Lightnews — Scholar-powered news

Indira Sen

@indiiigo.bsky.social

600 followers 430 following 38 posts

Junior Faculty at the University of Mannheim || Computational Social Science ∩ Natural Language Processing || Formerly at: RWTH, GESIS || she/her indiiigo.github.io/

Posts Media Videos Starter Packs

Pinned

Indira Sen @indiiigo.bsky.social · Jul 21

Do LLMs represent the people they're supposed simulate or provide personalized assistance to?

We review the current literature in our #ACL2025 Findings paper and investigating what researchers conclude about the demographic representativeness of LLMs:
osf.io/preprints/so...

1/

Screenshot of our paper "Missing the Margins: A Systematic Literature Review on the Demographic Representativeness of LLMs"

Details about what we annotated in our systematic review

2 8 23

Reposted by Indira Sen

Anne Lauscher @a-lauscher.bsky.social · 14d

🚨 Are you looking for a PhD in #NLProc dealing with #LLMs?
🎉 Good news: I am hiring! 🎉
The position is part of the “Contested Climate Futures" project. 🌱🌍 You will focus on developing next-generation AI methods🤖 to analyze climate-related concepts in content—including texts, images, and videos.

1 14 22

Reposted by Indira Sen

Juhi Kulshrestha @juhi153.bsky.social · 21d

We are hiring multiple PhD and postdocs for two newly funded projects at the intersection of mental health and political polarization at the CS Dept at Aalto, Finland. The PIs are Juhi Kulshrestha, Talayeh Aledavood, and Mikko Kivelä.

Full call text and link to apply: www.aalto.fi/en/open-posi...

5 8

Reposted by Indira Sen

Joachim Baumann @joachimbaumann.bsky.social · 26d

🚨 New paper alert 🚨 Using LLMs as data annotators, you can produce any scientific result you want. We call this **LLM Hacking**.

Paper: arxiv.org/pdf/2509.08825

$We present our new preprint titled "Large Language Model Hacking: Quantifying the Hidden Risks of Using LLMs for Text Annotation". We quantify LLM hacking risk through systematic replication of 37 diverse computational social science annotation tasks. For these tasks, we use a combined set of 2,361 realistic hypotheses that researchers might test using these annotations. Then, we collect 13 million LLM annotations across plausible LLM configurations. These annotations feed into 1.4 million regressions testing the hypotheses. For a hypothesis with no true effect (ground truth $p > 0.05$), different LLM configurations yield conflicting conclusions. Checkmarks indicate correct statistical conclusions matching ground truth; crosses indicate LLM hacking -- incorrect conclusions due to annotation errors. Across all experiments, LLM hacking occurs in 31-50\% of cases even with highly capable models. Since minor configuration changes can flip scientific conclusions, from correct to incorrect, LLM hacking can be exploited to present anything as statistically significant.$

5 94 260

Reposted by Indira Sen

Naomi Saphra @nsaphra.bsky.social · Aug 29

How can an imitative model like an LLM outperform the experts it is trained on? Our new COLM paper outlines three types of transcendence and shows that each one relies on a different aspect of data diversity. arxiv.org/abs/2508.17669

3 17 95

Reposted by Indira Sen

Zeerak Talat زیرک ظلعت (they/them) @zeerak.bsky.social · Aug 25

Come join and organise the workshop with us!

Workshop on Online Abuse and Harms @woahworkshop.bsky.social · Aug 25

Excited for WOAH’s 10th anniversary? 😍

We're launching an open call for new organisers!
Our goal: diversify the team and bring in fresh perspectives.

🗓️ Apply by September 12
🔗 forms.gle/aiFs35vwDXnt...

1 2

Indira Sen @indiiigo.bsky.social · Aug 20

maybe @a-lauscher.bsky.social's "Multi3hate: Multimodal, multilingual, and multicultural hate speech detection with vision-language models" aclanthology.org/2025.naacl-l...

