Wikipedia:Wikipedia Signpost/Single/2024-10-19

The 2024 election for the Board of Trustees of the Wikimedia Foundation is now complete. Preliminary results were announced earlier this week. 12 candidates ran for 4 Community- and Affiliate-selected Trustee seats in this election. The vote was conducted using a single transferable vote system, which transfers votes from eliminated candidates after each round to the voter's next preferred candidate.

The four candidates with the most votes were –

• Christel Steigenberger (User:Kritzolina, candidate statement)
• Maciej Artur Nadzikiewicz (User:Nadzik, candidate statement)
• Victoria Doronina (User:Victoria, candidate statement)
• Lorenzo Losa (User:Laurentius, candidate statement)

These four winning candidates will need to pass a background check and have all other bylaw requirements confirmed before they can be officially appointed at the board's December meeting.

Two of them (Doronina and Losa) are already on the board and were re-elected. The election came not long after the board's controversial refusal to ratify the "Movement Charter" after it had been worked on over several years (Signpost coverage: "Wikimedia community ratifies Movement Charter, Wikimedia Foundation rejects ratification"). As one Wikimedian noted on the Foundation-l mailing list: "Victoria and Lorenzo, who were greatly associated with the Movement Chart BoT veto some months ago, were reelected by the community, despite many predictions that they would suffer a big backlash for making public their positions, and a number of people was quick to predict their certain removal from office, especially Victoria."

A visual representation of the vote is given by the Sankey diagram above. Incoming Board member Nadzik also analysed some voting statistics on this page on Meta.

Congratulations to the winners and our thanks to the other candidates who ran in the election: Bobby Shabangu, Deon Steyn, Erik Hanberg, Farah Jack Mustaklem, Lane Rasberry, Mohammed Awal Alhassan, Rosie Stephenson-Goodknight, and Tesleemah Abdulkareem. – Sb, H, S

After a year of fewer RFAs, a flood of candidates have signed up in the upcoming trial of Administrator elections. With the call for candidates having closed on 14 October, 35 candidates will be seeking to be elected (or re-elected) as administrator. The full list of candidates can be found on the respective AELECT subpage.

There were a total of 14 RFAs this year until October. Considering AELECT, this brings 2024 to 49, reversing a multi-year trend of fewer admin candidates in successive years. The latest year with more RFAs was 2015, with 53 candidates over the year.

However, it is not clear how many of said candidates will end up passing. As far as is known, the previous record for the number of active RFAs at once is 28 on 6 December 2005, per this discussion in October 2007. The elections will also use private voting via SecurePoll, which may skew the results.

The discussion phase will be open for 72 hours, from 22 to 24 October. Anyone may ask questions at the election subpages, and candidates are "encouraged" to reply. Voting begins from 25 October and will continue till 31 October, closing at 23:59 UTC. After three days for discussion and questions, the candidate pages will be closed for discussion, and voting begins.

Admin elections and other RFA reform efforts were last covered by The Signpost in the 26 September issue. – S

Note: This section was prepared before the large number of applicants emerged for the administrator elections trial.

In prior Signpost coverage we discussed the declining number of active administrators:

• 2019-07-31 (first time under 500 active administrators)
• 2019-12-27 first time under 500 for a month
• 2023-01-01 first time under 500 for a year
• 2023-10-23 first time under 450
• 2023-12-04 more records in the 440s
• 2024-03-29 another new record: 441
• 2024-07-04 more records in the 430s
• 2024-07-22 another new record: 430
• 2024-09-04 another new record: 427
• 2024-09-26 another new record: 419

Now, on 7 October, Rick Bot has recorded 418 active administrators, another record low count. The monthly averages of data reported by the bot daily were:

2024

Period	daily average
January	465.7
February	473.3
March	448.2
April	438.3
May	438.3
June	434.6
July	435.4
August	432.5
September	425.4
October	421.0[a 1]

For comparison, using the same methodology, the average number of active administrators in the past several years (also shown in a chart created earlier through mid 2019) was:

Other years

Period	daily average
2017	543.0
2018	528.6
2019	515.1
2020	508.3
2021	487.2[b 1]
2022	470.0
2023	464.1

A trend of declining numbers of active administrators is apparent in both the monthly and annual sets of data. If no more changes occurred after the 10 October data cutoff date, then 2024 would tally a loss of 43 over the 2023 average. The next greatest annual decline in this table was about half as many: -21.1 between 2020 and 2021. – B

On September 27, Brazilian Wikimedian João Alexandre Peschanski was revealed to be one of the over 150 international researchers who co-signed a collective op-ed that deemed disinformation to be "one of the biggest short-term threats to humanity", and called for more protection and easier access to online data for those who study it and its perceived impact on public opinion.

This is a notable bit of news, since Peschanski—also known on-wiki as Joalpe—currently serves as the executive director of the Wiki Movement Brazil User Group, an association founded in 2013 and regularly affiliated to the Wikimedia Foundation since 2019.

The op-ed is available in several languages, including Spanish, German, Italian and Portuguese, and has been re-published by several media outlets. – O

The Product and Technology Advisory Council (PTAC) was appointed this week. This council was introduced as a one-year pilot program as part of Board of Trustees' three proposals in lieu of the Movement Charter.

Apart from WMF CPTO Selena Deckelmann and other WMF staff, the council includes:

• EPorto WMB and one more affiliate member
• Eugene233
• Lahi
• Putnik
• Sohom Datta
• TheDJ
• GorillaWarfare
• FJMustak
• Benjamin Mako Hill

According to the initial proposal, the 8 volunteer members would include 5 at-large technical contributors and 3 Wikipedia volunteers, at least one of who would be from the English Wikipedia. While it's not clear which category each selection was for, 4 volunteers have English Wikipedia as their most edited project – Sohom Datta, TheDJ, GorillaWarfare (also admin and OS), Benjamin Mako Hill.

PTAC was last covered by The Signpost in the 14 August issue. – S

The WMF published two bulletins since our latest Signpost issue – one for late September and early October respectively.

Of note were The Power of the Commons, an event hosted by the Wikimedia Foundation and other organizations as part of The Summit of the Future at the Headquarters of the United Nations and a new change to the Wikidata Query Service.

Highlights from the October issue include an e-mail from the "Wikimedia Foundation Board Governance Committee" and its Movement Charter debrief process (Spoilers: They'll collect feedback until January 2025).

