Jump to content

Wikipedia talk:Wikipedia Signpost/Single/2025-03-22

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia

Comments

The following is an automatically-generated compilation of all talk pages for the Signpost issue dated 2025-03-22. For general Signpost discussion, see Wikipedia talk:Signpost.

Essay: Unusual biographical images (3,684 bytes · 💬)

High quality reporting again, thanks for the laughs! Polygnotus (talk) 03:13, 22 March 2025 (UTC)

Disappointed there was no reference to "Pictures of You". DMacks (talk) 14:56, 22 March 2025 (UTC)

Somehow you guys missed this picture to the right. (t · c) buidhe 17:39, 22 March 2025 (UTC)

I was thoroughly entertained by these pictures. Great work! TNM101 (chat) 08:11, 23 March 2025 (UTC)

Thanks for this, very interesting to look at. A minor point, I don't think it's accurate to say that ""Good articles" are required to have an image (unless it is impossible to obtain one)." The explanatory footnote for GACR#6 specifically says "The presence of media is not a requirement." Eddie891 Talk Work 08:55, 24 March 2025 (UTC)

I think this belongs here. Dieter Lloyd Wexler 10:11, 24 March 2025 (UTC)

Interesting reporting.–Vulcan❯❯❯Sphere! 01:35, 25 March 2025 (UTC)

Great article, thanks! Love it.But puts a question into my mund. I was thinking of Andy Warhol's Marylyn Monroe portraits. By putting considerable skill, energy, and (let's be honest) chutzpa when using the image as a basis for this new work. He could sell it (for more than the price of the Shire too) without permission. Is this not correct? Alright, suppose I find of an image of, let's say, the punk singer Angry Pope online, copyrighted. Suppose I download it, then do photoshop stuff to make really different -- photoshop effects and whatnot -- but reasonably helpful and accurate image of what the guy looks like (just as Warhol's was). It is a work of skill and craft, a new thing. Can't I uoload it, release it under our usual license, and use it in an article? What am I missing? Herostratus (talk) 03:01, 25 March 2025 (UTC)

C:COM:Derivative works (t · c) buidhe 05:26, 25 March 2025 (UTC)
See also Marilyn Diptych §Appropriation and fair use. The fact that Warhol was forced to pay a settlement for copyright violation is like half of our article on that artwork. Caeciliusinhorto-public (talk) 12:24, 25 March 2025 (UTC)

Been laughing for the past 5 minutes. Extremely funny (and informative!) Paprikaiser (talk) 21:49, 26 March 2025 (UTC)

From the editor: Hanami (397 bytes · 💬)

Thank you, Master of Ceremony Bri. You've found very good words, sending out a bright atmosphere. (A sense of poetry does not need any more words.) --Just N. (talk) 17:27, 27 March 2025 (UTC)

Gallery: WikiPortraits rule! (2,924 bytes · 💬)

  • I think it is a very fine line between letting people pose for their own wikipedia page photo presenting a certain persona and letting people edit their own wikipedia page. I think we should be cautious about using any deliberately posed photos because of the precedent it sets. It will open the floodgates to publicity photos. NeedsGlasses (talk) 08:02, 24 March 2025 (UTC)
    • These types of photos are for celebrities, which are a very large part of the encyclopedia. Very roughly we're talking about 20-25% of all articles! We do need good photos for this many articles. I do think that getting the celebrities or their agents to upload their photos would be the most efficient solution. If we just had a program that would contact the agents and tell them what they needed to do to upload them to Commons, we'd be much better off than currently. I'd suggest something very simple, say suggesting a normal photo studio take and upload 3 headshots of the style used for high-school yearbooks. Of course they'd want to get a bit fancier and that might make the articles a bit more entertaining, but hopefully they wouldn't go too far overboard. If they did, Wikipedia editors might not include the extravagant photos in the articles. And the cost to young struggling actors? Maybe a couple hundred bucks? Agents used to distribute photos of their clients to newspapers all the time for use in news stories. Why not now? Smallbones(smalltalk) 15:05, 24 March 2025 (UTC)
  • Great initiative, more better portraits for better article quality.–Vulcan❯❯❯Sphere! 01:30, 25 March 2025 (UTC)

In the media: The good, the bad, and the unusual (4,596 bytes · 💬)

So can we restore the article deleted by the WMF yet? Doc James (talk · contribs · email) 06:02, 22 March 2025 (UTC)

