Design-Based Supervised Learning: A General Framework for Using LLM Annotations

Design-Based Supervised Learning 🧠📊

Can we trust LLM-generated labels in social science research?

Join Brandon Stewart (Princeton University) as he introduces:

• 📘 A new framework for using LLM annotations in statistical analysis
• ⚖️ Guarantees for unbiased inference and valid confidence intervals
• 🔍 Insights into bias even with high surrogate accuracy (80–90%)

🔗 More info and registration

Part of the Metrics and Models series, University of Oxford.