More notably, nominations are currently open for positions on AffCom, Ombuds commission and the trust and safety Case Review Committee. AffCom plays an advisory role in official accession of new affiliates to the movement; the ombudsfolk are tasked with investigating complaints about infringements of the WMF privacy policy and related policies by CheckUsers, oversighters and others with access to sensitive user data; and the case review committee represents the community for appeals of certain trust and safety office actions. Applications for the Affiliations Committee close on November 18, 2024, and applications for the Ombuds commission and the Case Review Committee close on December 2, 2024.

Wikimedia Foundation has filed an amicus brief in a Mexican case known as Richter v. Google – more information about the case here, it is compared to Section 230 in the United States. Although Mexico does not have statutory Section 230 protections, the Internet Society says that they are subject to similar content moderation restrictions and protections due to international treaties with the U.S. and Canada.

– B, S

The Wikimedia Foundation announced a new round of grants from its Knowledge Equity Fund (see previous Signpost coverage), awarding "13 organizations in 10 countries, supporting work to address knowledge gaps and create and share new knowledge" with $1.362 million in total.

It also published impact reports for some of the Round 2 grantees announced in August 2023, and noted some highlights from three such reports:

• Aliansi Masyarakat Adat Nusantara ("Alliance of the Indigenous Peoples of the Archipelago", AMAN) – Indonesia ($200,000 USD): Final narrative report, Final financial report
  • "They organized three training sessions for indigenous and citizen journalists in North Sumatra, Riau, and Nusa Bunga. They report that 'in 2023, as many as 154 articles have been written by indigenous journalists (JMA) and published on their website. The traffic involved 3,728 users with 20,537 reactions on the website (clicks, shares, comments, etc).' [...]"
• Black Cultural Archives – United Kingdom ($290,000): Interim narrative report
• Criola – Brazil (a civil society organization dedicated to advocating for the rights of Black women in Brazilian society, $160,000): Final narrative report
• Data for Black Lives – United States ($100,000): Final narrative report
• Create Caribbean ("the first digital humanities center in the Caribbean", based in Dominica, $75,000): no reports posted yet
  • "They received 60 nominations to their public call for participants for their oral history project. The team has conducted virtual and onsite interviews with people in Antigua, Jamaica, Puerto Rico, Dominica and St. Lucia to collect nontraditional sources of knowledge."
• Project Multatuli (Indonesia, $50,000): no reports posted yet
  • "with the grant, they have produced 16 stories, including two photojournalism essays, that touch on a wide range of issues in Indonesia, ranging from the role of indigenous women in society, public health, and culture to underprivileged youth, disability, and human rights. [....]"

Another round two grantee doesn't seem to have provided reports yet (but the Foundation "will share the rest of the final reports as we receive them"):

• Filipino American National Historical Society (United States, $70,000)

In the recent announcement, the Foundation notes that

Through multiple community conversations that we hosted in 2023, we heard feedback from volunteers about the goals and impact of the Knowledge Equity Fund which led us to make some key changes for our upcoming rounds of funding.

These changes include

• More consistent and clear communication about the Knowledge Equity Fund, its grants and impact
• Opportunities for movement groups to also receive grants for work they are doing to address knowledge equity
• Clearer measures of impact for Knowledge Equity Fund grants

The announcement also states that

With the conclusion of Round 3, the Fund now has $815,000 USD left [from the $4.5 million it was set up with in 2020]. The Equity Fund will run one last “round” in the next 4 months, where we will choose a handful of the most impactful grantees from the first rounds and provide them with a final “top up grant” to deepen their work with the movement and ensure that the content they create is present on the Wikimedia projects.

– AK, H

• Annual reports: Wikimedia Ukraine, Wikimedians for Offline Wikis User Group
• Articles for Improvement: This week's Article for Improvement is Longshore drift. Please be bold in helping improve this article! Next week's Article for Improvement (beginning 21 October 2024) is Image.

A monthly overview of recent academic research about Wikipedia and other Wikimedia projects, also published as the Wikimedia Research Newsletter.

A new paper titled "The Rise of AI-Generated Content in Wikipedia"^[1] estimates

"that 4.36% of 2,909 English Wikipedia articles created in August 2024 contain significant AI-generated content"

In more detail, the authors used two existing AI detectors (GPTZero and Binoculars), which

"reveal a marked increase in AI-generated content in recent[ly created] pages compared to those from before the release of GPT-3.5 [in March 2022]. With thresholds calibrated to achieve a 1% false positive rate on pre-GPT-3.5 articles, detectors flag over 5% of newly created English Wikipedia articles as AI-generated, with lower percentages for German, French, and Italian articles. Flagged Wikipedia articles are typically of lower quality and are often self-promotional or partial towards a specific viewpoint on controversial topics."

The researchers also conducted a small qualitative analysis "to better understand the motivations for using LLMs to create Wikipedia pages", by manually inspecting the edit histories of a smaller subset, namely "the 45 English articles flagged as AI-generated by both GPTZero and Binoculars" (corresponding to 1.5% of those 2,909), and looking at the other contributions of the editors who created them. In this rather small sample, they identify four different such motivations:

Advertisement

One prominent motive is self-promotion. Of the 45 flagged pages, we identify eight that were created to promote organizations such as small businesses, restaurants, or websites. [...]

Pages Pushing Polarization

[...] we also identify pages that advocate a particular viewpoint on often polarizing political topics. We identify eight such pages out of the flagged 45. One user created five articles on English Wikipedia, detected by both tools as AI-generated, on contentious moments in Albanian history [see also figure 2, below ...] In other cases, users created articles ostensibly on one topic, such as types of weapons or political movements, but dedicated the majority of the pages’ content to discussing specific political figures. We find two such articles that espouse non-neutral views on JD Vance and Volodymyr Zelensky.

Machine Translation

[...] We find three cases where users explicitly documented their work as translations, including pages on Portuguese history and legal cases in Ghana. [...]

Writing Tool

Other pages, which are often well-structured with high-quality citations, seem to have been written by users who are knowledgeable in certain niches and are employing an LLM as a writing tool. Several of the flagged pages are created by users who churn out dozens of articles within specific categories, including snake breeds, types of fungi, Indian cuisine, and American football players. [...]