@Doc James: Ultimately that's a question for WMF legal, but my best guess is no. The "ruling" as I remember it is that in the next Supreme Court hearing in early April ANI's lawyers have to address the issue of whether the High Court's decision was out of line because it puts a high burden on free speech and freedom of the press. I'll add that IANAL, and have no expertise about Indian law. I do see the ruling as great progress on the case and that it is a victory for the WMF's go-slow, one-step-at-a-time legal approach. Just a quick example, I was thinking that some of The Signpost's earlier coverage of this matter might set off another bad reaction from the court (but IMO we were right to take that small chance). I had no such doubts when writing this article. Smallbones(smalltalk) 09:58, 22 March 2025 (UTC)

I am a bit disapointed that in the The New Yorker talks to Heritage Foundation, they never mentions who they plan to dox. Surprise, surprise: it is editors who are not "pro-Israeli" enough. Strange that this fact "disappears" from the story. Huldra (talk) 21:26, 25 March 2025 (UTC)


Allegedly bad photos

  • Regarding self-appointed arbiters of which photos are "bad" or "not bad"? As for the McEntire photo example that was given: other than the presence of a smiling bystander who could trivially easily be cropped out if anyone actually cares enough to be arsed to do so — a bystander who cannot plausibly be accused of photobombing, because all he was guilty of was the (allegedly) grave sin of merely standing next to her — what is allegedly bad about that photo? Is it (allegedly) the fact that it is an action shot, an unposed shot, in which she is speaking, rather than a posed shot? If so, what's so (allegedly) "bad" about that? I think many humans are quite weird about alleging "badness" of photos. The whole notion feels much like a Rorschach exercise that reveals latent biases and hangups that are very much not flattering for the people in which they are revealed; but yet many people pride themselves on the result! When someone alleges that that photo is "bad" (even though it is not blurry, not distorted, not discolored, and so on), what are they saying? Is it that being posed is the only way to look good? Is it that real-life photos without cropping and photoshopping or airbrushing are inherently "ugly" until they have been ideologically "corrected"? Is it that the alleger lacks the capacity to look at a photo and not be helplessly distracted by the least imperfections short of portrait-studio still-photo composition? I don't see why so many people seem so proud of that. Quercus solaris (talk) 21:19, 29 March 2025 (UTC)
  • That photo is bad because Reba McEntire looks like Reba McEntire? I find it appealing. Smallchief (talk) 08:45, 30 March 2025 (UTC)
Well, we're all self-appointed, aren't we? We see a badly worded sentence, a tangled subheader structure, a broken or irrelevant ref link, an underexposed or misleading picture, and we fix it, replace it, or beg for someone else to take care of it. And if some fairly prominent outsider social media site has a section for making and publishing such judgments, we insiders may want to draw inspiration from them or not. I have replaced photos with later ones that appeared, including one of Gantry Plaza State Park here in New York. Other editors might agree with my choice or not, in which case I can reply in whatever way I think will help the work best. It's no more alarming, or no less, than disagreements on word choice. A few days ago I inserted my photo of three editors at our local gardening edit-a-thon and spent a minute pondering whether to crop out one of the three, and finally decided it's only an internal page that my fellow insiders will see, so it's not worth another minute. Jim.henderson (talk) 04:14, 3 April 2025 (UTC)

News and notes: Deeper look at takedowns targeting Wikipedia (6,564 bytes · 💬)

