaged. It also seems strange to me that an individual editor can make an identical edit as AI and if it's not problematic it can stay. As for WP:G15, I read that as pertaining to deletion of an entire article, not just parts. Sundayclose (talk) 21:20, 3 November 2025 (UTC)

Bathwater is an essay though... And It also seems strange to me that an individual editor can make an identical edit as AI and if it's not problematic it can stay doesn't apply here - multiple editors have now identified this as high-probability unreviewed LLM use and documented issues with it (NPOV, unsourced content, etc.) which means this is not equivalent to content written by a skilled human editor. There is fairly widespread community consensus that given the asymmetries involved (LLM content is much quicker to create than to fix) that there is considerable leeway in reverting edits composed of such content NicheSports (talk) 21:31, 3 November 2025 (UTC)

Essays are not without merit. Let me make sure I understand something. If there is content that has high "probability" of being AI generated and it's removed, does that mean I, as a skilled human editor, can't restore the content if I check if for verifiability, accuracy, adequate writing, and no policy violations? Sundayclose (talk) 21:35, 3 November 2025 (UTC)

If you want to go ahead and verify this content line by line, source by source, I will absolutely hold off on reverting it, and will add a note in the {{ai-generated}} tag that an editor is addressing the content. NicheSports (talk) 21:37, 3 November 2025 (UTC)

Sorry if I was unclear; I'm speaking in general, not just this article. So what's the difference in not removing AI content to begin with vs. removing it and then restoring it after carefully checking for problems? That seems like a silly exercise. Why not just remove the parts that don't pass muster and leave the parts that do? Sundayclose (talk) 21:40, 3 November 2025 (UTC)

I'll quote myself: There is fairly widespread community consensus that given the asymmetries involved (LLM content is much quicker to create than to fix) that there is considerable leeway in reverting edits composed of such content. You could also ask the same question about why G15 exists, but it does. I was serious about my suggestion/offer above btw, not being pointy - if you want to adopt this article and do the rigorous line-by-line verification, I will hold off on reverting. NicheSports (talk) 21:42, 3 November 2025 (UTC)

I also think there is a misunderstanding among editors who don't work with LLM content. LLMs hallucinate constantly. All the time. I have reviewed biographical stubs where 50%+ of the article claims were hallucinated - including date, location, and cause of death, even though these were all documented in both the nytimes and guardian obits for this individual. There are likely extensive hallucinations in the edits these students made to this article, which would take days and potentially library access to books to track down. There are also hundreds of these types of article expansions happening every day - just check the 1325 and 1346 edit filter logs, or look at the drafts at AfC. It is not realistic to apply WP:BATHWATER to these edits. NicheSports (talk) 21:46, 3 November 2025 (UTC)