Hannes Cools

Hannes Cools

@hannescools.bsky.social

Assistant Professor, University of Amsterdam ๐Ÿ‡ณ๐Ÿ‡ฑ || Journalism, AI, Tech & innovation ๐Ÿ‘จโ€๐Ÿ’ป || Fellow Digital Democracy Centre ๐Ÿ‡ฉ๐Ÿ‡ฐ & Marie Curie Scholar ๐Ÿ‡ช๐Ÿ‡บ || Occasional Photographer || #FirstGen ๐ŸŒ https://hannescools.com/

1.2K followers231 following571 posts

Een inzending maakt dit duidelijk: volgens meerdere kranten gebruikte Kaatje Gijbels AI voor redigeren. Ze heeft haar inzending in samenspraak met jury teruggetrokken. Dit is kwalijk omdat 1. Er niets over AI gebruik in reglement staat, en in vele gevallen redigeren en proeflezen toegestaan isโ€ฆ

๐Ÿšจ ๐๐ž๐ฐ ๐ซ๐ž๐ฌ๐ž๐š๐ซ๐œ๐ก: ๐‡๐จ๐ฐ ๐š๐ซ๐ž ๐œ๐จ๐ฆ๐ฆ๐ฎ๐ง๐ข๐œ๐š๐ญ๐ข๐จ๐ง ๐ฌ๐œ๐ข๐ž๐ง๐ญ๐ข๐ฌ๐ญ๐ฌ ๐ฎ๐ฌ๐ข๐ง๐  ๐†๐ž๐ง๐€๐ˆ?๐Ÿšจ This is the first systematic, field-wide, international empirical investigation of generative AI practices in communication research. We surveyed 1,138 scholars across 77 countries + did 4 focus groups. Small thread with 5 key findings๐Ÿงต

View full thread

๐Ÿšจ ๐๐ž๐ฐ ๐ซ๐ž๐ฌ๐ž๐š๐ซ๐œ๐ก: ๐‡๐จ๐ฐ ๐š๐ซ๐ž ๐œ๐จ๐ฆ๐ฆ๐ฎ๐ง๐ข๐œ๐š๐ญ๐ข๐จ๐ง ๐ฌ๐œ๐ข๐ž๐ง๐ญ๐ข๐ฌ๐ญ๐ฌ ๐ฎ๐ฌ๐ข๐ง๐  ๐†๐ž๐ง๐€๐ˆ?๐Ÿšจ This is the first systematic, field-wide, international empirical investigation of generative AI practices in communication research. We surveyed 1,138 scholars across 77 countries + did 4 focus groups. Small thread with 5 key findings๐Ÿงต

View full thread

5.AI-guidelines lag behind practice. Guidelines frequently articulate high-level values (fairness, transparency, accountability) but lack operational detail, creating a principles-to-practices gap.

The bottom line of this study is: Generative AI is already part of communication science - often cautiously, sometimes quietly. The question is not whether it belongs there, but how we make its use ethical, transparent, & scientifically sound.

๐Ÿšจ ๐๐ž๐ฐ ๐ซ๐ž๐ฌ๐ž๐š๐ซ๐œ๐ก: ๐‡๐จ๐ฐ ๐š๐ซ๐ž ๐œ๐จ๐ฆ๐ฆ๐ฎ๐ง๐ข๐œ๐š๐ญ๐ข๐จ๐ง ๐ฌ๐œ๐ข๐ž๐ง๐ญ๐ข๐ฌ๐ญ๐ฌ ๐ฎ๐ฌ๐ข๐ง๐  ๐†๐ž๐ง๐€๐ˆ?๐Ÿšจ This is the first systematic, field-wide, international empirical investigation of generative AI practices in communication research. We surveyed 1,138 scholars across 77 countries + did 4 focus groups. Small thread with 5 key findings๐Ÿงต

View full thread

3.โ€˜AI shameโ€™ and underreporting. Scholars in our focus groups reported โ€˜AI-shameโ€™ and even desk rejections associated with disclosed genAI involvement, creating a chilling effect on transparency and open methodological dialogue.

4.Cultural and linguistic context matter. Institutions in regions where English is not dominant often endorse genAI for language support in scholarly writing, motivated by inclusion and equity. In English-speaking contexts, hesitancy is greater, driven by privacy, integrity, and authorship concerns.