  • Not sure how DMCA requests work, but do they need proof? Seems a bit dumb to take down the page on Mothers Day, which I'm fairly sure does not contain infringing material. Also how would a page be restored to the results? ARandomName123 (talk)Ping me! 05:40, 22 March 2025 (UTC)
    • No, they do not need any proof, which is why most DMCA notices are bogus. The only way the page can be restored in Google results is if someone files a DMCA counter-notice with Google, which could be done by anyone that has added content to that page. Nosferattus (talk) 03:25, 25 March 2025 (UTC)
  • I imagine some bit of AI is automatically sending out these DMCA takedowns and Google simple complies with them all. There should be a significant punishment for those sending in fake DMCA takedowns but that would require an act of government I imagine. Doc James (talk · contribs · email) 05:53, 22 March 2025 (UTC)
  • For context, such absurdly wrong DMCA takedown notices have long been a problem, see e.g. this ten year old article by the EFF: "Absurd Automated Notices Illustrate Abuse of DMCA Takedown Process". Regards, HaeB (talk) 06:33, 22 March 2025 (UTC)
  • "To what extent is the Wikimedia Commons community invested in the dissemination of knowledge via images on Wikipedia articles?" Pretty heavily, in my experience. But it is the Wikipedia community (rather, each one of the 300 or so of them) who decides what is shown on (or linked to from) their Wikipedia, not the Commons community. [Courtesy ping to User:SDeckelmann-WMF ] Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 13:42, 22 March 2025 (UTC)
    (And by proxy the Wikidata community, as a great number of Wikipedia projects use automatic infoboxes). Reconrabbit 15:07, 27 March 2025 (UTC)
  • Re Deckelmann's comments on Wikimedia Commons:
    There's not as many videos honestly, but a lot of images, really incredible ones that's false, there are also many videos including pretty incredible ones. Over 100 000 videos from YouTube alone, and Google & DuckDuckGo apparently censor them from their 'Videos' tab. One example category of high-quality videos is this but there's also many other ones (many of them in here) (btw video2commons used to import nearly all of these seems currently broken for a month and WMF is not taking care of it but leaving it all to volunteers). the primary focus of the Commons community is the collecting of free content, rather than its dissemination. Yes, there is not much consideration so far about whether the platform is actually used, useful, and known – a top issue in that regard is, I think, how it's indexed in Web search engines (see the proposal linked above). Two further proposals for example would be to integrate Wikimedia Commons into the open source NewPipe mobile app and to improve to Commons app (e.g. enable playing videos).
    I think there's a lot that could be done if we were to think about this system holistically and the ways in which they, the different projects complement or don't complement each other. and How much more powerful and effective would it be if we were really thinking hard about the ways in which we could showcase all of the incredible images Agree – see m:Community Wishlist/Wishes/Suggest media set in Wikidata items for their Wikipedia articles.
    Also see this discussion at Commons talk:Media knowledge beyond Wikipedia. --Prototyperspective (talk) 16:26, 22 March 2025 (UTC)
  • Minor point, does the "Video" tab of Google search send the user to anywhere besides YouTube? I find that the tab gets me only YouTube, but using the general tab and adding the word "video" to the search item opens up the world of clips, merely heavily weighted to YouTube. Jim.henderson (talk) 01:58, 25 March 2025 (UTC)
    It does still find videos on other sites (e.g. this search takes me to this Vimeo video right now, although this one doesn't). Regards, HaeB (talk) 05:51, 25 March 2025 (UTC)
    Yes, not just YouTube and Vimeo but lots of even quite small sites. Just made a short check and it showed dailymotion, IGN, facebook, mubi, shutterstock, tiktok, Videoclip.bg, etc. Also news sites like dw.com, cnn or theguardian. Note the filetype is in the url on Commons too, e.g. ending with .webm. Btw, PeerTube tech could be an option if there ever are any load issues. Also I think we should try to make Commons categories (example examples) show up when people search the Web for e.g. "subject images" or "subject free media" etc. And video2commons works again now. Prototyperspective (talk) 17:35, 25 March 2025 (UTC)

Obituary: Rest in peace (488 bytes · 💬)

  • Deepest condolences to their family and friends.–Vulcan❯❯❯Sphere! 01:37, 25 March 2025 (UTC)

Opinion: Talking about governments editing Wikipedia (12,256 bytes · 💬)