These are among the first research results providing a quantitative answer to an important question that Wikipedia's editing community and the Wikimedia Foundation been weighing since at least the release of ChatGPT almost two years ago. (Cf. previous Signpost coverage: Community rejects proposal to create policy about large language models, "AI is not playing games anymore. Is Wikipedia ready?", and in this issue: "Keeping AI at bay – with a little help from volunteers", summarizing media coverage of WikiProject AI Cleanup). The "Implications of ChatGPT for knowledge integrity on Wikipedia" were also the topic of a research project conducted in 2023-2024 by UT Sydney researchers (funded by a $32k Wikimedia Foundation grant) which just published preliminary results where "Concerns about AI-generated content bypassing human curation" are highlighted as one of the challenges voiced by Wikipedians.

The new study's numbers should be valuable as concrete evidence that the generative AI has indeed started to affect Wikipedia in this manner (but might potentially also be reassuring for those who had been fearing Wikipedia would be overrun entirely by ChatGPT-generated articles).

That said, there are several serious concerns about how to interpret the study's data, and unfortunately the authors (a postdoc, a graduate student and an undergraduate student from Princeton University) address them only partially.

First, the researchers made no attempt to quantify how many of the articles from their headline result ("4.36% of 2,909 English Wikipedia articles created in August 2024 contain significant AI-generated content") had also been detected (and flagged or deleted) by Wikipedians. Inspecting the aforementioned smaller subsample of 45 (1.5%) articles where both detectors agreed, they found that

"Most of the 45 pages are flagged by moderators and bots with some warning, e.g., 'This article does not cite any sources. Please help improve this article by adding citations to reliable sources' or even 'This article may incorporate text from a large language model."

Even for this smaller sample though, we are not told what percentage of AI-generated articles survived.

In other words, the paper is a rather unsatisfactory read for those interested in the important question of whether generative AI threatens to overwhelm or at least degrade Wikipedia's quality control mechanisms - or whether these handle LLM-generated articles just fine alongside the existing never-ending stream of human-generated vandalism, hoaxes, or articles with missing or misleading references (see also our last issue, about an LLM-based system that generates gene articles with fewer such "hallucinated" references than human Wikipedia editors). Overall, while the paper's title boldly claims to show "The Rise of AI-Generated Content in Wikipedia", it leaves it entirely unclear whether the text that Wikipedia readers actually read has become substantially more likely to be AI-generated. (Or, for that matter, the text that AI systems themselves read, considering that Wikipedia is an important training source for LLMs - i.e. whether the paper is evidence for concerns that "The ouroboros has begun".)

Secondly and more importantly, the reliability of AI content detection software - such as the two tools that the study's numerical results are based on - has been repeatedly questioned. To their credit, the authors are aware of these problems and try to address them. For example by combining the results of two different detectors, and by using a comparison dataset of articles created before the release of GPT-3.5 in March 2022 (which can be reasonably assumed to be virtually free of LLM-generated text). However, their method still leaves several questions unanswered that may well threaten the validity of the study's results overall.

In more detail, the authors "use two prominent detection tools which were suitably scalable for our study". The first tool is

GPTZero [.....] a commercial AI detector that reports the probabilities that an input text is entirely written by AI, entirely written by humans, or written by a combination of AI and humans. In our experiments we use the probability that an input text is entirely written by AI. The black-box nature of the tool limits any insight into its methodology."

The second tool is more transparent:

An open-source method, Binoculars [...] uses two separate LLMs [...] to score a text s for AI-likelihood by normalizing perplexity by a quantity termed cross-perplexity [...] The input text is classified as AI-generated if the score is lower than a determined threshold, calibrated according to a desired false positive rate (FPR). [...] For our experiments, we use Falcon-7b and Falcon-7b-instruct [as the two LLMs, following the recommendation of the authors of the Binoculars paper.] Compared to competing open-source detectors, Binoculars reports superior performance across various domains including Wikipedia"

The "superior performance" of the Binoculars tool (online demo) for the Wikipedia "domain" sounds very reassuring, with both precision and recall at or near a perfect 100% according to figure 3 in the "Binoculars" paper^{[supp 1]}. But it refers to the performance on a 2023 dataset called "M4",^{[supp 2]} where the AI "articles" to be detected were generated in a rather simplistic manner. ("We prompted LLMs to generate Wikipedia articles given titles, with the requirement that the output articles contain at least 250 words", see also the results for e.g. ChatGPT). It seems unwise to assume that this is representative of all the ways in which actual editors try to use AI to generate new articles in August 2024. Indeed the authors explicitly acknowledge this in a different part of the "Rise" paper, pointing out they did not attempt to "simulat[e] the various ways Wikipedia authors might use LLMs to assist in writing—taking into account different models, prompts, and the extent of human integration, among other factors." As a small illustration of potential issues with this, the few concrete examples of articles detected as AI-generated that are included in the paper (figure 2, see above) all start with an infobox - something which ChatGPT can certainly generate if explicitly prompted to do so, but which seems to be absent from most or all of the examples in the M4 dataset.

What's more, as is evident from Figure 1 (above), both tools disagreed frequently, with GPTZero being much more detection-happy than Binoculars in English, French, and German, but much less in Italian. The authors acknowledge that "the tools we use are primarily for detecting AI-generated content in English. While GPTZero supports Spanish and French, it is not designed for other languages."

As mentioned in the paper's abstract (see above), the authors try to control for false positives by calibrating both detectors to a 1% false positive rate on the control dataset (of presumably AI-free Wikipedia articles from March 2022). A technical issue that appears to have been overlooked here is that this 2022 dataset was generated (by Hugging Face) from the source wikitext dumps using the well-known "mwparserfromhell" Python package, whereas the authors obtained their August 2024 articles by scraping the text rendered by the Wikipedia API and applying some of their own cleanup steps. LLM-based text classification tools can sometimes be quite sensitive to minor formatting aspects.

More importantly, it seems rather adventurous to assume that the articles from that March 2022 dataset are comparable in all relevant properties to the newly created articles from August 2024 (i.e. that the 1% false positive calibration on the former will mean a 1% false positive rate on the latter). The authors are clearly aware of this concept drift problem, but only make a very perfunctory attempt to address it:

"One concern is that pre-March 2022 pages may be more polished due to years of editing. However, we observe that a higher number of edits weakly correlates with a higher AI-detection score for pre-March 2022 articles (Appendix D), suggesting that the FPRs for those articles may even be inflated. While the base assumption cannot be watertight, we observe a relatively consistent distribution of page categories between the two data pools, and we rely on the consistency of our chosen tools’ reported FPRs."