๐Ÿšจ ๐๐ž๐ฐ ๐ซ๐ž๐ฌ๐ž๐š๐ซ๐œ๐ก: ๐‡๐จ๐ฐ ๐š๐ซ๐ž ๐œ๐จ๐ฆ๐ฆ๐ฎ๐ง๐ข๐œ๐š๐ญ๐ข๐จ๐ง ๐ฌ๐œ๐ข๐ž๐ง๐ญ๐ข๐ฌ๐ญ๐ฌ ๐ฎ๐ฌ๐ข๐ง๐  ๐†๐ž๐ง๐€๐ˆ?๐Ÿšจ This is the first systematic, field-wide, international empirical investigation of generative AI practices in communication research. We surveyed 1,138 scholars across 77 countries + did 4 focus groups. Small thread with 5 key findings๐Ÿงต

1. The paradox: Concern vs. Actual Use. Most researchers said genAI use in research is risky and inappropriate for many tasks. Yet they report genAI use across the research lifecycle with writing-related assistance as the most common use.

2. Institutional & editorial contradictions. Many unis encourage or tolerate genAI in teaching & some aspects of research, yet offer limited guidance on how to use it. At the same time, several journals and publishers impose restrictions, including explicit prohibitions on using genAI in peer review

In the meantime, OpenAI & Anthropic rolled out new models on Tuesday. Did anybody hear anything about this so-called AI slowdown lately? ๐Ÿ™‹โ€โ™‚๏ธ

๐Ÿ“š ๐Ÿ’ฅ ๐๐ž๐ฐ ๐›๐จ๐จ๐ค ๐œ๐จ๐ฆ๐ข๐ง๐  ๐จ๐ฎ๐ญ ๐ข๐ง ๐๐จ๐ฏ๐ž๐ฆ๐›๐ž๐ซ! ๐Ÿ“š ๐Ÿ’ฅ Our book '๐†๐ž๐ง๐ž๐ซ๐š๐ญ๐ข๐ฏ๐ž ๐€๐ˆ ๐š๐ง๐ ๐‰๐จ๐ฎ๐ซ๐ง๐š๐ฅ๐ข๐ฌ๐ฆ ๐…๐ซ๐จ๐ฆ ๐๐ซ๐จ๐๐ฎ๐œ๐ญ๐ข๐จ๐ง ๐ญ๐จ ๐๐ž๐ซ๐œ๐ž๐ฉ๐ญ๐ข๐จ๐ง' will come out almost 4 years after ChatGPT was introduced to the world. Together with @tjthomson.bsky.social & @ryanjthomas.bsky.social Preorder here: www.routledge.com/Generative-A...

View full thread

This book brings together fieldwork, conducted in nine countries (Australia, USA, UK, Germany, Norway, Switzerland, Denmark, the Netherlands, and France). And although GenAI is moving fast, we can clearly see some trends and patterns in how it's reshaping newsrooms and reader trust.

๐Ÿ“š ๐Ÿ’ฅ ๐๐ž๐ฐ ๐›๐จ๐จ๐ค ๐œ๐จ๐ฆ๐ข๐ง๐  ๐จ๐ฎ๐ญ ๐ข๐ง ๐๐จ๐ฏ๐ž๐ฆ๐›๐ž๐ซ! ๐Ÿ“š ๐Ÿ’ฅ Our book '๐†๐ž๐ง๐ž๐ซ๐š๐ญ๐ข๐ฏ๐ž ๐€๐ˆ ๐š๐ง๐ ๐‰๐จ๐ฎ๐ซ๐ง๐š๐ฅ๐ข๐ฌ๐ฆ ๐…๐ซ๐จ๐ฆ ๐๐ซ๐จ๐๐ฎ๐œ๐ญ๐ข๐จ๐ง ๐ญ๐จ ๐๐ž๐ซ๐œ๐ž๐ฉ๐ญ๐ข๐จ๐ง' will come out almost 4 years after ChatGPT was introduced to the world. Together with @tjthomson.bsky.social & @ryanjthomas.bsky.social Preorder here: www.routledge.com/Generative-A...

We dug into three big questions: ๐Ÿ”น What does generative AI actually produce in journalism? ๐Ÿ”น How do journalists themselves perceive and use these tools? ๐Ÿ”น How do news audiences react when they find out AI played a role?

