Beyond Right and Wrong: Evaluating Second-order Social Reasoning in Large Language Models
Key takeaway
Previous AI alignment efforts have focused primarily on first-order social norms -- teaching models what is socially acceptable or unacceptable (e.g., `do not steal').
Why it's on PDOOM
PDOOM selected this story for its alignment & control signals: Alignment.
AlignmentFrom ArXiv cs.AI
Previous AI alignment efforts have focused primarily on first-order social norms -- teaching models what is socially acceptable or unacceptable (e.g., `do not steal'). However, social intelligence depends not only on norm recognition, but also on anticipating who will enforce it and how (e.g., public shame or even imprisonment). These second-order expectations, known as metanorms, govern how people respond when social rules are broken. We introduce
By Sunny Rai, Jinyi Kuang, Reyhan Jamalova, Annie Lou, Cristina Bicchieri, Niyati Malhotra, Victor Hugo