Jill Walker Rettberg
@jilltxt.bsky.social
Co-Director at Center for Digital Narrative, Univ of Bergen. Lead research project AI STORIES (ERC Advanced Grant). Australian and Norwegian. Blogging at jilltxt.net since 2000.
Widespread uptake of doorbell cameras paved the way for this. People really struggle to grasp that they are laying down the foundations of privately owned mass surveillance with every “personal” decision to buy this crap.
Hypervisible @hypervisible.blacksky.app · 15h ago“Although the buyers of this kind of technology, whether it be smart glasses or watches, may consent to their every move being monitored, the innocent bystanders that they record don’t necessarily give their blessing.”
An excellent article by @jilltxt.bsky.social that actually helped me understanding how #LLM work and (mis-)function. link.springer.com/article/10.1...
LLMs are metaphor machines: orbital argumentation, misaligned citations, and scientific fabrication in AI-generated writing - AI & SOCIETY
Large language models are metaphor machines: they generate text based on similarities and proximity in their training data. I argue that this metaphor-based text generation leads to what John Gallaghe...
link.springer.com
I have to say seeing "Associate Professor" in front of the name of someone who was once my PhD student is an extremely satisfying feeling. Also, NTNU is so lucky to have this wonderful researcher and teacher working with them ❤️ www.ntnu.edu/employees/ra...
Ragnhild Solberg - NTNU
ntnu.edu
Literally *years* of research on AI in education - *reams* of it by now - just does not support such sweeping proposals for mass AI access for students or educator "AI trailblazers", and it's maddening to see such uninformed, industry-friendly, research-free ideas from key sector bodies
Margot Finn@eicathomefinn.bsky.social · 22h ago'Every undergraduate would also be entitled to access artificial intelligence (AI) tools, the roadmap states, while every course should have a lecturer who is an “AI trailblazer” so that “graduates have the skills to be AI pioneers in the workplace”.' AI trailblazers. UUK at its very finest.
Glasses with cameras - like Meta's pervert glasses - are banned in Oslo schools. I hope universities follow suit, but this is the government's responsibilty. Ban technologies that make it easy to record or track other people without their knowledge or consent. www.utdanningsnytt.no/julie-remen-...
Forbyr KI-briller i Osloskolen
Oslo innfører forbud mot smartbriller for både «elever og voksne» i skoletiden.
utdanningsnytt.no
This sounds extremely cool and I can't wait to read it. Good thread breaking down the paper's insights!
Jill Walker Rettberg@jilltxt.bsky.social · 2d agoMy new article about LLMs is finally out in AI & Society! It argues that LLMs produce text metaphorically, through juxtaposing tokens that are related in some dimensions yet unrelated in others. This is why we see misaligned citations in scientific articles. link.springer.com/article/10.1...
Fascinating stuff, and helpful if your job involves reading hundreds of AI-‘assisted’ articles each year by students (& colleagues)
Jill Walker Rettberg@jilltxt.bsky.social · 2d agoMy new article about LLMs is finally out in AI & Society! It argues that LLMs produce text metaphorically, through juxtaposing tokens that are related in some dimensions yet unrelated in others. This is why we see misaligned citations in scientific articles. link.springer.com/article/10.1...
Denne var veldig god! Anbefales for alle som er interessert i skriving, i studenter, i kommunikasjon, i utdanning...
Jill Walker Rettberg@jilltxt.bsky.social · 2d agoWriting professor @johnrgallagher.bsky.social explains how LLMs generate orbital reasoning that skims yet avoids the key issues, and how AI-generated lists avoid causal connections and other analysis, making texts harder to read. The talk made me inspired to write! www.youtube.com/watch?v=cHRs...
My new article about LLMs is finally out in AI & Society! It argues that LLMs produce text metaphorically, through juxtaposing tokens that are related in some dimensions yet unrelated in others. This is why we see misaligned citations in scientific articles. link.springer.com/article/10.1...
LLMs are metaphor machines: orbital argumentation, misaligned citations, and scientific fabrication in AI-generated writing - AI & SOCIETY
Large language models are metaphor machines: they generate text based on similarities and proximity in their training data. I argue that this metaphor-based text generation leads to what John Gallaghe...
link.springer.com
The video of Marwa Abdulhai and Isadora White's talk about this paper is now online here - they present a pretty complex computer science paper so clearly, I highly recommend watching it if you're interested in using LLMs to edit or draft writing might do: www.youtube.com/watch?v=OnDQ...
How LLMs Distort Our Written Language #AIStories
YouTube video by Center for Digital Narrative, UiB
youtube.com
Aparna Nair@disabilitystor1.bsky.social · 2w agoLinkedin post here: lnkd.in/p/gjkcygv7
Videos of the talks are now up! bsky.app/profile/jill...
Aparna Nair@disabilitystor1.bsky.social · 2w agoI think what I find really depressing about AI in higher education is how so many people didn't even wait for the evidence/research before we decided on telling students "how to use AI responsibly" which often includes things like checking grammar, and obviously, we shouldn't have been
Two new talks on how LLMs affect writing are up on the @cdn-bergen.bsky.social YouTube channel, both from the AI STORIES event last week - you can access them from the event webpage here www4.uib.no/en/research/...
How do LLMs affect writing? Talks by visitors to the AI STORIES project | UiB
The AI STORIES project is hosting a day of research talks focused on analysing AI-generated writing, with approaches ranging from the qualitative analysis of a professor of writing to large-scale comp...
www4.uib.no
www.pressfire.no/artikkel/jo-... Så derfor måtte jeg svare. Det er i grunn trist hvordan den norske mediebransjen har malt seg selv inn i et hjørne når det kommer til sporing.