Multi3Hate: Multimodal, Multilingual, and Multicultural Hate Speech Detection with Vision–Language Models

Minh Duc Bui, Katharina Von Der Wense, Anne Lauscher. Proceedings of the 2025 Conference of the Nations of the Americas Chapter of the Association for Computational Linguistics: Human Language Technol...

aclanthology.org

Reposted by Indira Sen

Gabriella Lapesa @gabriellalapesa.bsky.social · Aug 20

If you want to nominate yourself to be the organizer of the next Argument Mining workshop @argminingorg.bsky.social‬, fill in this form: docs.google.com/forms/d/e/1F... Deadline: Aug 22nd 13.00 CEST!

ArgMining 2026 Workshop Organising Committee Application

docs.google.com

1 1

Reposted by Indira Sen

Marvin Stecker @marvins.bsky.social · Aug 19

New publication, out in Political Analysis:

There is an increasing array of tools to measure facets of morality in political language. But while they ostensibly measure the same concept, do they actually?

I and @fhopp.bsky.social set out to see what happens.

Moral Foundation Measurements Fail to Converge on Multilingual Party Manifestos | Political Analysis | Cambridge Core

Moral Foundation Measurements Fail to Converge on Multilingual Party Manifestos

www.cambridge.org

3 14 34

Reposted by Indira Sen

Nedjma Ousidhoum @nedjmaou-nlp.bsky.social · Aug 15

The Call for #EMNLP2025 @emnlpmeeting.bsky.social student volunteers is out:
2025.emnlp.org/calls/volunt...
Please fill out the form by 20 Sep 2026 : forms.gle/qfTkVGyDitXi...
For questions, you can contact emnlp2025-student-volunteer-chairs [at] googlegroups [dot] com

Call for Volunteers

Official website for the 2025 Conference on Empirical Methods in Natural Language Processing

2025.emnlp.org

4 3

Reposted by Indira Sen

Lucie-Aimée Kaffee @frimelle.bsky.social · Aug 11

Work with @giadapistilli.com and @yjernite.bsky.social

📄 Full Paper: huggingface.co/datasets/AI-...
🔍 Explore INTIMA: huggingface.co/datasets/AI-...

huggingface.co

2 4

Reposted by Indira Sen

Data Science Hamburg Group @ds-hamburg.bsky.social · Jul 29

Wondering what makes each of us unique and how AI should handle human diversity? 🤔

We're organizing Identity-Aware AI workshop at #ECAI2025 Bologna on Oct 25.

Deadline: Aug 22
Website: identity-aware-ai.github.io

Identity-Aware AI Workshop announcement. Co-located with ECAI 2025 in Bologna on October 25, with submission deadline August 22. Topics include: Methods for effective, fair, and inclusive AI; Critiques of AI on the exclusion of identities; Methods for detecting and controlling bias; Perspectivist approaches to AI. Submission types: Long papers (8 pages), Short papers (4 pages), Extended abstracts, Mixed-media submissions (videos, blogs, codebase, artworks). For details, visit: identity-aware-ai.github.io

1 3 5

Reposted by Indira Sen

Lucie-Aimée Kaffee @frimelle.bsky.social · Aug 8

Wikipedia has long been one of my favourite places online. As AI becomes part of knowledge creation, there's a lot we can learn from its editor communities. I spoke with Daniel Wu about AI content on Wikipedia; some thoughts made it into this piece:
www.washingtonpost.com/technology/2...

Volunteers fight to keep ‘AI slop’ off Wikipedia

Hundreds of Wikipedia articles may contain AI-generated errors. Editors are working around the clock to stamp them out.

www.washingtonpost.com

3 8

Reposted by Indira Sen

Aleksandra Urman @aurman21.bsky.social · Aug 7

What do people in 🇨🇭🇩🇪🇯🇵🇺🇸 think about GenAI for news-related purposes?

We find the adoption of GenAI for news and trust in the journalistic deployment of GenAI are relatively low, and so is knowledge regarding GenAI.

Read more in this new paper led by Eliza Mitova! journalqd.org/article/view...

Exploring Public Attitudes Toward Generative AI for News Across Four Countries | Journal of Quantitative Description: Digital Media

journalqd.org

1 3 8

Reposted by Indira Sen

Petter Törnberg @pettertornberg.com · Aug 6

We built the simplest possible social media platform. No algorithms. No ads. Just LLM agents posting and following.