For many people, the fact that additional edits by human Wikipedia editors make the AI detection score go up in both detectors might increase skepticism about their overall validity. But the authors take it as an argument strengthening their paper's overall "Rise of AI-Generated Content" claim, by alluding to the possibility that its estimate might be too low. At various other points of the paper the authors likewise express awareness that their measurement method is subject to substantial errors and uncertainties, but claim that these can only go in their favor (i.e. could only mean that the actual rate of AI-generated articles is higher than their estimated lower bound). And there are other issues that likewise make one wonder a bit about the stringency of the peer review process that the paper has undergone, for example its claim that "The Wikipedia data we collect is under a Creative Commons CC0 License."

The study has only been been published as an arXiv preprint at the time of writing. But according to a remark in the accompanying code, it has been accepted at the "NLP for Wikipedia Workshop" at next month's EMNLP conference.

The authors have commendably published code used for the paper (although not under an open source license). Unfortunately though for readers who might want to replicate part of the paper's quantitative or qualitative analysis (or check whether some of the AI-generated articles it detected have slipped through Wikipedia's New pages patrol), none of the data underlying the paper's main results is being published (even though they were based entirely on public information from Wikipedia):

"Detecting AI may have unexpected negative consequences for people implicated as having generated that text. We have therefore been encouraged to omit any identifying information in the specific pages we discuss; however, we will provide more specific data to researchers upon request provided that it not be disseminated further."

But these concerns did not stop the authors from discussing the aforementioned concrete examples in a way that makes it very easy to identify involved users. (One reader of the paper has already done so, pointing out the specific longstanding sockpuppeting case that the editor featured in figure 2 was involved in.)

• See the page of the monthly Wikimedia Research Showcase for videos and slides of past presentations.

Other recent publications that could not be covered in time for this issue include the items listed below. Contributions, whether reviewing or summarizing newly published research, are always welcome.

From the abstract:^[2]

"This paper presents a new tool to support efforts to fill in these gaps by automatically generating draft articles and facilitating post-editing and uploading to Wikipedia. A rule-based generator and an input-constrained LLM are used to generate two alternative articles, enabling the often more fluent, but error-prone, LLMgenerated article to be content-checked against the more reliable, but less fluent, rule-generated article."

From the abstract:^[3]

"Wikipedia is composed from consensus. [...] While it is often positioned as a self-evident good, its usage on Wikipedia is not without concern. In this paper I mobilize Chantal Mouffe’s (2000) feminist critical political theory and Johanna Drucker’s (2014) methods of interface analysis to raise important questions about the relationship between consensus and peer production. [...] I identify the multitude of ways that Wikipedians perform consensus: not only through understanding and decision-making, but also through acts of composing, showing, processing, closing, and calculating. However, because Wikipedia’s socio-technical vision is over-determined by consensus, its political design is ill-equipped to address the political conditions of pluralist societies. As a result, I identify the reasons why Wikipedia should strengthen its democratic commitment by engaging with dissensus. By conducting this research, I demonstrate how consensus has transitioned from a democratic ideal into an interface and why it should be re-imagined within peer production projects."

From the abstract:^[4]

"Working from a feminist affective framework, this article reports on a study of student-editors' experience with Wikipedia-based writing, using their reactions to a key editing guideline, 'Be bold,' as an entry-point for examining their affective experience. The 'Be bold' guideline, which encourages would-be editors to 'just go for it,' is nearly as old as the English language version of Wikipedia itself yet has received little critical attention. Drawing on survey and focus group interviews from participants at the undergraduate and graduate level, this study's findings provide new understandings of novice editors’ affective experiences in Wikipedia while offering a critical analysis of the 'Be bold/ guideline."

See also a post by one of the authors in the "Wikipedia Weekly" Facebook group, with discussion

From the abstract:^[5]

"This article details settler colonial erasure of Native American and Indigenous histories, knowledges, and philosophies on [English] Wikipedia. I show that long-time Wikipedia editors follow the settler colonial logic of elimination to omit Native histories from Wikipedia’s American history pages; block Native and allied editors from adding scholarship that centers Native experience; and ban Native and allied editors from the website so that settlers can lay claim to digital space. To do so, I concentrate on Wikipedia’s United States and American history pages, and I detail editor discussions regarding Native histories on Wikipedia’s talk pages, noticeboards, and off-Wikipedia message boards where editors congregate. I supplement this information with interviews of Wikipedia editors engaged in editing these topics. [...] I ultimately provide suggestions for the Wikimedia foundation to combat settler colonial erasure on Wikipedia."

See also a "Public response to the editors of Settler Colonial Studies" by Wikipedia admin User:Tamzin in the June 8 Signpost issue, arguing that the paper's author had failed to disclose a conflict of interest and that the paper contained multiple factual errors.

From the abstract:^[6]

"This study examines the impact of integrating Google Translate into Wikipedia's Content Translation system in January 2019. Employing a natural experiment design and difference-in-differences strategy, we analyze how this translation technology shock influenced the dynamics of content production and accessibility on Wikipedia across over a hundred languages. We find that this technology integration lead to a 149% increase in content production through translation, driven by existing editors become more productive as well as an expansion of the editor base. Moreover, we observe that machine translation enhances the propagation of biographical and geographical information, helping to close these knowledge gaps in the multilingual context."

See also mw:Wikimedia_Research/Showcase#July_2024

From the abstract:^[7]

"[The] Wikipedia Revision History (WikiRevHist) [dataset ... can be a] valuable resource[] for NLP applications. [...] we report Blocks Architecture (BloArk), an efficiency-focused data processing architecture that reduces running time, computing resource requirements, and repeated works in processing WikiRevHist dataset. BloArk consists of three parts in its infrastructure: blocks, segments, and warehouses. On top of that, we build the core data processing pipeline: builder and modifier. The BloArk builder transforms the original WikiRevHist dataset from XML syntax into JSON Lines (JSONL) format for improving the concurrent and storage efficiency. The BloArk modifier takes previously-built warehouses to operate incremental modifications for improving the utilization of existing databases and reducing the cost of reusing others' works. In the end, BloArk can scale up easily in both processing Wikipedia Revision History and incrementally modifying existing dataset for downstream NLP use cases. The source code, documentations, and example usages are publicly available online and open-sourced under GPL-2.0 license."

