Juan S. Gómez-Cruces

Can language models identify political stances as accurately as experts? We are testing language models to identify political stances using a transparent and academically rigorous approach. With only a bit more calibration, we will be close to the mean of scores provided by experts.

OpenAI’s Sora Makes Disinformation Extremely Easy and Extremely Real

OpenAI’s Sora Makes Disinformation Extremely Easy and Extremely Real www.nytimes.com/2025/10/03/t...

Yesterday I presented our paper "Automated Political Stance Identification” at the #11ICSSM (@isa-sociology.org)

This interdisciplinary work uses a purpose-specific language model + Natural Language Inference (NLI) to detect stances in political texts.

📊 Are populist attitudes evenly spread across the ideological spectrum?

➡️ Using CSES data from 43 countries, E Tamaki &
@drjungphd.bsky.social find a non-linear relationship: populism is strongest at the ideological extremes, forming a U-shape pattern www.cambridge.org/core/journal... #FirstView

#dataquality #KODAQS
🎉 The KODAQS Data Quality Toolbox Has Launched! 🎉
Designed to support researchers with assessing the quality of their data, the Toolbox provides practical coding examples of quality indicators, and resources that make achieving reliable, high-quality data easier than ever.

From survey data to digital behavioral and linked data, the KODAQS Toolbox empowers you to analyze your data with precision at every stage, ensuring the validity of research findings. Ready to improve the quality of your data?

Explore the KODAQS Toolbox!

🔗 kodaqs-toolbox.gesis...

🚨 New preprint alert:

We show that web-browsing GPT and LLaMA models can infer social media user demographics with reasonable accuracy—using only usernames.

This opens new possibilities for social media research in the post-API era but raises important privacy concerns.

This summer I am teaching an LLMs for social research course to non-social scientists (mostly computer science people). I appreciate any advice, references, memes, etc.

Yesterday I presented this work in progress at #epsa2025. Thank you for the helpful and on point feedback.

I am looking forward to #WIConf25, where I'll be presenting my recent research on the impact of populist democracy on internet freedom.

Here you can learn more about the project we presented yesterday:

pathtopower.global/path-to-powe...

#PathtoPower

Thrilled to have been part of the panel discussion at “The Path to Power 2024: Democracy Disrupted?”, where we dove into the role of digital technology in the 2024 elections worldwide.
Many thanks to the HPI and GIZ for supporting this project.
#PathToPower

We are #hiring a funded #PhD in open-source software or alternative economic and production structures within digital tech as part of the Digital Technology, Governance and Policy Research Group @HPI_DE at @unipotsdam Application deadline is January 16th

tinyurl.com/fd2w4vu8

Large language models surpass human experts in predicting neuroscience results - Nature Human Behaviour Large language models (LLMs) can synthesize vast amounts of information. Luo et al. show that LLMs—especially BrainGPT, an LLM the authors tuned on the neuroscience literature—outperform experts in pr...

"Large language models surpass human experts in predicting neuroscience results" www.nature.com/articles/s41...