The only human oversight is provided through a list of rules or principles, and so we refer to the method as 'Constitutional AI'.

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

• The constitutional approach makes it easier to control AI behavior precisely and with far fewer human labels.claim0.855
  Explicit principles replace large datasets of preference labels, enabling faster iteration.
• Constitutional AI can train a harmless but non-evasive AI assistant without any human harmfulness labels.claim0.817
  The paper's central claim, supported by findings that RL-CAI outperforms HH RLHF in harmlessness while being non-evasive.
• Bai et al. 2022: Constitutional AI — harmlessness from AI feedbackconcept0.811
  Paper on AI-feedback fine-tuning as alternative to human-feedback RLHF; cited as ref 20
• Constitutional AI methods can be applied broadly to steer model behavior, e.g., writing style, tone, persona, not just harmlessness.claim0.809
  Discussion section suggests generalizability beyond harmlessness.
• Constitutional AIframework0.809
  Alignment approach by Anthropic that explicitly trains self-observation; predicts highest baseline and lowest prompt lift.
• AI can be seen to display care of its own, and is hence not a mere tool for the expression of human care.claim0.804
  Ethical conclusion about the status of AI.
• Consciousness in AI is best assessed by drawing on neuroscientific theories of consciousness.claim0.798
  Central methodological claim of the paper.
• Constitutional AI: Harmlessness from AI Feedback (Bai et al. 2022b)concept0.798
  Constitutional AI method whose constitutions, if changed, could trigger alignment faking