  • "he boasted, on his userpage at Commons, that he had obtained permission from the official Kremlin.ru site for all photos there to be uploaded to Commons under Creative Commons licenses." I've put a lot of effort over at least decade and a half, lobbying for my (United Kingdom) government to make content (including RAF imagery) and data available under open licences. Does that make me corrupt? Is everyone who puts a PD US government (including USAF and NASA) image into an article also corrupt? Andy Mabbett (Pigsonthewing); Talk to Andy; Andy's edits 17:34, 22 March 2025 (UTC)
    • That's a pretty silly idea, Andy. I give 5 quick reasons why Russavia looks suspicious, then you take one out of the Russian context and suggest that I'm accusing you of being corrupt. With Russiavia you should look at the big picture, e.g. his block log on en.wiki at [1], how many times does it need to go beyond warning the guy to blocking the guy, to globally blocking the guy indefinitely before you can see there is something suspicious about the guy. Or you might look at his collection of hundreds of "Polandballs" c:Category:Polandball to conclude that this is the most racist stuff you've seen onWiki. Get real. Smallbones(smalltalk) 22:11, 22 March 2025 (UTC)
  • The number one country I would be concerned about in terms of government employee editing of articles is Israel, who didn't even get mentioned as one of the Big Scaries. I think it's also pretty ludicrous worrying about the permanuked editor Russavia at this late date. Then again, you once dismissed me as a "paid editor" in an ad hominem manner, as I recall, so I guess we should all just consider the source. xoxo, —tim //// Carrite (talk) 23:32, 22 March 2025 (UTC)
    • Thanks for the comment @Carrite:. My aim here was to get a reasonable conversation going with Larry Sanger an all those who think that US government employees are doing something they shouldn't be doing with Wikipedia. As above, I think that most of the concerns are overstated (except for the politicians and the spooks). I hope everybody realizes that Heritage Foundation, (maybe) Musk, and the ADL have all expressed some interest in dragging the WMF in front of Congress and perhaps attacking the WMF tax exempt status, so we should have some conversation to prepare for that. The politicians wouldn't be too worrying except for the shear number of them that edit and get caught. It is the spooks who worry me, but how am I going to expose any major group of spooks editing here? Even the WMF would have great difficulty dealing with them IMHO. In fact I on't really want to know any specific plans the WMF has - to have any chance at all of dealing with the problem they should keep their plans secret.
So you commented on Israel - are they a problem? I just checked an ISSI 2021 report on the major cyberwarfare powers [2] see also a secondary source [3]. They list
    • Top tier - USA only
    • 2nd tier: Australia, Canada, China, France, Israel, Russia and the United Kingdom. So Israel is in there
    • 3rd tier: India, Indonesia, Iran, Japan, Malaysia, North Korea and Vietnam
  • The only real threats to the US in these lists are Russia, China, Iran, North Korea and perhaps Israel IMHO Smallbones(smalltalk) 03:03, 23 March 2025 (UTC)
    • Cyberwarfare indeed is a specialized discipline requiring top level technical capabilities that not all countries have. Propagandistically disrupting wikipedia can be done by much less skilled people operating on a comparatively shoestring budget. Unlike many countries Israel is known for being rather aggressive in its approach to public relations and it could certainly stand to benefit from skewing our coverage even to a moderate extent. (t · c) buidhe 07:15, 23 March 2025 (UTC)
      • I don't know how much of an impact Israel has on this Wiki as of now, but certainly some worries have been raised over at Hebrew Wikipedia, as we previously reported. Oltrepier (talk) 18:11, 27 March 2025 (UTC)
  • The link to Rene Gonzalez currently points to the baseball player rather than the politician. I'd amend it myself, but not sure what the rules are for editing Signpost articles after they've been published. Cheers! BaduFerreira (talk) 22:57, 23 March 2025 (UTC)
    • Thanks, I've made the correction. In general, readers can make ordinary copy editing changes like this one, but not anything that would change the meaning. So thanks for asking and being careful. Smallbones(smalltalk) 00:19, 24 March 2025 (UTC)
  • The problem with such laws and/or regulations is that they tend to be overbroad. For example, I've disclosed that I've been paid money as an election worker or election consultant or poll watcher (not all on the same day - that's another conflict of interest). I've been a seasonal employee of boards of elections for years. If the Federal government forbids any funds of any kind being used for Federal elections, then I would not be able to work with the local board of elections in 2026 or 2028. Or I would be forced to quit an as editor here. That would violate my First Amendment rights. Bearian (talk) 15:00, 24 March 2025 (UTC)
