Keep hypotheses out of the evidence lane
Distinguish evidence from hypotheses so a retrieval system does not ground answers in unverified assumptions.
Leila wants the assistant to support an executive recap without turning guesses into facts. Before: every captured item looks equally authoritative. After: evidence and assumptions are visibly separated before retrieval. Status became explicit, so the assistant can treat verified facts as evidence and hypotheses as assumptions. The evidence lane keeps source-backed material close to executive wording, while guesses stay useful without becoming false grounding. Open questions are preserved instead of hidden, which stops the model from filling uncertainty with invented confidence. Separate what you know from what you think before AI retrieves from it. source status hygiene Practiced labeling captured…
Sign up free: one personalized lesson every day, matched to your role and goals.
Already have an account? Sign in