Pathfinder / funder / opportunities / Assert, don't describe

Assert, don't describe

AI safetyRemoteEA Forum

Brief

Research note on how linguistic features in fine-tuning data can shift LLM reasoning about animal welfare; the provided source recommends asserting a moral position rather than neutrally describing scenes, and avoiding hedging and overly concrete sensory detail.

Founders

No founder records yet.

Snapshot

Hub
Remote
Source
EA Forum
Posted
6/5/2026