From the abstract:^[8]

"The article compares selected entries on Wikipedia concerning significant historical events in three language versions: Belarusian, Lithuanian, and Polish. [...] I apply the method of ideological critique to investigate whether national values influence the objectivity of Wikipedia articles written in local languages. A comparison of multilingual Wikipedia entries reveals the prevalence of “local” points of view on controversial historical events."

From the abstract:^[9]

"After 2007, researchers started to observe that the number of active editors for the largest Wikipedias declined after rapid initial growth. Years after those announcements, researchers and community activists still need to understand how to measure community health. In this paper, we study patterns of growth, decline and stagnation, and we propose the creation of 6 sets of language-independent indicators that we call “Vital Signs” [formerly available at https://vitalsigns.wmcloud.org/ ]. Three focus on the general population of active editors creating content: retention, stability, and balance; the other three are related to specific community functions: specialists, administrators, and global community participation. [...] We present our analysis for eight Wikipedia language editions, and we show that communities are renewing their productive force even with stagnating absolute numbers; we observe a general lack of renewal in positions related to special functions or administratorship."

See also:

• A Twitter thread by one of the authors
• The documentation page of the 83,400€ Wikimedia Foundation grant that gave rise to this paper
• Our coverage of an earlier paper from the same project: "Wikipedia Editor Drop-Off: A Framework to Characterize Editors' Inactivity"

From the abstract:^[10]

"we conduct the first data-driven study of ban evasion, i.e., the act of circumventing bans on an online platform, leading to temporally disjoint operation of accounts by the same user. We curate a novel dataset of 8,551 ban evasion pairs (parent, child) identified on Wikipedia [by "Wikipedia moderators", at Wikipedia:Sockpuppet investigations], and contrast their behavior with benign users and non-evading malicious users. We find that evasion child accounts demonstrate similarities with respect to their banned parent accounts on several behavioral axes — from similarity in usernames and edited pages to similarity in content added to the platform and its psycholinguistic attributes. We reveal key behavioral attributes of accounts that are likely to evade bans. Based on the insights from the analyses, we train logistic regression classifiers to detect and predict ban evasion at three different points in the ban evasion lifecycle. Results demonstrate the effectiveness of our methods in predicting future evaders (AUC = 0.78), early detection of ban evasion (AUC = 0.85), and matching child accounts with parent accounts (MRR = 0.97)."

See also earlier coverage on research about sockpuppets on Wikipedia

Supplementary references and notes:

The Independent Journal Review, borrowing from The Daily Caller, recently claimed that your Wikipedia donations might be funding "feminism and racial justice", instead of just keeping the lights on. The Commune Mag and OpIndia then joined the fray, alleging that Wikimedia's finances are tied to shadowy donors. Karah Rucker of Straight Arrow News listed left-leaning leaders like Art+Feminism and Black Lunch Table—programs we've previously highlighted in The Signpost. Meanwhile, in yet another familiar critique of Wikipedia's alleged political leanings, Voz branded the platform as “Wokepedia”, saying it is the world's largest encyclopedia in one sentence, and it "now resembles a sort of Orwellian authority" in the next one.

At this point, it's like watching a rerun. Every few weeks, a new outlet accuses the free encyclopedia of the mind-numbing non-napping known as part of the politics of perception. Next, Matt Walsh, a commentator for The Daily Wire (not the most reliable source, to be honest), described a scene from his own movie Am I Racist?, where a supposed "white-guilt" group tries to get him arrested by reading the Wikipedia article about him to the police. See previous Signpost coverage on Walsh's commentary. – JSG, S

WISH-TV reported on the latest edition of WikiConference North America, which took place in Indianapolis from October 3 to October 6. The station interviewed Nigerian Wikimedian James Popoola—a frequent contributor to Wikidata and The WikiVibrance Project—Justin Clark, the digital initiatives director at the Indiana State Library, and the conference's organizer, Dominic Byrd-McDevitt. As further highlighted in the article, the mayor of Indianapolis, Joe Hogsett, proclaimed October 4 as "Wikipedia Day", while the IU Indianapolis Library received $280,000 from the local Library Fund in order to improve information on Indiana's digital heritage on Wikipedia. – S

Devex reported on how Wikipedia's army of human volunteers is being hailed as its greatest weapon against the rise of artificial intelligence. Interviewed by the platform, Wikimedia Foundation CEO Maryana Iskander emphasized that while AI churns out a lot of "slop" by prioritizing speed over accuracy, Wikipedia's crowdsourced approach has kept it a beacon of reliable information. Despite the hype surrounding the AI's quick progress, Iskander notes that Wikipedia's human editors remain cool and confident, because, well, it turns out that good old-fashioned community curation still works.

Additionally, 404 Media has picked up on the formation of WikiProject AI Cleanup (see previous Signpost coverage), created by volunteers to tackle a growing problem: AI-generated content that introduces errors or misleading information into articles. ExtremeTech also highlighted the surge of "AI-generated garbage," including a fabricated article about a non-existent Ottoman fortress and incorrect information added to existing articles. Even with the challenges posed by AI, at least this human editor is confident that community-driven curation will remain the best guarantor of quality. – JSG

Ashley Rindsberg shows "How Wikipedia is Becoming a Massive Pay-to-Play Scheme" at Pirate Wires, writing that "a boomtown industry feeding an insatiable demand for services like article creation, editing, management and deletion has emerged." One of his first targets is Pakistani company Abtach, which is reportedly tied to "at least 130 different Wikipedia editing front companies that operate under domains like Wikicreatorsinc.com, Wikicreation.services, Wikipedia Pro, Wikipedia Legends, and USAwikispecialists.com." See this story in The Signpost for more details.

For examples of slightly more sophisticated paid editing, Rindsberg provides links to articles about British investment immigration consultancies, Canadian frozen foods producers, cellulite-busting self-massage accessories, custom T-shirt retailers, Swedish online travel agencies, German disinfectant brands, industrial waste management companies, RegTech software firms, as well as packaging producers, electronic device recyclers, and self-storage chains. That's just what he considers "black hat" paid editing.

