𤷠AI Tip: Give your AI permission to say "I don't know"
Models often fill gaps with confident guesses. Anthropic's guide to reducing hallucinations says explicitly allowing Claude to admit uncertainty "can drastically reduce false information."
Add one line to your prompt:
If you're unsure or the report lacks the information, say "I don't have enough information to confidently assess this."
Working with a long document? Go further: ask for word-for-word quotes first, then an answer based only on those quotes. Or after a draft, make it find a supporting quote for every claim and delete any claim it can't back up.
ā An honest "I don't know" beats a confident fake fact.
#AI
Models often fill gaps with confident guesses. Anthropic's guide to reducing hallucinations says explicitly allowing Claude to admit uncertainty "can drastically reduce false information."
Add one line to your prompt:
If you're unsure or the report lacks the information, say "I don't have enough information to confidently assess this."
Working with a long document? Go further: ask for word-for-word quotes first, then an answer based only on those quotes. Or after a draft, make it find a supporting quote for every claim and delete any claim it can't back up.
ā An honest "I don't know" beats a confident fake fact.
#AI
