what are the prospects of critical AI practices? we explore on friday 26th june, convened with @lboungr.bsky.social @hetsalwel.bsky.social @noortjem.bsky.social @dntrcc.bsky.social @esthrr.bsky.social Daniela van Geenen, Tobias Blanke, Sabine Niederer & more! ✨ jonathangray.org/2026/06/19/c...

Really interesting. Liable for content! The court treated the AI overviews as Google's own content and rejected Google's argument that users were responsible for fact-checking the results themselves the-decoder.com/landmark-ger...

Landmark German ruling declares Google's AI Overviews are Google's own words and makes it liable for false answers

A German regional court has ruled that Google is directly liable for the content of its AI search overviews. According to the court, previous limited liability protections for search engine operators ...

the-decoder.com

Rich Sutton warned that computational scale ultimately outperforms human expertise. Does his "bitter lesson" hold true when reproducing our own custom methods with LLMs? We find: the bitter lesson is sometimes sweet, sometimes bitter, and sometimes inapplicable. deep-culture.org/the-occasion...

It’s not just training data. It’s what gets incentivized — that’s why LLMs often write exactly like this. This isn’t common Internet speech. It’s human response ratings — and it’s rewarding the system for the appearance of nuance. ✨⬇️

Appreciate the distinction between "representational" and "operational: readings laid out by @ddzeeuw.bsky.social + @hetsalwel.bsky.social here. Resonates with what @bentarnoff.com + I have been groping towards in Musk's relationship to social media and genAI. podcasts.apple.com/us/podcast/n...

Preview | Theorists Daniël de Zeeuw & Sal Hagen on the post-discursive web (NM 93) 2026

Podcast Episode · New Models · April 12 · 22m

podcasts.apple.com

What’s the difference between latent space, vector space, feature space, and embedding space? @louravn.bsky.social spelled it out for you in this short blog post deep-culture.org/whats-the-di...

🐈‍⬛🐈New 1.54 4CAT update 🐈🐈‍⬛ github.com/digitalmetho... 📽️ Show downloaded media in the Explorer. This works for media uploads and social media datasets if images/videos have been downloaded through 4CAT. ✏️ New "Upload annotations" processor that lets you add new data fields to existing datasets. (1/2)

Release v1.53 4CAT in bloom · digitalmethodsinitiative/4cat

This 4CAT updates mostly comprises updates and bug fixes for processors and data sources, as well as improvements to the Web UI (particularly the Explorer, which can now display downloaded dataset ...

github.com

Social media is no longer social. Most of it is passive viewing of videos and pictures from people we've never met. But we're still studying social media like it's 2010. We've entered the post-social media era — and research needs to catch up. osf.io/preprints/so... preprint w/ Richard Rogers

enjoyed this "wayback studio" moment at @internetarchive.eu with @anthbrtn.com @iiil.li @idilg.bsky.social @hetsalwel.bsky.social @esthrr.bsky.social @digitalmethods.net @d-eepculture.bsky.social @publicdatalab.bsky.social @digitalculturekcl.bsky.social & many more! ✨

Internet Archive Europe@internetarchive.eu · 6mo ago

How do researchers, journalists & artists work with the Wayback Machine? At a recent gathering in Amsterdam, we explored how web archives support digital investigations, platform research & media art — and how we can build better tools & collaborations. 👉 www.internetarchive.eu/2026/02/18/w...

Together with Bastian August and Brogan Latil, I spent decidedly too much time trying to understand how Andrew Tate remains present on YouTube after his two channels were suspended. Full paper in New Media & Society: journals.sagepub.com/doi/10.1177/...

A small 4CAT update just in time for the holidays - v1.52! 📊 Processors for regression analysis and dataset statistics 🧭 Test LLM prompts from our prompt library with the PromptCompass processor 🤖 Use Deepseek and Gemini 3 APIs with LLM processors 🐛 Many bug fixes! ➡️ github.com/digitalmetho...

