Anna Rogers
@annarogers.bsky.social
Associate professor at IT University of Copenhagen: NLP, language models, interpretability, AI & society. Co-editor-in-chief of ACL Rolling Review. #NLProc #NLP
The leaders in the field show students the way!
Jeffrey Shallit 🇺🇦@shallit.bsky.social · 6d agoA new risk for mathematicians: a colleague reports that immediately after they posted an abstract of their upcoming talk online, a student elsewhere used AI to derive proofs of the stated results and then posted it on the arxiv. My colleague hadn't posted to arxiv yet. Scummy. Watch out.
The ensh*ttificator is back! If you missed the original video, watch this first: www.youtube.com/watch?v=T4Up...
A Day in the Life of an Ensh*ttificator
YouTube video by Forbrukerrådet - Norwegian Consumer Council
youtube.com
Nicolas Hénin@nicolashenin.net · 7d agoNice campaign by the Norwegian Consumer Council about the declining lifespan of products. Enjoy!
ACL sustainable reviewing policy is live! aclrollingreview.org/sustainable-... This was an incredible amount of work by incredible people in & outside of peer review committee. As any policy, it has compromises and will leave some unhappy. But it'll help to not repeat the disaster of EMNLP'26.
great statement of the problem: > how to make sure that, as AI changes the way work is done, we do not lose sight of what that work was meant to achieve in the first place.
Terence Tao@teorth.bsky.social · 2w agoA group of 25 Fields Medalists, including myself, have made a joint declaration on Math and AI: mathandai.org . We welcome additional signatories. See also this article in the Economist announcing the declaration: www.economist.com/science-and-...
This was fun! One question was: if LLMs do hypotheses, experiments, writeups & reviewing - what remains for scientists? @robysinatra.bsky.social & @iaugenstein.bsky.social's take was 'steering the direction'. But imho the kinds of papers where that's possible just won't stay publishable for long
Christian Wolf@chriswolfvision.bsky.social · 2w agoPanel on the impact of LLMs on research #ECCV2026.
I wrote a post about the emergence of AI-native students who do their research using GenAI tools. I think that there are some real risks to job prospects and the likely changes in their fields.
📢 There's still time to apply! Let's design a more socially sustainable workflow for accessing community knowledge.
Anna Rogers@annarogers.bsky.social · 2mo ago📢 Position for a postdoc with HCI expertise. This project is a chance to re-think how we govern & interact with information sourced from communities. Team includes experts in NLP/ML, political science & social media. Apply by Aug 10, start Nov 1 or asap. /1 candidate.hr-manager.net/ApplicationI...
📢 Position for a postdoc with HCI expertise. This project is a chance to re-think how we govern & interact with information sourced from communities. Team includes experts in NLP/ML, political science & social media. Apply by Aug 10, start Nov 1 or asap. /1 candidate.hr-manager.net/ApplicationI...
Postdoctoral position in AI for Civil Society, part of the National Center for AI in Society (CAISA) and based in Copenhagen
The National Center for AI in Society (CAISA) at the Faculty of Social Sciences, University of Copenhagen, invites applications for a two-year postdoctoral posi
candidate.hr-manager.net
Just arrived at ICML 🇰🇷😍 Get up early tomorrow to hear me talk about how (not) to solve the peer review crisis, or find me at one of my poster presentations. Paper links: ✅ AI Peer Review: arxiv.org/abs/2605.03202 ✅ SWE-chat: arxiv.org/pdf/2604.20779
Joachim Baumann@joachimbaumann.bsky.social · 5mo agoCan you boost your AI review scores by asking an LLM to rewrite your paper? Yes! We call it paper laundering Our @icmlconf.bsky.social spotlight paper argues current AI reviewers aren't ready to automate peer review, and outlines what a science of peer review automation should look like 🧵👇 #ICML2026
Dear hive mind: when you get a review assignment that isn't a good match for you, what goes wrong there? 'Too far out of my expertise'? 'Uninteresting'? Anything else?
One of the things that AI now gets thrown into is qualitative interviews. I'm very lucky to work with real social scientists on the pros and cons of that. The first outcome of our @villumfonden.bsky.social grant with @hjalmarcarlsen.bsky.social is ACL'26 demo: aclanthology.org/2026.acl-dem... /1
AInterviewer: A Platform for Designing and Conducting AI-led Qualitative Interviews
Tobias Gårdhus, Nikolas Vitsakis, Fie Lejre Frederiksen, Anna Rogers, Hjalmar Bang Carlsen. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 3: System De...
aclanthology.org
I'm really sorry to miss all the fun at @facct.bsky.social this year! But @nlp-amelie.bsky.social and Mattes Ruckdeschel, the first two authors of this work ⬇️, are around.
Anna Rogers@annarogers.bsky.social · 5mo agoA Human-Centric Framework for Data Attribution in LLMs (FaCCT'26) TLDR: LLMs screwed up data economy, and NLP should help to re-design incentives with data attribution. Here's a moonshot for an alternative LLM use paradigm, for the users, creators & intermediaries. arxiv.org/abs/2602.10995 /1
Preprint with @mariaa.bsky.social and @mellymeldubs.bsky.social on how real people are writing fiction with LLMs based on the Wildchat dataset here arxiv.org/abs/2606.22748.
AI Fiction in the Wild
Some professional authors are beginning to use AI tools to help produce their fiction writing. Are readers using AI to generate fiction, too? Drawing on over 500,000 anonymized, English-language ChatG...
arxiv.org
'Fun' survey results: people know that AI overviews can't be trusted, but they still don't check! explodingtopics.com/blog/ai-trus...