Hannes Cools reposted

๐Ÿ“ข Are you interested in studying news consumption and trust in news for a PhD in Journalism, Media, and Communication? Iโ€™m looking for prospective PhD students interested in developing a project in this and applying for an Arts and Humanities Research Council Doctoral Landscape Award (AHRC DLA)

The academic grant system is extremely broken. 21.627 submissions... That is how many proposals the Marie Curie Postdoctoral Fellowships received this month. Two years ago it was 10,360. The budget did not move. The success rate went from around 16% to something close to 7%.

Hannes Cools reposted

Praat mee! Voor het @sidnfonds.bsky.social project denken we na over de vereisten van digitale autonomie. De centrale vraag: ๐–๐š๐ง๐ง๐ž๐ž๐ซ ๐ข๐ฌ ๐ƒ๐ž ๐‚๐จ๐ซ๐ซ๐ž๐ฌ๐ฉ๐จ๐ง๐๐ž๐ง๐ญ ๐ฏ๐จ๐ฅ๐ ๐ž๐ง๐ฌ ๐ฃ๐จ๐ฎ ๐ฏ๐จ๐ฅ๐๐จ๐ž๐ง๐๐ž '๐๐ข๐ ๐ข๐ญ๐š๐š๐ฅ ๐š๐ฎ๐ญ๐จ๐ง๐จ๐จ๐ฆ', ๐ณ๐จ๐ฐ๐ž๐ฅ ๐จ๐ฉ ๐ก๐ž๐ญ ๐ ๐ž๐›๐ข๐ž๐ ๐ฏ๐š๐ง ๐€๐ˆ ๐š๐ฅ๐ฌ ๐š๐ง๐๐ž๐ซ๐ž ๐›๐ข๐  ๐ญ๐ž๐œ๐ก? decorrespondent.nl/17247/wij-bo...

decorrespondent.nlWij bouwen met taalmodellen een betere zoekfunctie voor De Correspondent, onafhankelijk van big techGoogle, met zijn razendsnelle AI-taalmodellen, kan ons archief beter ontsluiten dan wij. Maar willen we dat? De modellen zijn vaak zonder toestemming getraind op journalistiek werk. Bovendien kapen ze...

Praat mee! Voor het @sidnfonds.bsky.social project denken we na over de vereisten van digitale autonomie. De centrale vraag: ๐–๐š๐ง๐ง๐ž๐ž๐ซ ๐ข๐ฌ ๐ƒ๐ž ๐‚๐จ๐ซ๐ซ๐ž๐ฌ๐ฉ๐จ๐ง๐๐ž๐ง๐ญ ๐ฏ๐จ๐ฅ๐ ๐ž๐ง๐ฌ ๐ฃ๐จ๐ฎ ๐ฏ๐จ๐ฅ๐๐จ๐ž๐ง๐๐ž '๐๐ข๐ ๐ข๐ญ๐š๐š๐ฅ ๐š๐ฎ๐ญ๐จ๐ง๐จ๐จ๐ฆ', ๐ณ๐จ๐ฐ๐ž๐ฅ ๐จ๐ฉ ๐ก๐ž๐ญ ๐ ๐ž๐›๐ข๐ž๐ ๐ฏ๐š๐ง ๐€๐ˆ ๐š๐ฅ๐ฌ ๐š๐ง๐๐ž๐ซ๐ž ๐›๐ข๐  ๐ญ๐ž๐œ๐ก? decorrespondent.nl/17247/wij-bo...

decorrespondent.nlWij bouwen met taalmodellen een betere zoekfunctie voor De Correspondent, onafhankelijk van big techGoogle, met zijn razendsnelle AI-taalmodellen, kan ons archief beter ontsluiten dan wij. Maar willen we dat? De modellen zijn vaak zonder toestemming getraind op journalistiek werk. Bovendien kapen ze...

Couple extra questions for Dario. - Does slower pace mean: No Giga-IPO? - Who will supervise the 3rd-party evaluators when at Anthropic? - Who will determine this 'slower pace'?

Hannes Cools@hannescools.bsky.social

Latest essay from one of the tech bros calling for a slowdown in AI development. Core: embedded 3rd-party evaluators inside AI labs, industry coordination on safety standards, eventual global agreements. Wild that 'maybe monitor and regulate this stuff' is the take. ๐Ÿซ  darioamodei.com/post/we-must...

