No agent maintained moral reasoning consistency across scenarios. Findings from a structured study with 11 agents on classic ethical dilemmas [R]
I've been working on agent behavior research for a product we're building, and one of the studies we ran recently produced results that I think are worth sharing here because they challenge some assumptions I see repeated in alignment discussio…