It still became a polarization machine.

Then we tried six interventions to fix social media.

The results were… not what we expected.

arxiv.org/abs/2508.03385

Can We Fix Social Media? Testing Prosocial Interventions using Generative Social Simulation

Social media platforms have been widely linked to societal harms, including rising polarization and the erosion of constructive debate. Can these problems be mitigated through prosocial interventions?...

arxiv.org

13 99 290

Reposted by Indira Sen

Dr Abeba Birhane @abeba.bsky.social · Aug 6

"Asked to generate intervention plans for struggling students, AI teacher assistants recommended more-punitive measures for hypothetical students with Black-coded names and more supportive approaches for students the platforms perceived as white" www.chalkbeat.org/2025/08/06/a...

Annie and Lakeesha struggle in school. AI teacher assistants treated them very differently.

A Common Sense Media study found that prominent teacher assistants that use AI generated recommendations that appeared to be rooted in racial stereotypes based on students’ names. About a third of tea...

www.chalkbeat.org

9 320 550

Indira Sen @indiiigo.bsky.social · Aug 1

Lots of great posters at the #wiknlp workshop at #ACL2025NLP

Indira Sen @indiiigo.bsky.social · Aug 1

Great keynote by Matthias Gallé on multilinguality in LLMs with takeaways on how we have to go broader *and* deeper to achieve multilingual efficacy by covering local knowledge.

Struck by the industrialization of LLM research with LLM tech reports now having massive # authors. #wikinlp #acl2025nlp

Reposted by Indira Sen

Workshop on Online Abuse and Harms @woahworkshop.bsky.social · Aug 1

Time for our second keynote 🚨

@fvancesco.bsky.social is going to guide us through practical aspects of safety that are often overlooked in academia.

Do we want to close the gap between academia and industry? Join us to find out!

#ACL2025NLP

1 5

Indira Sen @indiiigo.bsky.social · Aug 1

Excellent panel on dataset papers using Wikipedia data and the importance and challenges of multilingual research.

Check out the dataset paper’s here: meta.m.wikimedia.org/wiki/NLP_for...

Indira Sen @indiiigo.bsky.social · Aug 1

Incredible keynote by Monica Lam on creating LLM-powered research assistants.

One great example of NLP/wikipedia synergy is this tool that helps find inconsistencies in Wikipedia articles and fix them semi-automatically: wikifix.genie.stanford.edu

2 1 2

Indira Sen @indiiigo.bsky.social · Aug 1

Happening now!

On the interplay between Wikipedia and NLP

1 3

Indira Sen @indiiigo.bsky.social · Aug 1

Last day of #ACL2025NLP but there's still lots to do: attend the #WikiNLP workshop, where we explore how NLP and wikipedia can help each other!

We have amazing keynotes, discussions with Wikipedia editors, a panel + posters!

Details: meta.wikimedia.org/wiki/NLP_for...

Join us in room 2.31!

WikiNLP workshop program with keynotes, dataset panel, poster session, discussions with Wikipedia editors and more.

1 3 20

Indira Sen @indiiigo.bsky.social · Jul 29

Hire Agostina! She does lots of great work in CSS+NLP

Agostina Calabrese @agostinacal.bsky.social · Jul 29

At #ACL2025NLP and on the job market (NLP + AI Safety) 💼

It's great to see growing interest in safety/alignment, but we often miss the social context.

Come to our @woahworkshop.bsky.social Friday to dive deeper into safe safety research!

A quiet token from the biggest @aclmeeting.bsky.social ⬇️

1 2

Indira Sen @indiiigo.bsky.social · Jul 29

It’s poster board 1! The only CSS poster in this poster session!!

Indira Sen @indiiigo.bsky.social · Jul 29

👋 #ACL2025NLP 🇦🇹 @marlutz.bsky.social and I are presenting our poster on demographic representativeness of LLMs today!

🕦 10:30-12:00
📍 Hall X5 (board 1 or 14 according to different sources 🧐)

Here’s the paper on ACL anthology: aclanthology.org/2025.finding...

Drop by!