Lightnews — Scholar-powered news

Reposted by Thomas Davidson

Workshop on Online Abuse and Harms @woahworkshop.bsky.social · 9d

Should WOAH start a mentorship programme? 🤔

As the workshop grows, reviewer expectations are rising.
We don’t want contributors from adjacent communities penalised by *CL norms.

Senior PhDs and beyond could be mentors.

Share your thoughts:

👉 forms.gle/safif3rU2rs5...

4 4

Reposted by Thomas Davidson

Neal Caren @haphazardsoc.bsky.social · 12d

Carolina Sociology is hiring!

• Tenure-track in computational sociology or advanced quantitative methods
• Open rank in population health.

unc.peopleadmin.com/postings/307...
unc.peopleadmin.com/postings/307...

Assistant Professor

This Assistant Professor position will teach graduate and undergraduate courses, conduct research in the field of Sociology, participate in departmental service, and mentor graduate students.

unc.peopleadmin.com

1 38 30

Reposted by Thomas Davidson

Étienne Ollion @eollion.bsky.social · 15d

Come work with us @crestsociology.bsky.social

We are looking for a post-doc with a background in social sciences, interested in NLP and in media & journalism studies.

Details here: www.css.cnrs.fr/post-doctora...

Post-Doctoral Position – Media / Journalism – CSS @ IP-Paris

Site web de l'axe sciences sociales computationnelles du CREST-CNRS. Cours et tutoriels pour l'analyse des données numériques en sciences sociales.

www.css.cnrs.fr

32 28

Thomas Davidson @thomasdavidson.bsky.social · 16d

Big professional news!

Screenshot of an email that reads:

Call to Join Editorial Board - ASS

Dear Dr. Thomas Davidson,

I hope this message finds you well.

We are pleased to invite you to consider joining the Editorial Board of Archives of Social Science (ASS). As a specialist in the field, your expertise and insights would be invaluable in shaping the journal's scientific rigor and global reach.

ASS is an open-access platform dedicated to publishing high-quality research across all domains of social science, such as sociology, psychology, anthropology, political science, economics, education, communication studies, social work etc. We strive to foster collaboration and visibility among professionals worldwide.

1 3

Reposted by Thomas Davidson

Laura K. Nelson @lauraknelson.bsky.social · Sep 8

📊The Sociology Department at Stony Brook University invites applications for *two* tenure-track Assistant Professor positions to begin in Fall 2026. [They] seek scholars who specialize in issues of global inequality and justice, broadly construed.

apply.interfolio.com/172669

Apply - Interfolio {{$ctrl.$state.data.pageTitle}} - Apply - Interfolio

apply.interfolio.com

21 13

Reposted by Thomas Davidson

Thomas Davidson @thomasdavidson.bsky.social · 21d

Preparing my CSS class for next week and learned that Spotify put limits on its API that prevent most interesting queries developer.spotify.com/blog/2024-11...

This was my main go-to after the Twitter API shut down and now I have to find a new application. Does anyone have any other API reccs?

2 1 1

Thomas Davidson @thomasdavidson.bsky.social · 21d

Preparing my CSS class for next week and learned that Spotify put limits on its API that prevent most interesting queries developer.spotify.com/blog/2024-11...

This was my main go-to after the Twitter API shut down and now I have to find a new application. Does anyone have any other API reccs?

2 1 1

Thomas Davidson @thomasdavidson.bsky.social · 26d

Interested to read this more closely. One issue I worry about using LLMs for text analysis is the slippage between annotation and classification, particularly for zero-shot learning (is it annotation or classification?). Certainly appears to underscore the critical need for validation.

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.$

3

Reposted by Thomas Davidson

LSE Methodology @lsemethodology.bsky.social · 27d

We're hiring an Assistant Professor in Computational Social Science ❗

📚 jobs.lse.ac.uk/Vacancies/W/...

Apply before 26 October and join an internationally outstanding group of social science methodologists 🌎

we're hiring assistant professor in computational social science, applications close 26/10/2025

57 47

Reposted by Thomas Davidson

Friedolin Merhout @fmerhout.bsky.social · 27d

Come join me in wonderful Copenhagen! 🇩🇰

My department is looking to fill at least two positions - any specialization and any level! The University of Copenhagen aims to be the best place for the best ideas. What’s yours?

Apply by Nov 15.