The more sophisticated "white hat" editors are linked to articles on Bain & Co, Yelp, Qualcomm, Kaspersky Lab, software company Forcepoint, the RSA Conference, as well as a New York Times exec and corporate clients like Reddit, MetLife, Accenture, Intel, IBM, Hubspot, Hilton, Vox Media, Dick's Sporting Goods, United Airlines, Amdocs, Gallup, Allergan, Breyers, Vimeo and Waymo.

Along the way, he mentions several of the better known scandals involving paid editing, including MyWikiBiz (see related Signpost coverage), Legal Morning (see previous Signpost coverage) and WhiteHatWiki (see see previous Signpost coverage). It's an excellent introduction to paid editing on Wikipedia, and you shouldn't be surprised if you see some of these names again.

In the words of Rindsberg, "the question this leaves us asking is whether we can really apply the historic term 'encyclopedia' to a sprawling network of thousands of articles carefully pruned by the PR departments of billion dollar companies, or if Wikipedia is something else entirely." – S

A crossword published in USA Today by constructor Ada Nicolle featured an unexpected nod to Wikipedia. One clue highlighted Annie Rauwerda, founder of the popular Depths of Wikipedia accounts on Instagram and other social media platforms, which showcase quirky and obscure content from the site. Other clues celebrated Icelandic-Canadian heritage, such as a reference to Gimli, Manitoba—the Canadian town with the highest population of Icelanders outside of Iceland itself.

In a recent interview with The Michigan Daily, Rauwerda reflected on how her account has grown into a hub for curious readers and Wikipedia enthusiasts, emphasizing the thrill of discovering—and correcting—obscure content. She also discussed the unique challenges of maintaining the account while navigating Wikipedia's complex editing rules, and the role her platform plays in demystifying the editing process and encouraging new contributors. From clickbait headlines to deep dives into obscure history, Depths of Wikipedia has evolved into a cultural phenomenon that brings new visibility to the weird and wonderful corners of our beloved encyclopedia. – JSG

According to a report by Stefan Weber, an Austrian media researcher noted for his work as a "plagiarism hunter", current US Vice President and presidential candidate Kamala Harris "copy-pasted a Wikipedia article" into her 2009 book Smart on Crime (co-authored with Joan O'C. Hamilton), alongside plagiarism from other sources. The allegations were publicized by US conservative activist Christopher Rufo in what may be intended as an October surprise ahead of the November 5 presidential election. (Rufo and a co-author had previously investigated plagiarism in academic publications by Harvard University president Claudine Gay, contributing to her resignation.)

The New York Times summarized a different plagiarism expert's initial reaction as stating that "the errors were not serious, given the size of the [book]." However, a day later the same expert (Jonathan Bailey, who runs a site called "Plagiarism Today") followed up to clarify that his statement had been based on the examples provided in Rufo's post only, and that after reviewing the full dossier by Weber, he judged "the case [to be] more serious than I commented to the New York Times" although he still maintained that "the pattern points to sloppy writing habits, not a malicious intent to defraud." Specifically, Bailey stated that

"The most serious allegation concerns Wikipedia. Harris’ book contained roughly two paragraphs copied from Wikipedia without citation. To be clear, that is plagiarism. It’s compounded by the fact that Wikipedia is typically not seen as a reliable source, and, according to Weber, there was an error in the information."

The Harris campaign rejected the plagiarism allegations outright as "a partisan attempt to weaponize a 15-year-old work", as summarized in a Washington Post article. Conversely, Harris' political opponent JD Vance seized on them by posting on Twitter/X "I wrote my own book, unlike Kamala Harris, who copied hers from Wikipedia." The WaPo article also reported an unnamed source as claiming that while Harris' involvement in the book included "reviewing drafts", she "was not involved in the formatting of outside excerpts and citations".

The Wikipedia article in question is Midtown Community Court, which according to Weber was plagiarized in this revision. – H

• Wikipedia paces to victory: In an unexpected twist, Wikipedia isn't just for late-night research dives anymore—it's also the name of a harness racing horse! Standardbred Canada reports the nobly named Standardbred pacer, a three-year-old son of Mcwicked, came from five lengths back in the stretch to win an Ontario Sire Stakes Gold division race, overcoming Crush Kill Destroy, Unrivaled Hanover, Legal Attack and Chain Gang. The next time someone will say “Wikipedia is fast,” they won't just be talking about the servers.
• Camille Herron's race for wiki supremacy: Multiple magazines and news outlets, including Runner's World, Canadian Running Magazine, Women's Health, and Athletics Illustrated all reported that ultrarunner Camille Herron went from trails to keyboards, getting flagged for editing her own Wikipedia page to emphasize her achievements, as well as removing some of the achievements in the article about Kilian Jornet — turning this into a race of reputation, rather than endurance. Wikipedia's neutrality got a workout as Herron's edits sparked controversy over athletes curating their own narratives. Actions have consequences, though—Herron was swiftly dropped by her primary sponsor, Lululemon, making this yet another high-stakes lesson in online reputation management gone wrong.
• Recent research piece picked up: The Arabian Post picked up on a study comparing how people perceive credibility between Wikipedia, ChatGPT, and Alexa. While their coverage didn't mention The Signpost directly, it did link back to our Recent research piece from two issues ago. So, even though we weren't named, The Signpost served as a bridge between the original Nature article and international media coverage.
• Wikipedia "winner" goes viral!: A viral screenshot of a Wikipedia page claiming Abhijeet Sawant as the winner of Bigg Boss Marathi season 5 has stirred up social media, leading to debates on whether it's true or just another case of Wikipedia vandalism. The buzz even got coverage by The Free Press Journal, which pointed out the ongoing confusion.
• Goa's newest archive: Indian outlet Goamankat Times highlighted local Wikipedian Tanmay Pereira Naik's efforts to expand and improve articles related to Goa on Wikipedia, which include everything from documenting local history and politics to covering underrepresented topics like traditional cuisine and Goan personalities.
• Wikipedia training camp: Nigerian newspaper The Punch reported that a group of 20 participants attended a Wikipedia editing workshop in Nigeria, hosted by the Tyap Wikipedia User Group. The workshop aimed to improve digital literacy and expand representation of Nigerian topics on the encyclopedia.
• Portland City Auditor reopens Gonzalez case: After initially determining there was insufficient evidence to determine that the Portland, Oregon mayoral candidate broke campaign finance law by hiring a firm to burnish his Wikipedia page, the city auditor reopened the case. KOIN reports that "Gonzalez's office had paid a company called Codename Enterprises, operating under WhiteHatWiki, to assist with edits and contentious matters on Gonzalez's Wikipedia page." See also the report in The Oregonian and previous Signpost coverage here, here, and here. The next installment of this long running saga is expected with the release of the revised City Auditor's report on or before Halloween.
• "Under the radar" page move rankles: Scholar Asaf Romirowsky criticised the article on Israeli apartheid in an interview for the Jewish Journal, saying that the Wikipedia article's title and opening paragraph are "a work of fiction"; the same article also claims that the Requested move to the current title was accomplished "under the radar". This is not the first page move controversy reported by The Signpost; find more about previous instances here, and here.
• Spotlight on Bangladeshi Authors: An article in The Daily Star explored how Wikipedia categorizes prominent Bangladeshi literary figures like Abbasuddin Ahmed and Humayun Kabir. The piece highlighted issues with labeling writers based solely on linguistic or geographical ties, which sometimes results in misclassifications—such as identifying them only as "Bengali" or "Indian", rather than acknowledging their heritage. The article argues for more precise representation on Wikipedia to honor these authors' contributions to the cultural and literary history of Bangladesh, and to avoid losing the nuances of their identities due to historical geopolitical changes.
• High interest in 5784: In its year-end Internet culture wrapup, the Jerusalem Post noted that "Israel's Wikipedia entry... [got] approximately 14,769,946 views in a year, more than Israel's entire population." Other Wikipedia pageviews were noted, as well as Google search and YouTube popularity, to gauge interest in various Israel-related topics (and some Jewish public figures in the US and elsewhere).

