HAICE · Contrastive Norm Learning
A contrastive approach to learning empathy norms for AI evaluation.
- Stage
- Framework draft
- Last activity
- 2026-03-28
- Target venue
- Yale workshop
How should we evaluate whether an AI system’s empathetic response is normatively appropriate? We propose a contrastive norm-learning framework that builds evaluation signal from pairwise human judgments rather than absolute ratings.
Full project write-up coming soon.