Call for two or more open-rank academic positions in Sociology

jobportal.ku.dk

2 25 39

Thomas Davidson @thomasdavidson.bsky.social · 29d

Ha thanks, Rohan. It’s a short one but it scratches the surface of what we can do with LRMs. And I haven’t seen any evaluations like this in the CS literature. Let me know if you have any feedback

2

Thomas Davidson @thomasdavidson.bsky.social · 29d

Substantively, the results show how reasoning effort and traces could help tasks like content moderation

There are, of course, caveats: LRMs do not replicate human cognition, the models have limited capabilities, and reasoning traces are not always faithful

Preprint: arxiv.org/pdf/2508.20262

arxiv.org

1 1

Thomas Davidson @thomasdavidson.bsky.social · 29d

Analysis of the reasoning traces for Gemini 2.5 shows that the model identifies second-order factors when faced with these decisions, helping to address common false positives like flagging reclaimed slurs as hate speech (Warning: offensive language in example)

1 1

Thomas Davidson @thomasdavidson.bsky.social · 29d

On a content moderation task, humans take longer and LRMs use more tokens when offensiveness is identical or fixed.

This suggests that LRM behavior is consistent with dual process theories of cognition, as the models expend more reasoning effort when simple heuristics are insufficient

1 2

Thomas Davidson @thomasdavidson.bsky.social · 29d

The results are consistent across three frontier LRMs: o3, Gemini 2.5 Pro, and Grok 4

1 1

Thomas Davidson @thomasdavidson.bsky.social · 29d

New pre-print on large reasoning models 🤖🧠

To what extent does LRM behavior resemble human reasoning processes?

I find that LRM reasoning effort predicts human decision time on a pairwise comparison task, and both humans and LRMs require more time/effort on challenging tasks

1 1 6

Reposted by Thomas Davidson

Per Engzell @pengzell.bsky.social · Sep 1

WE ARE HIRING! 2 Lecturers in Quantitative Social Science. Want a friendly interdisciplinary department in one of the world's most vibrant cities? This just might be for you.

Apply by: 10 Oct

www.ucl.ac.uk/work-at-ucl/...

3 160 150

Thomas Davidson @thomasdavidson.bsky.social · Aug 5

This is pretty much what the earlier work on debiasing word embeddings was doing. And it turns out that it doesn't work...

Cue one of the best CS paper titles aclanthology.org/N19-1061/

4

Thomas Davidson @thomasdavidson.bsky.social · Aug 1

Contributors include: @sxz.bsky.social @ajalvero.bsky.social @eollion.bsky.social @lpargyle.bsky.social @davidbroska.bsky.social @austin-van-loon.bsky.social @bstewart.bsky.social @hwaight.bsky.social @solmg.bsky.social @tinalaw.bsky.social @lauraknelson.bsky.social @oms279.bsky.social

5

Thomas Davidson @thomasdavidson.bsky.social · Aug 1

Thanks for reading this thread. The entire special issue can be found here: journals.sagepub.com/toc/smra/54/3

If you want to talk more about AI and sociology, I'll be at ASA and will be giving a talk on some ongoing AI research in the Saturday morning session on Culture and CSS

2

Thomas Davidson @thomasdavidson.bsky.social · Aug 1

Of course, the methodological focus means that we don’t cover other important topics like impacts of AI on society and the politics of AI. There is a lot more work to be done.

See the recent Socius special issue for more perspectives on the sociology of AI journals.sagepub.com/topic/collec...

journals.sagepub.com

1 1

Thomas Davidson @thomasdavidson.bsky.social · Aug 1

If you want to learn more about these articles, check out our editors’ introduction, where we provide an overview and discuss some central themes.

One point we emphasize is how the model ecosystem has matured & open-weight models are viable for many problems journals.sagepub.com/doi/10.1177/...

1 1

Thomas Davidson @thomasdavidson.bsky.social · Aug 1

Stuhler, Ton, and Ollion show how LLMs enable more complex information extraction tasks that can be applied to text corpora, with an application to the study of obituaries journals.sagepub.com/doi/abs/10.1...

1 3

Thomas Davidson @thomasdavidson.bsky.social · Aug 1

A study by Than et al. revisits an earlier SMR paper, exploring how LLMs can be used for qualitative coding tasks, providing detailed guidance on how to approach the problem journals.sagepub.com/doi/full/10....

1 1 1

Thomas Davidson @thomasdavidson.bsky.social · Aug 1

Law and Roberto showcase how vision language models like GPT-4o can extract information from satellite images, applying these techniques to study segregation in the built environment journals.sagepub.com/doi/full/10....

1 1