• If sockfarms don't succeed, try, try again: University of Sydney researcher Olga Boichak writes in Foreign Policy about "How Russia Invaded Wikipedia", commenting on the nationwide splinternet and the creation of Ruviki, the government-backed fork of ru.wiki.
• The silent majority?: The Wikipedia article for the 2024 Silent Hill 2 remake was recently locked after persistent disruptive editing aimed at lowering review scores. According to IGN, the edits skewed the ratings from various sources to make the remake seem less well-received. In a move straight out of the game's own playbook, admins had to put the page under "lock and key"—much like the puzzles and locked doors players navigate in Silent Hill 2—to keep the chaos at bay.
• Are press releases news?: Outlets such as Business Standard and NewsTap both covered the Wikimedia Technology Summit 2024, with the former publishing a press release and the latter seemingly slightly rewriting it. The summit was hosted with IIIT Hyderabad and focused on enhancing inclusivity and innovation within Wikipedia and other Wikimedia projects. The press release highlighted sessions on developing tools to better support diverse communities and promoting equal participation in tech development.
• Check the stats first: UnHerd reported that Wikipedia recently renamed its article on the UK's grooming gang scandals to "Grooming gang moral panic in the United Kingdom", sparking controversy. The change, which was based on discussions citing sources like a 2020 Home Office report, was intended to reflect how the media and public framed the issue. Critics argue that the new title downplays the severity of the abuse, causing debate over whether Wikipedia is succumbing to political correctness at the cost of historical accuracy. The Wikipedia article highlights another government report and statistics that show that White men in the UK sexually abuse children more than Muslims or South Asians.
• The court reads: Following the last issue's coverage of the ongoing legal case between India's Asian News Network and the Wikimedia Foundation, a dedicated article was written and published on 10 October 2024. However, the article has now been mentioned in the court on 14 October, and wasn't exactly well-received, as reported by Bar and Bench and Live Law. On 16 October, the court pretty much ordered a deletion of the article, which was snow-kept at a short-lived AfD.

In early April, a few of us in the WikiCup thread of the Wikimedia Discord were discussing the problem of systemic bias on Wikipedia regarding postcolonial, economically emerging, and/or non-Anglophone countries. This is a common topic of complaint among both Wikipedians and non-Wikipedians, and for good reason; it's still a major weak spot for the English Wikipedia community. Epicgenius, one of the judges of this year's WikiCup, first suggested a contest focusing on developing countries on April 5th. Over the next few days, we brainstormed different ways to run this hypothetical contest: the name, the inclusion criteria, and scoring. Quite a few people chimed in, and Ixtal and I decided to BOLDly set up a project page, aiming for July–September to run the contest. Soon thereafter, we decided to recruit a third coordinator, TechnoSquirrel69; that ended up being a very good decision.

Initially we were thinking about calling it the "Global South WikiContest", but we soon realized that "global south" is too nebulous and subjective of a grouping, so we settled on the IMF's list of developing countries as a basis. We also decided to give bonus points for "least-developed" countries and for "higher-level" articles about broad topics relating to these developing countries, as both of those factors tend to be more challenging to write about. Some fun themed awards were also of course necessary.

While planning this back in April, we weren't expecting more than maybe twenty participants. Interest picked up in the Discord as July approached, and we refined the details on the contest's talk page. When we actually opened up for signups and TechnoSquirrel69 set up a watchlist notice, we ended up with around eighty participants, far more than we were prepared for. This actually ended up being fine as many of those who signed up didn't submit anything, but it was very exciting to us coordinators, as it indicated a desire to work on systemic bias.

The course of the contest was largely hiccup-free, and we didn't have to decline very many submissions. We weren't able to get a bot up and working in time either, so we manually updated the leaderboard as submissions came in. As I mentioned above, I was very happy that we had brought on TechnoSquirrel as a third coordinator, because I moved across the Atlantic Ocean at the end of August and was unable to properly keep up with submissions and discussion for a time.

After the contest wrapped up on September 30th, we opened a discussion on the talk page to get some feedback, and we've already gotten some great criticism, praise, and ideas.

I'm very happy we did this experimental contest. We ended up with three new featured articles (Qalaherriaq, Siege of Baghdad, and Genghis Khan!), ten new featured lists, 88 new good articles, and a truckload of article reviews and DYK nominations, all relating to parts of the world underserved by our encyclopedia—that's nothing to turn one's nose up at! I'd like to express my appreciation for my fellow coordinators, and Chaotic Enby for making our special "belt buckles" inspired by the map's resemblance to one.

Here are some super-basic stats! There's some overcount for sure (I am merely human, and didn't put in any effort to split dual-GA nominations and dual-DYK nominations), but here's some basic estimates for everybody to enjoy! (In terms of creations. Sorry, reviewers, the overcount would have gotten too bad if I'd included your activities!)

