Andreas Bjerre-Nielsen

@andbjn.bsky.social

associate prof. of econ and social data science | topics: education & algorithms | methods: stats, machine learning & econ. theory | dad

A

While human behavior and the data describing it evolve over time, fairness is often evaluated at a single snapshot. Yet, as we show in our newly published paper, fairness is dynamic. We studied how fairness evolves in dropout prediction across enrollment and found that it shifts over time.

Bild
A

Can AI interviewers match humans in qualitative interviews? We found: 🤖 AI asked more questions, yielding longer interviews 🧠 Responses to specific questions were longer in human-led interviews 📏 But response quality—specificity & relevance—was similar See more below 👇

OSF

osf.io

Hjalmar Bang Carlsen@hjalmarcarlsen.bsky.social · last yr.

🚨 New preprint! "Comparing AI-led to human-led chat-based interviews" Can LLMs collect interview data similar to human RAs? Turns out… yes (mostly). Note: these were text interviews—not directly comparable to in-depth, face-to-face qualitative interviews. Led by Semra Guven & @gaardhus.dk 🧵👇