Catarina Leão

@catarinaleao.bsky.social

Fellow in Advanced Quantitative Methods, Department of Methodology (LSE) | DPhil in Politics (Oxford) Political Economy of Autocratic Legacies + Democratic Erosion catarinaleao.com

V-Dem v.16 is now released, w. updated scores for 2025. Most countries remain stable from 2024, but some clear improvements (on diff’t democracy indices), incl. S. Korea, Sri Lanka & Mauritius. Largest declines came, by far, in the US. This figure shows 10-yr changes on the Liberal Dem. index.

Bild

Demographic cues (eg, names, dialect) are widely used to study how LLM behavior may change depending on user demographics. Such cues are often assumed interchangeable. 🚨 We show they are not: different cues yield different model behavior for the same group and different conclusions on LLM bias. 🧵👇

Bild

Portugal is having a presidential election tomorrow (Sunday). It is a fairly unusual one, which is representative of how much the political landscape has changed in the country. Here is some quick context about it, in case that is of interest: 1/9

Calling it “a regime-change war” fully obscures what’s actually happening right now. I doubt Trump is concerned with democratization as much as he is with waging a war to distract from problems at home, with oil extraction as a perk.

The Atlantic@theatlantic.com · 7mo ago

“Whether the outcome is ultimately good for Venezuelans, as I hope, or bad, Trump has betrayed Americans,” Conor Friedersdorf argues. “He could have tried to persuade Congress or the public to give him permission to use force. He didn’t bother.”

Very important work imo. We know surprisingly little about what “good” description actually looks like.

Daniel de Kadt@dandekadt.bsky.social · last yr.

🚨 “Good Description” with @annagbusse.bsky.social 🚨 What sets 'good' description apart from 'mere' description? We develop a framework for evaluating descriptive research, whether we are doing it as scholars or assessing it as readers. Two main contributions... 🔗📄 tinyurl.com/gooddesc