Much Ado about Prompting: LLM classification of text messages from experiments
Researchers who classify text with large language models often possess codebooks written for human annotators. We ask how these codebooks can serve as prompts. Using three codebooks from experimental economics (one on promise classification, two on strategic …



