ahp-comparison-matrixlisted

# AHP comparison matrix ## Scale | Answer | Value | Reciprocal filled at M[j][i] | |---|---|---| | About the same | 1 | 1 | | More | 5 | 0.2 | | Much more | 10 | 0.1 | Only these three answers exist. Never offer other magnitudes. ## Asking - Group criteria into clusters first. **Around 5 per matrix is the sweet spot** (10 pairs). Pairs grow quadratically, and answers get noticeably less considered past roughly ten questions. This is guidance, not a hard limit — if the user wants 7 or 8, run it, but say once that consistency usually suffers and offer to cluster instead. - Define every cluster in plain language **before** the first pair. Name, what it covers, what high and low look like. One short block each. ### Do not use the interactive picker `AskUserQuestion` caps options at 4. The scale needs 5. Routing the fifth through "Other" makes four answers one click and the fifth free-text typing — and people pick what's visible, which tilts every pair toward whichever criterion is listed first. Over ten pairs that bias is larger than the signal being measured. Ask in **plain text** instead, with a symmetric 1-5 spectrum the user answers by number. ### The prompt format Every pair, no variation. **Write it as ordinary prose and a normal markdown list — never inside a code block.** These are questions for a person to read, not code; monospace makes them look like output to skim past rather than a question to answer. The rendered result should look like this: > **