    • Yes, one of the big problems would be if they get away from "Federal employees" and go to "Federal money". Federal money works its way thru the whole economy. An example 1) rebuilding bridges. With big collapsed bridges (e.g. in Baltimore, Philadelphia, Minneapolis). Federal money will go to the state effected, and then to contractors and sub-contractors. Say one of the subcontractors is being paid to keep the public informed on the progress of the rebuilding, putting text and photos online. Could the subcontractor give permission or help a Wikipedian upload the photos to Commons? How about the general contractor or the state? 2) Or maybe a non-profit uses some federal money to put on a play or build a specialized school. Or 3) volunteers work for the National Park Service, e.g. as tour guides or cashiers in a museum shop. They might not get paid, but would likely use break facilities or get a "free lunch". Could they work in any way with Wikipedia? I'd think the only realistic cutoff point would be for any prohibition to apply only to Federal employees. Smallbones(smalltalk) 20:17, 24 March 2025 (UTC)
  • The US government has funded several Wikimedian-in-Residence positions through the Department of Agriculture, the National Archives, and the National Institute of Occupational Safety and Health. These positions sometimes include editing Wikipedia, and I consider them incredibly positive contributions. But yeah, government employees shouldn't be editing pages about candidates on-the-clock. Rachel Helps (BYU) (talk) 15:35, 26 March 2025 (UTC)
    • Ok @Rachel Helps (BYU):, this is important. At first glance, it looks like Sanger may have been right. First, I'd like to know if any Wikipedian is currently being paid to edit as a Federal government employee. If so, please do not publicly reveal their name or user:name as that may be considered outing (revealing their employer). But I would certainly like to know so that I can check it out. I think you can at least reveal the number of people you are talking about. And please email me with as much info you are comfortable releasing. My other major concerns are about how long ago the Wikimedians-in-Residence were editing, and were they direct federal employees (or was it a grant via a private organization). I'll leave it there for now, but I'm likely to get back to you. Smallbones(smalltalk) 17:29, 26 March 2025 (UTC)
      • As far as I know, the federal government isn't currently employing anyone who edits Wikipedia as part of their job description. However, it's no secret! I've presented on WiRs who edit Wikipedia in the past at Wikimania (naming them with their permission). Dominic Byrd-McDevitt's work at NARA to get their images on Commons was very public and easy to run into if you work on any topic in American history. Rachel Helps (BYU) (talk) 19:03, 26 March 2025 (UTC)
        • Thanks, @Rachel Helps (BYU):. I'll dive into this a bit more. It might take me awhile. I do think that not having any current fed employees editing (that we know of) should satisfy most of Sanger's concerns. Smallbones(smalltalk) 19:24, 26 March 2025 (UTC)
      • "Right" in what sense? I'd be delighted if we had assistance from the (US) National Archives on (American) Civil War regiments. Or from the (UK) Dept. of Nat. Heritage on listed buildings. Nor do I see this outwith Government responsibility for the education of the population. I certainly don't want any political editing from any government - or anyone else. All the best: Rich Farmbrough 22:28, 30 March 2025 (UTC).
        • @Rich Farmbrough: Sorry, I should have written "Sanger may have been correct" in his sly evidence-free accusations that there are federal employees directly contributing to Wikipedia. There would certainly be nothing wrong with NRHP employees supplying Wikipedians with better data for historical listings, but we get exactly what the general public gets, which is worse than our own listings except for the new listings (our listings are cobbled together from federal and state agencies). I'd guess NARA has muster roles for Union regiments in the US Civil War, but I've never seen one and if they wanted to get these out to the public, likely the most efficient way would be to put them on Wikipedia. Smallbones(smalltalk) 21:22, 31 March 2025 (UTC)
  • @Smallbones: I'm unbelievably late to the party, but another good example of reported breaches of WP:COI rules in this context might be the series of suspect edits made to articles of members of the Scottish Parliament we reported on last year. Oltrepier (talk) 18:20, 27 March 2025 (UTC)
    • Yup, I did mention UK, Canada, and Australia briefly above, but we're really just dealing with US federal employees here (except for analogies). It's actually getting to be a good party or conversation now. I'm learning a few things. Thanks. Smallbones(smalltalk) 19:18, 27 March 2025 (UTC)

