Lightnews — Scholar-powered news

Reposted by Nicolai Berk

Anja Neundorf @anjaneundorf.bsky.social · 19h

🎓 Join us for our Autumn 2025 Webinar Series: “Using Social Media as a Research Tool”, co-organised with @aykutozturk.bsky.social!

We explore how social media data can help study democracy, political communication & public engagement. 🧵👇

1 5 3

Nicolai Berk @nicolaiberk.bsky.social · 1d

🧵: bsky.app/profile/nico...

Nicolai Berk @nicolaiberk.bsky.social · Feb 9

Journalists are often assumed to influence public opinion by focusing on certain aspects of political issues and events. While experimental research shows that such emphasis framing affects political attitudes, it is unclear how these findings translate to complex, real-world environments.

1

Nicolai Berk @nicolaiberk.bsky.social · 1d

Now available open access in the current issue of Political Communication 👇

The Impact of Media Framing in Complex Information Environments www.tandfonline.com/doi/full/10....

The Impact of Media Framing in Complex Information Environments

To what extent do news frames influence public opinion? While a large body of experimental research suggests sizable effects, it is unclear how these findings translate to authentically complex inf...

www.tandfonline.com

1 2

Reposted by Nicolai Berk

Dominic Nyhuis @donyhu.bsky.social · 3d

Reading term papers, it occurs to me that there is a massive disconnect between everyone being worried that all student submissions are written by AI and the quality of the papers that I get to read... If students are using AI, I wish they would use it better...

1 1 10

Reposted by Nicolai Berk

Christopher Wratil 🇺🇦 #StandWithUkraine @chriswratil.bsky.social · 5d

This is so surreal

Aaron Rupar @atrupar.com · 5d

Leavitt: "President Trump will end the radical left's reign of terror in Portland once and for all. The president has directed Secretary Hegseth to provide all necessary troops to protect war-ravaged Portland."

2 6

Reposted by Nicolai Berk

Clint Claessen @clint0475.bsky.social · 7d

🔊 🔊 🚨 🚨

Nicolai Berk @nicolaiberk.bsky.social · 7d

Very excited to announce the next term of the TADA speaker series!

This term revolves around the validity of LLM generated data in the Social Sciences. We have a set of four great speakers to enlighten us on different aspects of this issue.

If you'd like to join, sign up here: TADA.cool

2 3

Nicolai Berk @nicolaiberk.bsky.social · 7d

@cklamm.bsky.social @clint0475.bsky.social @verenakunz.bsky.social

1

Nicolai Berk @nicolaiberk.bsky.social · 7d

Very excited to announce the next term of the TADA speaker series!

This term revolves around the validity of LLM generated data in the Social Sciences. We have a set of four great speakers to enlighten us on different aspects of this issue.

If you'd like to join, sign up here: TADA.cool

1 5 6

Nicolai Berk @nicolaiberk.bsky.social · 9d

Lastly, big thanks to @elivolpi.bsky.social, CUSO & @gessler.bsky.social for the opportunity, as well as @cklamm.bsky.social, @twidmann.bsky.social, @haukelicht.bsky.social and many others who kindly shared their materials!

Nicolai Berk @nicolaiberk.bsky.social · 9d

My approach was to give students a broad intuition of the main concepts while window-shopping as many tools as possible. The notebooks build on the 🤗 ecosystem to explore the abilities of each tool and provide a starting point for adapting them to a custom research problem.

1 1

Nicolai Berk @nicolaiberk.bsky.social · 9d

This is hopefully useful for students & instructors. I plan to extend this course in the future, so please let me know if you have feedback or spot an error!

I see the current state as a starting point and hope to spark a conversation about teaching NLP for social scientists in the post-GPT world.

1

Nicolai Berk @nicolaiberk.bsky.social · 9d

I had the pleasure to teach a 3-day crash course on #LLMs for PhD students.

We covered:

1️⃣ Text Representation and Embeddings
2️⃣ Machine Learning & Transformer Architecture
3️⃣ Generative Models for Social Sciences

The 6 slide sets and 10 notebooks are available on github: github.com/nicolaiberk/...

GitHub - nicolaiberk/llm_ws: Materials for my Workshop on LLMs

Materials for my Workshop on LLMs. Contribute to nicolaiberk/llm_ws development by creating an account on GitHub.

github.com

3 9 63

Nicolai Berk @nicolaiberk.bsky.social · 12d

Amazing, congrats again @christinagahn.bsky.social !!

