iiinova — where your digital self keeps evolving

Patrick Gildersleve @gildersleve.uk · 15/01/2026

Happy 25th birthday to Wikipedia! 🥳 A fitting moment to share 1. Their great site to mark the occasion: wikipedia25.org 2. A paper in Big Data & Society, published over the winter break, where we develop Wikipedia as a “Refractive Dataset”, led by @beeeeeers.bsky.social: doi.org/10.1177/2053...

$A paper screenshot: Refractive datasets as a sensemaking methodology in closed data ecosystems Anna Beers, Viviane Ito, Agustin Orozco, Patrick Gildersleve, Pablo Aragón, and Francesca Tripodi Abstract As digital platforms restrict their APIs, researchers face diminishing options for studying social phenomena in digital environments. During what has been called the post-API era, researchers have found themselves looking for reliable data sources in an unreliable and frequently changing platform data ecosystem. In this context, we propose analyzing refractive datasets as a methodology for researchers to understand the dynamics of closed data platforms. Refractive datasets come from platforms with relatively more open data policies, and their analysis sheds light on platforms with more restrictive data policies. Like a prism, refractive datasets reflect but also transform data-based phenomena unfolding on closed platforms. Using refractive datasets from Wikipedia and Google Trends, we present three studies to demonstrate our methodology. We first show how refractive data from Wikipedia's multiple language editions can be used to understand a fractured global platform ecosystem in a case study of hydroxychloroquine, a purported COVID-19 medicine. Second, we use Google Trends to show how similar refractive analyses can be used to understand information lost to platform deletion, in a profile of an online panic over the drug brand Galaxy Gas. Finally, we show how Wikipedia data can be used as a grounding point for a refractive analysis of how new generative algorithms reproduce and distort data across the social web. We discuss how refractive datasets can be a way for researchers to “sensemake” in increasingly opaque big data environments, enabling interpretivist analyses which aim to generate new hypotheses rather than verify existing claims.$

072

Patrick Gildersleve @gildersleve.uk · 03/09/2025

A depressing juxtaposition of news stories, there is such a disconnect in the national conversation. A public figure directly calling for violence against trans people is met with police sympathy and instead we get phone ban campaigns to tackle online harms 🙃.

A screenshot of the BBC news app showing two stories. The first “Met chief calls for law change after Graham Linehan arrest”, with a photo of Graham Linehan. The second “Watch: Mother of Brianna Ghey calls for smartphone ban in schools”, with an image showing Brianna and Esther Ghey.

170

Patrick Gildersleve @gildersleve.uk · 22/07/2025

For even more WikiResearch at #IC2S2, I'm presenting our work on WikiReddit on Wednesday 14:30 in 'Social Media II' 🧑‍💻 I'll explain how these complementary platforms are powered by the magic of ✨𝐩𝐞𝐝𝐚𝐧𝐭𝐫𝐲✨, and how our dataset can be used to study cross-platform flows of information and attention 🔀

060

Patrick Gildersleve @gildersleve.uk · 22/07/2025

Great session on Wikipedia at #IC2S2 with some very cool research. Strong themes of multilingual analysis and coordination dynamics @smfsamir.bsky.social @feloe.bsky.social

1100

Patrick Gildersleve @gildersleve.uk · 26/06/2025

Had a great time meeting everyone and seeing all the interesting work @icwsm.bsky.social. I presented our study on the Wikireddit dataset - exploring Wikipedia’s role in fact-checking, discussion, and cross-platform attention on the web. Thank you to the organisers! 📄: ojs.aaai.org/index.php/IC...

Presenter (Patrick Gildersleve) in front of a screen summarising the WikiReddit Dataset project. The slide describes it as "Every Wikipedia mention and link on Reddit, 2020-2023", includes some example usage, describes the scale of the dataset, and offers suggested use cases.

083

Patrick Gildersleve @gildersleve.uk · 24/06/2025

Important starter pack research in action! I received lots of follow notifications during this talk - hopefully this work prompts more ICWSM activity here!

050

Patrick Gildersleve @gildersleve.uk · 24/06/2025

Very cool work at #ICWSM tracking information manipulation in response to national regulation on the Russian Wikipedia fork From Mykola Trokhymovych, @elaragon.bsky.social, @e-migrante.bsky.social & others

Presenter in front of screen at ICWSM.