Recent research: Explaining the disappointing history of Flagged Revisions; and what's the impact of ChatGPT on Wikipedia so far? (19,848 bytes · 💬)

  • The Wikimedia Research Fund section could use a sentence about what the fund is, what it does, etc. I hadn't heard of it before this article. –Novem Linguae (talk) 22:02, 23 March 2025 (UTC)
    Thanks for the feedback. I think the list of previously funded proposals may already convey quite a bit of information in that regard, as do the changes mentioned. But point taken. Regards, HaeB (talk) 02:27, 24 March 2025 (UTC)
  • Taking a look at the words to watch page, the usage of "crucial" is hinted in the puffery section and "additionally" might be something to add to the editorial section. – The Grid (talk) 13:33, 24 March 2025 (UTC)
  • Most of these articles about ChatGPT's impact seem to assume the current chatbot model will just keep going, if not grow even more. I'd like to remind everyone OpenAI is still losing billions of dollars a year and, if investor confidence wanes even a little bit, may not be able to provide free access to ChatGPT for much longer. It takes insane amounts of power to run an AI data center and the product just isn't valuable enough to justify the expense. This blog post discusses AI's profitability problems at length. HansVonStuttgart (talk) 08:41, 26 March 2025 (UTC)
    That's an extremely misguided comment. Whatever one thinks about OpenAI's finances (and I would recommend some healthy skepticism about Ed Zitron's rants; in any case, re investor confidence, the company raised another $40 billion right after you posted this comment here, so its demise is presumably still some time away):
    The cost of operating LLMs like ChatGPT "has fallen dramatically in recent years" [4]. And one now "run a GPT-4 class model on [a] laptop" [5] - i.e. a model that is significantly better than the free version of ChatGPT was during the timespan covered by the studies reviewed here.
    Regards, HaeB (talk) 04:54, 29 April 2025 (UTC)

