"Investigating extreme cases in Wikipedia talk pages: Some insights on user behaviours"
uplopen.com/chapters/e…
e.g. "the most prolific users, the longest threads (in terms of total duration, number of posts or number of distinct users involved) and the longest monologues"

"Demographic disparity in Wikipedia coverage: a global perspective" (top 12 languages) epjdatascience.springeropen.com/articles/1…
- Women slightly overrepresented (not underrepresented) among living article subjects since ~2015, but still have shorter articles
- Developing countries overrepresented

"Investigating How LLMs Impact Participation in [Wikipedia]" (interviewing 16 editors) https://arxiv.org/abs/2509.07819v1

ChatGPT etc "enhance contribution quality for experienced editors" & "lower entry barriers for newcomers", but newbies struggle to align LLM outputs w Wikipedia policies

"Invitation to give feedback to Wikimedia Research Fund 2024-2025 proposals" until May 12 lists.wikimedia.org/hyperkitty/l...
(13 proposals under review, with budgets ranging from $8,571 to $149,976 USD)

In the new edition of our monthly newsletter:
* How readers use Wikipedia health content
* Scholars are generally happy with how their papers are cited on Wikipedia
* Several other papers about references on Wikipedia
and more:
meta.wikimedia.org/wiki/Researc...

In the March issue of our research newsletter:
* Flagged Revisions: Explaining the disappointing history of a community-requested software feature
* A roundup of several recent papers investigating the impact of ChatGPT on Wikipedia so far
and more: meta.wikimedia.org/wiki/Researc...

In the new issue of our monthly newsletter:
▸ What's known about how readers navigate Wikipedia
▸ Italian Wikipedia is the hardest to read
▸ "open access articles are extensively and increasingly more cited in Wikipedia" than those behind a paywall
meta.wikimedia.org/wiki/Researc...

In the latest issue of our newsletter:
▸ GPT-4 is better at writing edit summaries than human Wikipedia editors
and other new research findings
meta.wikimedia.org/wiki/Researc...

In the latest issue of our monthly newsletter:
▸ "Wikipedia editors are quite prosocial", especially community "superstars" – but editors motivated by "social image" may put quantity over quality
And other recent research publications
meta.wikimedia.org/wiki/Researc...

Topics in the newsletter's September issue were:
▸WikiCrow: Article-writing AI is less "prone to reasoning errors (or hallucinations)" than human Wikipedia editors
▸Using Wikipedia's categories and lists to build a knowledge graph separate from Wikidata
and more meta.wikimedia.org/wiki/Researc...

And ICYMI, our October newsletter:
▸ "Rise of AI-Generated Content": The AI-Wikipedia ouroboros begins to devour itself - or does it?
▸ Using LLMs to improve coverage of underrepresented groups
▸ Feminist critiques of Wikipedia's "be bold" and lack of "dissensus"
meta.wikimedia.org/wiki/Researc...

In the latest issue of our newsletter on research about Wikipedia, Wikidata and other Wikimedia projects:
▸ "SPINACH": LLM-based tool makes Wikidata's data more accessible, translating "challenging real-world questions" to SPARQL queries
and other new publications
meta.wikimedia.org/wiki/Researc...