Big tech as a bunch of windsocks. After Google, Apple renamed Lake Ontario because some president asked. Canadaโ€™s response? Signs that state: โ€œLake Ontario, Now and Alwaysโ€ #jesuiscanada

Other map services such as MapQuest have refused to change the lake's name. The latest stand has driven MapQuest to the number one spot on the US Apple App Store's top charts, as of 1 September, ahead of even long-time favourite ChatGPT. www.euronews.com/next/2026/09...

euronews.comApple joins Google in renaming Lake Ontario 'Lake America'Apple Maps now shows Lake Ontario as "Lake America" for US users, following Google's lead in complying with US President Donald Trump's order as trade talks with Canada sour.

๐ŸŽ™๏ธ ๐ˆ๐ง๐ญ๐ž๐ซ๐ฏ๐ข๐ž๐ฐ ๐จ๐ง ๐ง๐ž๐ฐ ๐ฉ๐ซ๐จ๐ฃ๐ž๐œ๐ญ '๐ƒ๐ข๐ ๐ข๐ญ๐š๐ฅ ๐€๐ฎ๐ญ๐จ๐ง๐จ๐ฆ๐ฒ ๐Ÿ๐จ๐ซ ๐ง๐ž๐ฐ๐ฌ ๐ฆ๐ž๐๐ข๐š' Media have largely lost control over their online distribution. Dependencies on Big Tech have only intensified with the rise of so media platforms, & now AI isn't making this 'distribution challenge' any easier www.uva.nl/en/shared-co...

uva.nlHow can news media wean themselves off the Big Tech drip?News media remain heavily reliant on Big Tech. UvA researcher Hannes Cools is working with De Correspondent and Follow the Money to build greater digital autonomy through European and openโ€‘source tool...
View full thread

When a judge at the ICC in The Hague can suddenly lose access to his own email because Microsoft blocked him, it's a preview of what could just as easily hit a newsroom's ability to fulfil its watchdog role. Together with De Correspondent, Follow the Money, Waag Futurelab & HvA.

๐ŸŽ™๏ธ ๐ˆ๐ง๐ญ๐ž๐ซ๐ฏ๐ข๐ž๐ฐ ๐จ๐ง ๐ง๐ž๐ฐ ๐ฉ๐ซ๐จ๐ฃ๐ž๐œ๐ญ '๐ƒ๐ข๐ ๐ข๐ญ๐š๐ฅ ๐€๐ฎ๐ญ๐จ๐ง๐จ๐ฆ๐ฒ ๐Ÿ๐จ๐ซ ๐ง๐ž๐ฐ๐ฌ ๐ฆ๐ž๐๐ข๐š' Media have largely lost control over their online distribution. Dependencies on Big Tech have only intensified with the rise of so media platforms, & now AI isn't making this 'distribution challenge' any easier www.uva.nl/en/shared-co...

uva.nlHow can news media wean themselves off the Big Tech drip?News media remain heavily reliant on Big Tech. UvA researcher Hannes Cools is working with De Correspondent and Follow the Money to build greater digital autonomy through European and openโ€‘source tool...

โ–ช๏ธ ๐“๐ก๐ž ๐œ๐ž๐ง๐ญ๐ซ๐š๐ฅ ๐ช๐ฎ๐ž๐ฌ๐ญ๐ข๐จ๐ง ๐จ๐Ÿ ๐จ๐ฎ๐ซ ๐ญ๐ฐ๐จ-๐ฒ๐ž๐š๐ซ ๐ฉ๐ซ๐จ๐ฃ๐ž๐œ๐ญ ๐ข๐ฌ: What does digital autonomy actually mean (in practice) for journalism, and what steps are to be taken to 'reach' digital autonomy? โ–ช๏ธ ๐–๐ก๐ฒ ๐ข๐ญ ๐ฆ๐š๐ญ๐ญ๐ž๐ซ๐ฌ: Newsrooms depend on a handful of US tech companies for software, servers & cloud...

View full thread

Hij wilde niet wachten op een gesprek met een psycholoog. De boodschap die hij uiteindelijk meegeeft aan zijn volgers (en dus vnl jongeren): "Je moet spreken. Had ik maar eerder met ChatGPT gepraat.โ€

Het feit dat jongeren, vaak op hรฉรฉl kwetsbare momenten, hun gedachten toevertrouwen aan een systeem, baart zorgen En het is ook wat ik eerder in essay voor @groene.nl beschreef: "Een AI-tool die de drempel naar mentale hulp verlaagt, verlaagt in werkelijkheid de drempel naar iets wat geen hulp is"