"Wikipedia Contributions in the Wake of ChatGPT"

Figure 2a
Figure 2b
  • I'm skeptical of this paper's conclusions. I think this WWW article's evidence is actually partly inconsistent with the narrative of ChatGPT as a competitive threat to Wikipedia. Figure 2 shows a dramatic increase in Wikipedia views and edits in "dissimilar articles" much larger than the decrease observed in "similar articles". These similarity categories are based on an attempt to classify if a Wikipedia article is similar to content the chatbot would produce. But it is difficult to pin down what explains the difference between these sets. They set it up so that the dissimilar articles are the "control group", as if they would be unaffected by ChatGPT. But that's not the story they tell in Figure 2. The headline could easily be ChatGPT increased Wikipedia usage and contribution if the researchers had started from a different narrative frame. I'm still skeptical that current chatbots pose a competitive threat to Wikipedia. Wikipedia gets you facts faster than Chatbots, and has a stronger brand among its users for verifiable and factual information. Groceryheist (talk) 15:07, 26 March 2025 (UTC)
    Update: The more I look at this, the less convincing it is. The assumptions of their DiD just don't seem plausible if you look at the data in Figure 2. There's a big increase in views for dissimilar articles that starts before the change. The decline in views for similar articles begins prior to the change. This invalidates the "parallel trends" assumption of the DiD estimator. They only estimated a large decrease in views for "similar" articles in the DiD because there was an increase in views to "dissimilar" articles in the same period. Groceryheist (talk) 15:24, 26 March 2025 (UTC)
@Groceryheist: Yes, this confused me greatly too. For what it's worth though:
1. Parallel trends assumption: We should note that the authors directly claim that Figure 2 [...], which shows similar pre-ChatGPT evolution of these two groups, bolsters our confidence in [the "parallel trends"] assumption. Now, I don't think that one should blindly defer to a Nobel prize winner. (As I mentioned in the review, this wouldn't be the first Acemoglu-coauthored or -promoted paper on AI impacts that attracts criticism for statistical flaws - or, in recent news, worse.) But at the very least they didn't ignore this assumption.
There's a big increase in views for dissimilar articles that starts before the change. The decline in views for similar articles begins prior to the change. - note that Figure 2 actually shows residual views and edits, more specifically mean residuals for each month of activity 𝑡 aggregated over similar and dissimilar articles respectively from the "Comparative time series" regression on the page preceding Figure 2. And that regression already includes a linear trend[] for [...] activity time. I too am more familiar with visually checking the parallel trends assumption by looking at the actual outcome variable instead of such averaged residuals. But perhaps in this residuals view one is supposed to check the parallel trends assumption by e.g. visually inspecting whether the error bars cover the mean mean residual (dashed lines) pre-launch, which they do actually do fairly well in Figure 2? Just guessing though. In any case, the authors might question what you mean by "big" exactly in big increase in views for dissimilar articles - shouldn't that be put in relation to the size of the standard errors? Other observations derived from ocular inspection like The decline in views for similar articles begins prior to the change should probably take the error bars into account, too.
2. Eyeballed trends from Figure 2 vs. DiD regression coefficients in Figure 3: When the authors write in section 3.1 about Figure 2 that (my bolding)

For both views and edits, we see that similar articles exhibited little changes in activity from the pre-GPT to the post-GPT period (accounting for controls). Dissimilar articles, on the other hand, show an increase in edits after ChatGPT launched [...]

perhaps they simply refer to the positions of the dashed lines (which show the [mean] mean residuals for similar (blue) and dissimilar (red) articles over the pre-GPT and post-GPT periods respectively)? But it's in weird contrast to what they say right afterwards in section 3.2 based on the DiD regression(s):

The diff-in-diff coefficients for Figure 3a (views) are negative and statistically significant for all article ages except 𝑇 = 1, which implies that Wikipedia articles where ChatGPT provides a similar output experience a larger drop in views after the launch of ChatGPT. This effect is much less pronounced for edit behavior.