Slide content:

Summary

First study of how original Wikipedia content has been forked and manipulated to meet the requirements of a national regulation.
• RWFork edits target top visited and controversial pages
• Editors follow office-hour activity patterns
• Changes are systematic, mostly focused on the Russo-Ukrainian War

RWFork dataset is available for further study:
https://zenodo.org/records/15073728

031

Patrick Gildersleve @gildersleve.uk · 10/02/2025

On exploratory analysis, we find: 📉 Declining activity in posts, but stable performance of Wikipedia content on Reddit 🔗 Strong correlations between Wikipedia and Reddit activity 🌍 Intriguing asymmetric patterns of cross-lingual linking, dominated by English

Plot showing the daily count of posts and comments that mention Wikipedia (in text or as a link) over 2020-2023. The number of comments is steady at around 6000-7000, whereas the number of posts steadily decreases from around 300 to 200.

Figure showing the daily page views to Wikipedia articles on the day of posting and in the week after posting relative to the week before posting as a KDE plot. A small number of points in the extremes of the distributions are cut for visual clarity. We find that for links in Reddit posts, we observe a 45% increase in page views on the day of posting, and a 6% increase in the week after posting, as compared to the week
before posting. For links in Reddit comments, we observe a 45% increase in page views on the day of posting, and a 5% increase in the week after posting, as compared to the week before posting.

Figure showing the proportion of links to English Wikipedia from non-English Reddit posts vs the total number of links in that post language. Larger languages such as Chinese, German, French link to English Wikipedia around 20% of the time, whereas for smaller languages, English is used a majority of the time.

130

Patrick Gildersleve @gildersleve.uk · 10/02/2025

📂 The SQL database contains: - All Reddit posts and comments mentioning Wikipedia 2020-23, including hyperlinks, hashed IDs, and metadata. - Edit history and page view activity for Wikipedia articles at the time of posting, page IDs, and redirects.

A table from the paper detailing a summary of all Wikipedia mentions, those including links, and those that map to Wikipedia articles by
Reddit format.

| Posts | Comments | Total
# mentioning Wikipedia | 335,897 | 10,264,340 | 10,600,237
# with Wikipedia links | 286,359 | 9,465,316 | 9,751,675
Total # of Wikipedia links | 658,493 | 11,573,36 | 12,231,860
# unique Wikipedia links | 295,439 | 1,890,497 | 1,954,003
# unique Wikipedia articles | 252,846 | 1,196,494 | 1,260,479

110

Patrick Gildersleve @gildersleve.uk · 10/02/2025

📄 Excited to share a new dataset preprint! WikiReddit: Tracing Information and Attention Flows Between Online Platforms A collaboration with @beeeeeers.bsky.social, Viviane Ito, Agustin Orozco, and @ftripodi.bsky.social at @citap.bsky.social .

A screenshot of the start of the paper "WikiReddit: Tracing Information and Attention Flows Between Online Platforms"

1176

Patrick Gildersleve @gildersleve.uk · 14/08/2024

🔍With FORUM, we find dramatic differences in how different, even simple, algorithms prioritise different kinds of comments, as measured by their sentiment, readability, lexical diversity, and similarity to the article

100

Patrick Gildersleve @gildersleve.uk · 14/08/2024

📈We develop the Feature-Oriented Ranking Utility Metric (FORUM 🥁) to evaluate how well a ranking algorithm (most upvoted, chronological, etc) returns comments by a certain feature (sentiment, readability, etc) compared to best/worst-possible and random baseline

100

Patrick Gildersleve @gildersleve.uk · 14/08/2024

🕵️With data from @derStandardat, we find, compared to readers, journalists prefer more positive, timely, complex, direct responses, while readers favour comments similar to article content from elite authors

100

Patrick Gildersleve @gildersleve.uk · 14/08/2024

🚨New preprint! 📰The News Comment Gap and Algorithmic Agenda Setting in Online Forums w/ Flora Böwing We study: 1) The differences in comment preferences between journalists and readers 2) How comment ranking algorithms prioritise different kinds of comments to display to users 🧵

262

Patrick Gildersleve @gildersleve.uk · 18/07/2024

Excited to be at #IC2S2 2024! I’ll be presenting “The News Comment Gap and Algorithmic Agenda Setting in Online Forums” in session 3C, Friday 11am. Say hi if you want to connect👋

Interior of the (empty) Irvine auditorium at UPenn. “IC2S2” is displayed on a large projector screen

120

Patrick Gildersleve @gildersleve.uk · 20/12/2023

Super proud of my MSc supervisees who graduated from LSE last week. All achieved distinctions in their dissertations and degrees, including the top MSc project award! 🎉

120