How to move along with the audiovisual, vibey, and AI-infused feeds of today's platforms? In this new article with @ddzeeuw.bsky.social and Tommaso Venturini, we theorize 'digital rhythmanalysis' as a potential new quali-quantitative approach to measure and 'feel' rhythms online: bit.ly/pns-dra

4CAT now contains everything for quite sophisticated classification pipelines with humans and (local) LLMs! I myself used it to classify what types of migration were mentioned in Dutch news articles (see www.groene.nl/artikel/onde...). See all changes below 👇

4CAT@handle.invalid · 9mo ago

Just released: 4CAT 1.51 “The R is in the month” Highlights: 📊 Classification evaluation: subpar F1 scores, now in 4CAT 🤖 New LLM processor options #️⃣ Search by #hashtag on Telegram 🖼️ Image hasher: check for duplicate images! 🪰 A LOT of small bug fixes and improvements github.com/digitalmetho...

Voor @groene.nl onderzochten we hoe het migratiedebat de laatste jaren is gegroeid, verhard en verbreed. We analyseerden 150.000 nieuwsartikelen en 20.000 Tweede Kamer-spreekbeurten en gebruiken een (lokale!) LLM om te zien welke migratietypen de meeste aandacht kregen. Lees het artikel hieronder 👇

De Groene Amsterdammer@groene.nl · 9mo ago

Onderzoek | Spraakverwarring, overmatig gebruik van bepaalde woorden en dramatische metaforen: mede door de taal die de afgelopen 25 jaar werd gebruikt door politici en media zijn we problemen rond migratie gaan beschouwen als een crisis. buff.ly/hDiGiLV

Mijn analyse over een vrolijke liberaal in een steeds rechtser land staat nu alvast online. www.groene.nl/artikel/een-....

Rob Jetten wacht een onmogelijke taak

Rob Jetten loste zijn ideologische ankers, pakte het vooruitgangsoptimisme van de VVD en leende gemeenschapszin van het CDA. Zo dreef hij naar een steeds leger midden, waar hij iedereen moet bekoren d...

groene.nl

gefeliciteerd schier

Sal Hagen@hetsalwel.bsky.social · 3y ago

Nog steeds aan het bijkomen van de race tussen Schier en Vlieland

On election day, AI Overviews are still erratic, shown in ~10% of our queries. Some issue-related queries are unanswered but "which party is for defence" recommends VVD and JA21 at the top and includes outdated stances on JSF investments and the PvdA as a separate party. Nothing on NATO norms.

Sal Hagen@hetsalwel.bsky.social · 9mo ago

With @best.wel.moe and @nieuwsuur.nl, we studied Google's AI Overviews for political searches in the NL. We found that larger parties often benefit and that Google changed its use of AI Overviews *three times* during the run-up to the Dutch elections. Read the article here: nos.nl/nieuwsuur/ar...

Genuinely what the fuck is going on in Albania

With @best.wel.moe and @nieuwsuur.nl, we studied Google's AI Overviews for political searches in the NL. We found that larger parties often benefit and that Google changed its use of AI Overviews *three times* during the run-up to the Dutch elections. Read the article here: nos.nl/nieuwsuur/ar...

Google zet gekleurde AI-antwoorden midden in campagnetijd aan, uit, en weer aan

GroenLinks-PvdA, D66 en de VVD worden in de antwoorden op algemene politieke vragen het vaakst genoemd. Kleinere partijen komen veel minder vaak aan bod.

nos.nl

Keywords are not neutral. With Tobias Blanke, we investigate the power of keywords as interpretive tools in our work "Keywords in Digital Humanities: A critical assessment of computational techniques for mapping security and freedom in historical debates". You can read the article now on DHQ!

Super fun to contribute to this brief interview with Terminally Online, a student-led organization on New Media at the University of Amsterdam. We spoke about my wonderful team‘s research (@d-eepculture.bsky.social), who inspires me, and some of my own work. 🔗: www.linkedin.com/pulse/interv...