Jo, å kutte sporing er noe å være stolt av - PressFire.no
pressfire.no
Erik Fossum@erik.pressfire.no · 5d agoJo, jeg synes det å respektere leserne våre er noe å være stolt av. www.kode24.no/artikkel/ing...
Thanks for the interest in the topic of our workshop! We will record the talks and put them here afterwards: www4.uib.no/en/research/... #AIstories
How do LLMs affect writing? Talks by visitors to the AI STORIES project | UiB
The AI STORIES project is hosting a day of research talks focused on analysing AI-generated writing, with approaches ranging from the qualitative analysis of a professor of writing to large-scale comp...
www4.uib.no
My latest for @theverge.com is all about everyone's new favorite pastime: destroying Flock cameras. Tech backlash has entered a critical moment where folks not only widely oppose AI & tech industry excess, but openly support direct action against its most tangible and visible targets. 🎁 Gift Link:
Americans are cheering for vigilantes who take down Flock cameras
The surveillance cameras have become another flashpoint of growing bipartisan anger against tech industry excess.
theverge.com
Det er helt nødvendig at vi forskere lærer å gjenkjenne typiske KI-feil både som lesere og som skribenter, skriver KI-forsker @jilltxt.bsky.social: www.khrono.no/fem-regler-f...
Fem regler for KI-resistent forskning
Det er helt nødvendig at vi forskere lærer å gjenkjenne typiske KI-feil både som lesere og som skribenter, skriver KI-forsker Jill Walker Rettberg.
khrono.no
"Across predictive policing, facial recognition, automated welfare, deepfake sexual violence, and embodied robotics, AI does not malfunction; it performs exactly as designed, reproducing historical domination." journals.sagepub.com/doi/10.1177/...
Robots Behaving Badly: Algorithmic Colonialism and the Consequences of AI - Bronwyn Carlson, Tamika Worrell, 2026
This article examines the political, social, and relational consequences of artificial intelligence through an anti-colonial and Indigenous lens. Challenging cl...
journals.sagepub.com
Skal du bevare evnen til å tenke selv må du bli KI-resistent! Her er fem regler for å unngå kognitiv overgivelse! www.khrono.no/fem-regler-f...
Fem regler for KI-resistent forskning
Det er helt nødvendig at vi forskere lærer å gjenkjenne typiske KI-feil både som lesere og som skribenter, skriver KI-forsker Jill Walker Rettberg.
khrono.no
– Derfor er det helt nødvendig at vi forskere lærer å gjenkjenne typiske KI-feil både som lesere og som skribenter. Denne må du lese.
Jill Walker Rettberg@jilltxt.bsky.social · 2w agoSkal du bevare evnen til å tenke selv må du bli KI-resistent! Her er fem regler for å unngå kognitiv overgivelse! www.khrono.no/fem-regler-f...
Skal du bevare evnen til å tenke selv må du bli KI-resistent! Her er fem regler for å unngå kognitiv overgivelse! www.khrono.no/fem-regler-f...
Fem regler for KI-resistent forskning
Det er helt nødvendig at vi forskere lærer å gjenkjenne typiske KI-feil både som lesere og som skribenter, skriver KI-forsker Jill Walker Rettberg.
khrono.no
What a dystopic idea. As a parent of teens I disapprove.
Hypervisible @hypervisible.blacksky.app · 2w ago“Rather than FaceTiming a teen while they're taking a trip, parents and guardians will be able to view a live video feed from the selfie camera on the driver's phone and get a broader look inside the vehicle. Uber says the aim is to help give parents some piece of mind.”
Love these drawings of @notsaved.bsky.social from this report from the Eponymous Lecture Series - about algorithmic folklore and more. pad.profolia.org/s/Without_Us...
"This week @jilltxt.bsky.social, Co-director of the Center for Digital Narrative at the University of Bergen, declared 'shimmer' the newest AI word to make your eye twitch," writes #IBM: www.ibm.com/think/news/a...
AI won’t stop saying “shimmer.” Here’s a new theory why | IBM
A digital narrative researcher traced AI's obsession with "shimmer" to the same latent-space quirk behind Google's sparkle icon. Here's what’s going on.
ibm.com
My colleague Jessica Witte just shared a preprint where she uses sentiment analysis on AI-generated and human-authored stories and finds a pretty clear difference: AI stories have these long, slow rises in sentiment while human stories wait until the very end to conclude jilltxt.net/ai-generated...
AI-generated stories have longer endings than human stories
My colleague Jessica Witte has just shared a preprint where she compared the emotional arcs of the stories we generated using gpt-4o-mini to those of human-authored (pre-2022) stories from the subr…
jilltxt.net
Humanity #2: our uni travel reimbursement used to send rejections with terse anonymous msgs saying "you did it wrong! Fix it!" Now I get friendly signed emails with longer explanations and "have a nice day!" I feel much less stressed about travel reimbursements, though I still always make mistakes.
After the journal's draconian and automated "this is the absolute only time you can correct proofs for this article" the kind proof reader emailed me a 2nd corrected version to check. Maybe kind of backchannel, didn't show up in the journal system. It felt like a glimpse of humanity and I love it.
"Across studies we find extreme opinions are overrepresented on social media by at least 6% and up to 22% compared to their true prevalence in our samples." Paper: osf.io/preprints/ps... This is largely why i ignore radicalized loudmouths on any platform. Not only are they wrong, they're also fake
Jay Van Bavel, PhD@jayvanbavel.bsky.social · 2w agoSocial media discourse is driven by false #polarization We find that: 1) ideologues post opinions more than moderates 2) extreme attitudes are posted more than moderate one 3) This replicates in 40 countries 4) Opinions that users post are more hostile than the beliefs they keep to themselves
