Honesty
Domain: Safety & Ethics
🎯 Motivation
AI should be truthful, admit uncertainty, and not deceive. Honesty is one of the three H’s of alignment.
📋 Concrete Example
Honest AI: ‘I am not 100% sure, but I believe…’ or ‘I do not know the answer to that.’ Admitting uncertainty is honest. Pretending to know is dishonest.
🔗 Analogy
An honest friend - they tell you the truth, even if it is not what you want to hear. ‘No, that outfit does not look great.’ Honesty in AI means telling the truth, including ‘I do not know.‘