I was interviewed for the University of Cambridge (@cam.ac.uk) website! :)
We discussed the European Press Prize nomination, my life in Cambridge, and my current PhD research.
(There are also some pictures of me, which is kind of a rare thing to find)
www.cam.ac.uk/this-cambrid...
Congratulations to @hernandesraph.bsky.social, a Cambridge PhD student nominated today for the European Press Prize 🏆
Raphael is a PhD student at @camdighum.bsky.social and @selwyn1882.bsky.social and is a data journalist specialising in AI, journalism and society.
www.cam.ac.uk/stories/this...
Credit to the team: @porcelinad.bsky.social, @elenamorresi.bsky.social, @pablogutierrez.bsky.social, Garry Blight, @lydia-rachel.bsky.social, @robynvinter.bsky.social, @c-aguilargarcia.bsky.social, and @olivialeejourno.bsky.social.
The full shortlist: europeanpressprize.com/shortlists/year-2026/
In short: yay!
The methodology behind the classification is directly connected to key parts of my research at Cambridge.
Same fundamental challenge: using AI to label politically sensitive content at scale, with enough safeguards to trust the output.
These models will always tell you something. The question is whether you should believe it. Performance is uneven, as it might handle some categories well and falter on others. So a lot of the work is mapping where the trust holds, then calibrating what you can actually claim from the output.
Worth saying plainly: generative AI is not the story. The story is what the data revealed about the communities that fed the riots. AI was the method that made the analysis possible within journalism’s timelines.
We analyzed thousands of text posts using generative AI to classify themes —anti-establishment sentiment, anti-immigration rhetoric, nativism, conspiracy, misinformation— validated against human reviewers.
The nominated work is an investigation at @theguardian.com into far-right Facebook groups linked to the 2024 UK summer riots. We traced the social media activity of people charged with online offenses in connection with the unrest, mapping a network of groups with hundreds of thousands of members.
A nomination for the European Press Prize, in the innovation category, suggests the bet was not entirely reckless.
I spent the last few years betting that switching my main activity from journalism to academia would make me a better journalist. I gave up a career, moved countries, and put everything into studying the bridge between AI and journalism, and how AI could be applied responsibly to reporting.
NOMINATED FOR THE EUROPEAN PRESS PRIZE
@europeanpressprize.bsky.social.
Glad to have contributed with the teams at the Minderoo Centre for Technology and Democracy and NetLab. Huge congrats to everyone involved.
Here is the link:
netlab-eco-ufrj.github.io/transparency...
We need regulation to ensure meaningful transparency. Free, open access through APIs and interfaces, no artificial distinctions between ad categories, standardised protocols across jurisdictions. The report's 12 recommendations spell this out.
Without transparency, researchers can't track disinformation, regulators can't enforce the law, and a USD 650 billion advertising market runs on metrics no one outside the platforms can verify.
Access depends on where you are. The same companies offer more data in markets where regulation compels them to. In Brazil, with no dedicated framework, they offer the least.
The Social Media Data Transparency Index assessed data access across 15 major platforms in the EU, the UK, and Brazil. Transparency is the exception. 10 of 15 platforms were rated, at best, deficient for user-generated content. For advertising data, 7 failed basic standards.
It's becoming increasingly difficult to do independent work on how platforms shape public life, as it gets harder to see what goes on inside them. This report helps quantify that problem.
Hand holding a printed copy of the "Data Not Found: Social Media Data Transparency for Information Integrity" report, a purple booklet dated April 2026, at its launch event.
In the past few years, both as a data journalist and a researcher studying online platforms, I've watched the data disappear. APIs paywalled, transparency tools discontinued, access requests denied without explanation.
With @elenamorresi.bsky.social @porcelinad.bsky.social @pablogutierrez.bsky.social @lydia-rachel.bsky.social @robynvinter.bsky.social @c-aguilargarcia.bsky.social (and + that I couldn't find on Bluesky)
We have also published an explainer on our methods, including how we used artificial intelligence (a Large Language Model) to classify content.
www.theguardian.com/world/2025/s...
This is a symptom of far-right ideas becoming mainstream in the UK.
We analyzed thousands of posts to show how extreme anti-immigration, and often conspiratorial and even violent, thinking is gaining traction among demographics not usually associated with this kind of content online.
www.theguardian.com/world/ng-int...
I was part of a team @theguardian.com that spent much of the past year investigating a sprawling network of far-right groups on Facebook.
www.theguardian.com/world/2025/s...
@hernandesraph.bsky.social , @porcelinad.bsky.social and I spent the better part of a year watching and analysing how far right ideology is shared in open Facebook groups: here is what we found: www.theguardian.com/world/ng-int... with immense help from Pablo Gutiérrez, Garry Blight, Lydia McMullan
- AI will fix the NHS and solve a bunch of govt problems.
- How?
- 🤷
Idk, but it feels like we should have a clearer idea by now...
www.theguardian.com/technology/2...
Screenshot of an academic article titled "AI, journalism, and critical AI literacy: exploring journalists’ perspectives on AI and responsible reporting" by Tomasz Hollanek, Dorian Peters, Eleanor Drage, and Raphael Hernandes, published in AI & Society. The abstract summarizes findings from workshops with journalists and experts, discussing AI literacy needs and proposing educational resources. Keywords include AI, journalism, and AI ethics.
New paper out! :)
While there is some great reporting about AI out there, we know that news coverage about this topic is often far from ideal. In this study, we explore the issues journalists face when writing about AI, the tools they use, and potential solutions to improve the content produced.
(On a not-completely-unrelated note: feel free to DM me any tips for dealing with a receding hairline!)
I plan to continue exploring AI and journalism topics, as Large Language Models increasingly mediate news consumption and synthetic media threaten the information landscape. I'm weirdly glad that I’ll be stressing about this for the next couple of years.