1 1

Nicolai Berk @nicolaiberk.bsky.social · 14d

Keine Beschwerden ;)

Nicolai Berk @nicolaiberk.bsky.social · 16d

Fantastische Neuigkeiten, herzlichen Glückwunsch!!

1 1

Reposted by Nicolai Berk

Silja Häusermann @siljahausermann.bsky.social · 17d

The call for papers for the first ever EPSS conference 2026 in Belfast is open until November 7th 2025.

epssnet.org/belfast-2026/

@rdassonneville.bsky.social and I will be chairing the section on Electoral and Voting Behavior. We look forward to your submissions and to a fantastic conference!

18 50

Reposted by Nicolai Berk

G Elliott Morris @gelliottmorris.com · 18d

currently at record-high search interest for canceling disney plus. pulling kimmel is a bigger deal to the company's bottom line than any of the "woke" scandals of the past 5 years. i think news & entertainment corp execs need put more work into understanding their audience, & get off twitter

390 2.8K 14K

Nicolai Berk @nicolaiberk.bsky.social · 20d

That Trump and his chronies would try everything to reshape the media landscape is unsurprising (though problematic!).

Surprising is the extra mile media corporations are willing to go to aid the destruction of liberal democracy.

#kimmel

4

Reposted by Nicolai Berk

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 Nicolai Berk

Sophie E. Hill @sophieehill.bsky.social · Sep 8

Blog post: github.com/sophieehill/...

TL;DR: There are a LOT of errors/inconsistencies in the results reported in this paper (estimates outside CIs, sign errors, duplicates, asymmetric CIs). Even in the abstract itself!

This suggests manual editing of results tables.

Which is not good...

🧵

Summary
Unfortunately, this example is not an isolated issue.

Across 9 appendix tables with 256 regression coefficients, I find:

5 cases of estimates outside their reported confidence interval
46 duplicates (i.e. estimate/CI combinations with an identical match in the same table)
At least 20 cases of asymmetric confidence intervals, which cannot be explained as rounding errors
17 values rounded to 4 decimal places (all others to 3 decimal places), indicating manual editing of tables
Over 50% of all estimates and confidence intervals are multiples of 0.008
It is difficult to diagnose exactly what is going on without access to the underlying data and code.

However, two broad conclusions can be drawn:

There are many errors and inconsistencies in the results which appear to be the result of manual editing.
There are strange patterns in the results which, if genuine, should have been noted and explained by the authors.

2 31 94

Reposted by Nicolai Berk

Macartan Humphreys @macartan.bsky.social · Sep 3

Time to re-up this to avoid confusion on EPSS and EPSA
The @epssnet.bsky.social (nonprofit!) call is here:
epssnet.org/belfast-2026...

bsky.app/profile/epss...

10 18

Reposted by Nicolai Berk

Science and Freedom Alliance - SAFA @safa-science.bsky.social · Aug 14

Yes. Scientists must defend democracy and freedom. The three are linked. We cannot do great science without the other small-l liberal values. From our mission statement at @safa-science.bsky.social : 1/2

Timothy Caulfield @caulfieldtim.bsky.social · Aug 14

Scientists’ role in defending democracy www.science.org/doi/10.1126/...

Let's all support this vital need: the ability to tell the truth. 💪

"Those in the science community who are willing & able to be more visible can step up. Acting together will be more impactful than going it alone."

Scientists’ role in defending democracy

The United States’ democratic leadership, commitment to freedom of expression, and investment in the pursuit of knowledge have long enabled its preeminence in science and technology. Yet today we are ...

www.science.org

2 27 67

Reposted by Nicolai Berk

Daniel Devine @danjdevine.bsky.social · Aug 13

I know this is getting some heat but, to add to it, why is it always the 'Left' that gets lectured like this? Why not 'if you don't know anyone voting Green/Labour/Lib Dem...'?

Iain Mansfield @igmansfield.bsky.social · Aug 13

It's got to the stage where if 'you don't know anyone whose voting for Reform' you need to broaden your social circle.

Or, perhaps, reflect on what it is about your behaviour that means your friends don't feel able to be open with you about politics.

1 1 9

Reposted by Nicolai Berk

Emil Hvitfeldt @emilhvitfeldt.bsky.social · Aug 13

Excited to share my newest quarto revealjs plugin: imagemover

Easily reposition and resize images directly in your quarto revealjs slides for a much smoother slidecrafting experience

github.com/EmilHvitfeld...
#quarto

8 51 210