Overall, members of the drive managed to create or substantially improve content concerning a grand total of 86 different countries. 86! That's pretty impressive and, coincidentally, the telephone country code for China. A pretty amazing coincidence, considering which country's coverage members of the drive improved the most.

The top five most-improved countries were:

1. China, with 1 featured list, 6 GAs, and 13 DYK appearances
2. India, with 1 featured list, 15 GAs, and 1 ITN appearance
3. Philippines, with 5 GAs and 7 DYK appearances
4. North Korea, with 2 GAs and 9 DYK appearances
5. Kiribati, with 10 GAs and 4 DYK appearances

and then, more individually:

By terms of GAs submitted:

1. India with 15
2. Tied: China, Kiribati, and Haiti with 6
3. Philippines with 5

By terms of DYKs submitted:

1. China with 13
2. North Korea with 9
3. Philippines with 7

But just because we improved those countries so much, doesn't mean we forgot about the rest. Members of the drive found time to get either a single GA, FA, or DYK for 49 individual countries. Most impressive of those were the two countries that got "only" a featured article: Iraq and Greenland, brought to you by AirshipJungleman29 and Generalissima. If you haven't already, you should take the time to read both of their articles. They're some of the best that Wikipedia has to offer.

Members of the drive also managed to get 3 important deaths (those of presenter Aparna, wrestler Afa Anoaʻi, and former PM of Lebanon Salim Al-Huss) featured In The News. These appearances were brought to us by two editors in particular: Vacant0 and Jaguarnik. Everybody else, this is who you have to beat next year!

We also got plenty of lists: 10 in total, covering 8 different countries! (Three of these lists covered Ukraine) To the uninitiated such as myself, the featured list process may appear somewhat strange, but not to editors such as MPGuy2824, with their impressive 3 FLS and 7 FLRs, Vanderwaalforces, who wrote Wikipedia's first ever Nigeria-related FL, Dantheanimator with the 3 Ukraine-related FLs, and 48JCL with their fascinating List of World Heritage Sites in Botswana.

Let's look at a breakdown by continent:

The continent we worked on the most was Africa, with 36 countries represented. Second place was Asia with 19, followed by the Americas with 15. Oceania got 10, and Europe came in last with 7 countries represented. I bet no other Wiki-drive has ever had that result before!

In terms of content, a special shout-out has to go to Vigilantcosmicpenguin for their series about abortion in various African countries- not only did they bring Abortion in Africa through the DYK process, but they brought 11 different country-specific articles through as well. Not satisfied with that, they also had to pause and turn Abortion in Sierra Leone into a Good Article too! In terms of eliminating systemic bias, the importance of their contributions cannot be overstated.

And while we're talking about Olympic feats, let's pause and appreciate the efforts of Arconning. They made 1 Olympic related Featured List, 7 Olympic related good articles, and then went above and beyond to get 3 of those articles through the DYK process as well. Talk about gold medalists!

But speaking of sports, BeanieFan11 saw the need to improve the articles on athletes from developing nations, and they more than rose to the challenge. But in between their GAs on NFL members, and DYKs on various athletes (did you know that Olympian judokan Valentin Houinato, from Benin, is also a journalist?), they also substantially expanded or created articles on three different politicians! One of those politicians being current Prime Minister of Equatorial Guinea, Manuel Osa Nsue Nsua.

While we're learning so many cool things, let me tell you all how our coverage of Kiribati improved so much. Thebiguglyalien is responsible for five of our new Kiribati-themed good articles, and all four of its DYK appearances. And did you know how they got all those DYK appearances in just one hook? Find out for yourself here!

In a similar vein, if India placed so high, it's only because of dedicated writers like Magentic Manifestations, who made India-related articles about a wide array of topics, from dairy engineers to crocodile trusts, into some Good-capital-G Articles.

And while I may not have focused on reviews too much, I am going to take a minute to highlight the contributions of our most prolific reviewer, Simongraham. They wrote Good Articles on seven different species and genuses of arthropods (mostly jumping spiders, from what it seems), and two different Soviet ships. Those with arachnophobia beware, but those who create species stubs, be even warier, for this is the type of editor we all wish we had more of.

Hopefully all the spider-phobes haven't left by now, because who isn't going to love this next topic? That's right, who wants to read about chocolate? Yue managed to write three articles on chocolate production, chocolate smuggling, and a chocolate manufacturer, and they even made one into a Good Article. But they didn't stop there! Those of you interested in the flags of the world, you need look no further than their new Good Articles on the flags of Togo, North Korea, São Tomé and Príncipe, or Rwanda.

This Wikicup may not have seen many—or, actually, any—Good Topics, but one user came close: Chipmunkdavis! They wrote five Good Articles about fisheries in the Philippines, after finding out just how lacking we were. Generalissima's been prodding them to turn the entire thing into a Good Topic. And who knows, hopefully she'll have convinced them by next year? No pressure or anything, but all eyes are on you now, Chipmunkdavis!

Di did what they do best and told us all about the time "that German officials exiled the Samoan king from his own kingdom". In light of the recent Olympics, Riley1012 helped to expand our coverage of artistic gymnasts from around the world. TheNuggeteer brought you coverage of Philippines storms, Averageuntitleduser wrote Good Articles on important Haitian women, and Cambalachero proved that you can write a Good Article on pop-culture topics after all! PerfectSoundWhatever introduced us to the life and works of Kenyan musician KMRU, Noorullah21 wrote an impressive 1k+ word GA on the Khalji Revolution, and Skyshifter got meta and treated us to hoax article! As in, a good article about a hoax article. You'll just have to read it for yourself. Fritzmann2002 specialised in article and list reviews, but took the time to write some good articles on plants from Syria and Turkey.

Some of us were busy during the contest, however—but in a drive such as this, literally every attempt counts. With that in mind, let's take some time to appreciate Sohom Datta's review of a GA about voting in India, TappyTurtle's review of the article about the Wikipedia hoax, Zanahary review of an article about an Indian god of war, SunTunnels's article about a South African speed-climber, and Queen of Hearts's article on a crab named after a League of Legends character.

I'd also like to say thank you to our lovely co-ordinators, Sawyer777, Ixtal, and TechnoSquirrel69. It may sound cheesy, but it's true—none of this would have happened without your ideas and dedication.

...if it doesn't look like anything, squint harder. As always, click on the squares to type your letters, but don't press enter. You've been warned.