Interview with Louis Ravn

Louis Ravn is a PhD Candidate on Deep Culture at the University of Amsterdam, working in the fields of science and technology studies and critical data/AI studies. Moreover, he is assistant editor for...

linkedin.com

Dit is de omgekeerde wereld. De academische vrijheid wordt bedreigd doordat onze onderzoekers die zich uitspreken in het publieke debat over “omstreden” onderwerpen als snel het doelwit worden van online bedreigingen, met name uit extreemrechtse hoek.

'Nederlandse universiteiten schuwen debat met andersdenkenden'

Is er op Nederlandse universiteiten nog wel ruimte voor een open debat? "Radicale vrijheid van meningsuiting bestaat in Nederland niet."

nos.nl

🚨The Critical AI Seminar Series returns in 2025/2026 ✨ This year we will have four talks and are beyond excited to welcome @amoorelouise.bsky.social & @alexcampolo.bsky.social, Fabian Offert, Helene Ratner & @nannathylstrup.bsky.social and @thaophan.bsky.social ✨ Sign up: bit.ly/uvacriticala...

The last few weeks I worked 4CAT's LLM features. We're not free from Silicon Valley APIs yet, but you have more options in terms of providers (including Mistral) and it's also possible to use local LLMs through LM Studio and Ollama. Check it out:

4CAT@handle.invalid · last yr.

🧠 New: 'LLM Prompter' processor A much more robust and flexible way to analyse datasets with LLMs • More APIs: OpenAI, Google, Anthropic, and Mistral • Support for LM Studio and Ollama to use local (and free) LLMs • Batching of input values to increase speed and reduce cost • Structured outputs

🔎New paper! Guillén Torres and I study "enrichment" on Google and Bing as a form of content moderation. Main finding: enrichment can be erratic, with components like "AI Overviews" even being banned for certain political queries like those including "trump". Pre-print: ssrn.com/abstract=528...

Op nos.nl ging de laatste vijftien jaar even vaak over klimaatverandering als over Erik ten Hag. (2853 vs. 2766 artikelen/updates) @xrnl.bsky.social @nieuws.nos.nl

Tired of looking at Excel rows when investigating social media data? 4CAT's new Explorer feature is here to help!

4CAT@handle.invalid · last yr.

🔎The Explorer lets you examine datasets in a rich environment where social media posts are "reassembled" to look like their original appearance. You can sort data, make manual annotations like field notes, or let (LLM-)processors add categories and notes. How to use: github.com/digitalmetho...

yes een nieuwe kans voor Schier

Sal Hagen@hetsalwel.bsky.social · 3y ago

Nog steeds aan het bijkomen van de race tussen Schier en Vlieland

Na ons artikel uit 2021 over klimaatverslaggeving (www.groene.nl/artikel/we-h...) was het tijd voor een data-update. Gelukkig is meer aandacht voor de klimaatcrisis in de vier grootste kranten. Maar er is nog steeds een golfbeweging gaande en aandacht is de laatste maanden weer afgenomen.

Jaap Tielbeke@jaaptielbeke.bsky.social · last yr.

Het probleem is niet dat journalisten zich laten leiden door de belangen van vervuilende bedrijven, maar een mediadynamiek die het moeilijk maakt om de klimaatcrisis de aandacht te geven die ze verdient. www.groene.nl/artikel/onge...

We’re hiring: Lecturer in Cultural Data & AI and Archival & Information Studies (Media Studies, @uvahumanities.bsky.social)! Join @tpoell.bsky.social, @riederb.bsky.social, myself, & other wonderful colleagues in the Critical Data & AI team. Applications due 28 May. werkenbij.uva.nl/en/vacancies...

Vacancy — Lecturer in Cultural Data & AI and Archival & Information Studies

Would you like to teach on cutting-edge theories, methods, and questions in Cultural Data & AI and Archival & Information? Then we have the perfect challenge for you!

werkenbij.uva.nl