宛瑾 Wan Jin

Finally see a nice protocol of using LLM to annotate the dataset! Interesting points:
1. Use chain-of-thought in classification to make the reasoning visible - sth often missing in human coding
2. Asking LLM to quote the text for reasoning & not hallucinate can decrease hallucinations

Establishing the construct and predictive validity of brief measures of affective polarization Measuring affective polarization, defined as the liking for one's political ingroup and the dislike for political outgroups, poses methodological challenges in multiparty systems: evaluations of seve...

1/7
First paper of my dissertation now out in EJPR with @bnbakker.bsky.social & @gijsschumacher.bsky.social

Can we measure affective polarization in systems with 6, 10, or even more parties without overburdening respondents or compromising validity? We show how.

Paper: doi.org/10.1111/1475...

If you know me, you know I am obsessed with building dashboards. 😅

Well, I am excited to share that @ernesto-deleon.com, Max Paulus, & I were awarded a small grant to build an #rstats package to make them accessible to other researchers.

And we’re hiring a Research Assistant to help us do that!

There is a study comparing using GenAi and traditional search engine for fact-checking and it turns out search engine is better in accuracy

Very interesting and relevant topic, check it out👇