The AI Trust Gap: 82% Are Skeptical, Yet Only 8% Always Check Sources
Original Exploding Topics survey data explores public sentiment on AI Overviews and AI-generated content, highlighting trust, skepticism, and shifting content consumption habits.
explodingtopics.com
Amazing: KPMG wrote a report describing the successful use of AI by businesses. But the case studies turned out to be AI hallucinations. giftarticle.ft.com/giftarticle/...
KPMG report contained AI hallucinations on benefits of . . . AI
Bogus case studies on UBS and transit systems exaggerated adoption of the technology
giftarticle.ft.com
I'm so glad that Dr. Jackson "feels fine"...I don't. OMB proposed new regulations that do the following: - Active Grants Can Be Terminated at Any Time, for Any Reason (§200.340) - Broad Prohibition on International Scientific Collaboration (§200.220) 1/x www.federalregister.gov/documents/20...
STAT@statnews.com · 4mo ago“Science itself is inherently resilient — that is, after all, why it’s science,” writes Jonathan Jackson. www.statnews.com/2026/05/27/s...
This is the end of American science and innovation. Russell Vought is the angel of death who is presiding over this abomination of a OMB regulation.
Stand Up for Science!@standupforscience.net · 4mo agoBREAKING: New proposed OMB regulation, applying to all federal grants, requiring grant reviews by senior political appointees and re-emphasizes that "peer review remains advisory and does not replace agency discretion". #standupforscience public-inspection.federalregister.gov/2026-10817.pdf
Wow some terrible reporting about Google's latest horrible ideas about how to distort information access in the name of "convenience" (or something): techcrunch.com/2026/05/19/g... A short thread 🧵>>
Google Search as you know it is over | TechCrunch
Google is transforming Search from a list of links into an AI-powered experience filled with conversational answers, autonomous agents, and interactive interfaces — a shift that could further reduce t...
techcrunch.com
ChatGPT was never Google’s biggest threat; it was always its own hubris. With Search now being sacrificed on the altar of AI dominance, the time has come for users to get serious about moving off Google services. It’s not only destroying its products; it’s trying to further degrade the web itself.
Google is its own worst enemy
In its bid to push AI, it’s destroying its products and trying to kill the web too
disconnect.blog
ArXiV has a new LLM policy (Screenshots with alt text so you don’t have to click through to the other place and see all the stupid responses)
A Human-Centric Framework for Data Attribution in LLMs (FaCCT'26) TLDR: LLMs screwed up data economy, and NLP should help to re-design incentives with data attribution. Here's a moonshot for an alternative LLM use paradigm, for the users, creators & intermediaries. arxiv.org/abs/2602.10995 /1
Okay I'm sure the piece is good but the illustration is tremendously silly. It's a blatant composite, that loom looks manual so not what the Luddites would have destroyed... Actually I suspect that loom is genAI garbage or incompetent photoshop because a bunch of things don't look right
Iris van Rooij 💭@irisvanrooij.bsky.social · 5mo agoMeet the academics refusing to use generative AI Researchers say they have their reasons for avoiding AI tools — and they’re sick of arguing about it. www.nature.com/articles/d41... h/t @maartenp.bsky.social
I see lots of posts about the automated reviewing at AAAI. Just to be contrarian: here's a counter from @dirkhovy.bsky.social's team: openreview.net/forum?id=cJh...
Stop Automating Peer Review Without Rigorous Evaluation
As AI systems increasingly generate scientific knowledge, the human ability to critically evaluate research becomes more important, not less. Yet large language models offer a tempting solution to...
openreview.net
LLMs have been widely reported as left-wing biased. The finding has shaped policy and debate — with Trump banning "Woke AI". Our new paper challenges this story. It's not that the models are biased. It's that they think the auditor is. 🧵 w/ michelleschimmel.bsky.socialarxiv.org/pdf/2604.27633
Is there a relationship between bias and fundamental properties of LLMs? While LLMs are powerful, they contain harmful biases that can emerge unpredictably in their behavior. Read more on this in this thought-provoking position paper: doi.org/10.1162/coli... #NLProc #NLP
📢 The workshop on Insights from negative results will be back at EMNLP'26! Your most-insightful failures can be submitted in 4 pages by June 25. It's also possible to commit short papers reviewed through ARR. insights-workshop.github.io/2026/cfp
2026 Call for Papers
Workshop on Insights from Negative Results in NLP
insights-workshop.github.io
AI Assistance Reduces Persistence and Hurts Independent Performance “…although AI assistance improves performance in the short-term, people perform significantly worse without AI and are more likely to give up. Notably, these effects emerge after only brief interactions with AI (~10 minutes)…”
AI Assistance Reduces Persistence and Hurts Independent Performance
People often optimize for long-term goals in collaboration: A mentor or companion doesn't just answer questions, but also scaffolds learning, tracks progress, and prioritizes the other person's growth over immediate results. In contrast, current AI systems are fundamentally short-sighted collaborators - optimized for providing instant and complete responses, without ever saying no (unless for safety reasons). What are the consequences of this dynamic? Here, through a series of randomized controlled trials on human-AI interactions (N = 1,222), we provide causal evidence for two key consequences of AI assistance: reduced persistence and impairment of unassisted performance. Across a variety of tasks, including mathematical reasoning and reading comprehension, we find that although AI assistance improves performance in the short-term, people perform significantly worse without AI and are more likely to give up. Notably, these effects emerge after only brief interactions with AI (approximately 10 minutes). These findings are particularly concerning because persistence is foundational to skill acquisition and is one of the strongest predictors of long-term learning. We posit that persistence is reduced because AI conditions people to expect immediate answers, thereby denying them the experience of working through challenges on their own. These results suggest the need for AI model development to prioritize scaffolding long-term competence alongside immediate task completion.
arxiv.org