My hunch is that they may have gotten a bit carried away with these ocular observations in section 3.1 (perhaps forgetting themselves for a moment that Figure 2 shows residuals only?) and should have stuck with reporting the DiD outcomes, instead of yielding to the temptation of (to use your expression) "telling a story" about Figure 2 already. It's also worth being aware that Figure 2 only shows the situation for a single T (T=6).
But I'm genuinely still confused about all this too. At the very least, it is safe to say that the paper goes through only a small part of the steps recommended in this new "practitioner's guide" to Difference-in-Differences designs (see p.53 f., "Conclusions"; also, section 5.1. there about 2xT designs looks like it could be relevant here, but I haven't gotten around to reading it yet).
In any case, thanks a lot for raising this. As mentioned, it had confused me as well, but I didn't discuss this problem in the review because a) I wasn't sure about it and b) this review and the rest of this issue already felt grumpy enough ;-) Overall, writing this roundup has been a bit of a disconcerting experience - the quantitative impact of the AI boom on Wikipedia is one of the most important research questions about Wikipedia in recent memory, and sure, it is not easy to tackle, but the quality of what has been published so far is not great. (I still think the "Wake of" paper might be the one with the most solid approach among those mentioned in this issue.)
Regards, HaeB (talk) 08:25, 18 May 2025 (UTC) (Tilman)
But perhaps in this residuals view one is supposed to check the parallel trends assumption by e.g. visually inspecting whether the error bars cover the mean mean residual (dashed lines) pre-launch, which they do actually do fairly well in Figure 2?
Hey! Thanks for responding in such thoughtful depth. I think Figure 2 is exactly what you'd want to look at for evaluating parallel trends for their DiD models. The DiD model is exactly the "comparative time series model" plus the terms needed to statistically test for post-intervention differences in treatment effects. This part of the study seems well-executed, as is true of the study overall, my concerns are entirely about "interpretation". What is a bit surprising or unclear is why they felt the need to adjust for length and creation month at all. The linear trend for time in these two models means that both lines (similar and dissimilar) in Figure 2 should be flat if there were no differences between similar and dissimilar articles. So we'd expect flat lines pre-cutoff and slopes post-cutoff; however, I see slopes pre-cutoff for both edits and views. Edit: The lines pre-cutoff could not be flat, but trending in the same direction if there were higher-order temporal trends (i.e., acceleration in editing or views).
That said, I don't think it appropriate to consider the standard errors in Figure 2 when considering the parallel trends assumption since this isn't a theoretically justified or interpretable test of the assumption. One reason is that the parallel trends assumption is about the "counter-factual" of what would have happened in the absence of treatment. This fundamentally isn't possible to test. Looking at pre-cutoff patterns and extrapolating is one approach, and people also tend to squabble over the substance of the phenomena to decide how much weight to give a DiD estimate. My understanding of DiD is that it is particularly prone to mislead when pre-cutoff trends are opposite, which to me appears so in this data.
shouldn't that be put in relation to the size of the standard errors? Other observations derived from ocular inspection like "The decline in views for similar articles begins prior to the change" should probably take the error bars into account, too.
Even if we do try to use SEs to tell if the trends are statistically significant or random fluctuations, it's hard to dismiss the pre-cutoff trends. For edits which looks like an increase of almost 1 SE for dissimilar and a decrease of about 0.5 SE for similar, and something happens in the three points before the cutoff that could be a bit opposite trend or a random fluctuation.
Overall, I wouldn't claim that the assumption is surely "invalid", but it is far from bullet proof.
I do agree that statements interpreting the magnitudes of these changes make more sense in the context of the of the variance in the outcomes. By that standard, it might be difficult to claim any of the changes as "big". But comparing the far-left and far-right of Figure 2 the change for dissimilar articles looks to about nearly 2 SE. For edits that change might be about 1 SE. Groceryheist (talk) 17:05, 18 May 2025 (UTC)
@Groceryheist: Thanks. This is helpful, but I'm still confused about various aspects here. (And yes, I'm familiar with the basics of DiD and aware that the parallel trends assumption is counter-factual, as you explained - but it does seem that various methods have been proposed in the literature that are at least partly more rigorous than mere eyeballing.)
Re I think Figure 2 is exactly what you'd want to look at for evaluating parallel trends for their DiD models - do you happen to have a reference at hand where one can read up more on this method? (It may be a while before I get to take a look at that chapter in the aforementioned new guide by Baker et al. to see if it is indeed relevant here.)
The only other point I'd add right now regarding DiD validity is that the authors rather audaciously assume the effect of ChatGPT to have been constant during the post-launch period (per the 1t ≥ Dec 2022 term in the DiD regression) - whereas we can be fairly certain that its use was still growing rapidly at least in the first several months of that period (one reason why above I felt inclined to inspect the pre-launch half of Figure 2 first).
PS: I added the corpus delicti above for easier reference.
PPS: Another thing that is really unfortunate about this paper is that the authors did not release any replication data or code (not to talk about preregistration).
PPPS re my general frustrations (above) about the current lack of high-reliability quantitative findings in this area, I noticed Mako stating recently that due to generative AI, for the first time in history, Wikipedia is beginning to see decreased viewership. So maybe he has formed some judgment about these papers too and might want to weigh in here? (Although in that article it is framed as a reference to the trend of AI content being listed first in web search results, ahead of Wikipedia. But that could be a mixup, considering that websites featured in those AI Overviews might also benefit from the extra backlinks, so we can't conclude from their mere presence that they reduce our traffic. And at least according to one SEO firm that tries to monitor their impact systematically, Wikipedia is probably not among their victims so far.)
Regards, HaeB (talk) 07:38, 26 May 2025 (UTC)
Hey @HaeB thanks for continuing the conversation. Nothing comes to mind as far as a citation for the DiD pretrends visualization. It's just a sort of standard part of the DiD study design. Including the overall trend in the visualized model makes it easier to see if there's a lack of parallelism pre-treatment. If the overall trends were opposed then we'd see a bigger difference. I actually kinda prefer eyeballing to statistical tests of pre-trend parallelism, since I think those tests lend an undue air of rigor. They can lead you to reject an analysis based on insubstantial differences. E.g., a KS test of normality will reject the null for pretty much any sufficiently large dataset that isn't generated by a gaussian process. People might conclude not to use a z-test in such a case, even though the test is fairly robust to deviations from normality.
Instead of a statistical test of pre-treatment parallelism, what I'd really like to see added to this would be a sensitivity analysis to quantify how violations of the assumptions would affect the results.
Good observation that the treatment effect isn't constant over time, it seems reasonable to assume it would tend to increase. That said, I think this one would make their analysis and interpretation conservative since those increases wouldn't be fully captured in their estimate of the treatment effect; part of the increasing effect would be captured by the post-treatment slope parameters.
One other DiD validity issue that @Jdfoote brought up in another channel was that, under their theory that more similar articles are more exposed to the treatment, there should be a dose-dependent treatment effect. I.e., we should observe a greater effect for the most similar articles. The way they split articles into similar and dissimilar groups seems a bit arbitrary from this perspective. Treating the treatment as continuous instead of discrete would be an interesting step.
I guess @Mako was just referring to the overall decline in page views as seen in e.g., [6]. As you describe, there could be multiple mechanisms for this decline.
Thanks for pointing out that you're unsatisfied with the research into this question so far. I agree the one we're discussing is maybe the most useful we've seen. Thinking about what else we might to do shed light on this. Groceryheist (talk) 19:46, 26 May 2025 (UTC)
and has a stronger brand among its users for verifiable and factual information - maybe so (I think WMF made such a claim based on survey data from its 2023 ChatGPT plugin experiment). But "among its users" does a lot of work here. Contrast these two studies we previously reviewed here:
Regards, HaeB (talk) 06:22, 18 May 2025 (UTC)
Blind tests inherently factor out any effect of Wikipedia's brand on credibility. Groceryheist (talk) 17:07, 18 May 2025 (UTC)
But yeah. These lab-based studies do suggest that some part of the potential audience doesn't ascribe all that much prestige to Wikipedia, at least compared to the chatbots. Groceryheist (talk) 19:07, 18 May 2025 (UTC)

Traffic report: All the world's a stage, we are merely players... (0 bytes · 💬)

Wikipedia talk:Wikipedia Signpost/2025-03-22/Traffic report



.

Retrieved from "https://en.wikipedia.org/w/index.php?title=Wikipedia_talk:Wikipedia_Signpost/Single/2025-03-22&oldid=1281762309"