Deciding with incomplete information

A scholar transcribes a damaged manuscript. Some letters are visible; others are inferred from a familiar phrase. The draft marks those categories separately. A targeted image of a damaged letter could distinguish two readings, whereas a sharper image of an already clear margin would not settle that uncertainty. The scholar can circulate a provisional transcription while stating which new observation would warrant a correction.

Manuscript work	Decision work
Visible letter	Observed information
Inferred missing letter	Assumption
Quality of the available image	Basis for confidence
Image of the disputed mark	Decision-changing information
Provisional reading and correction condition	Announcement and revisit trigger

The analogy concerns evidence, not the preservation or handling of historical objects. A word can remain ambiguous after additional inspection, just as another test may not resolve a decision. More detail is not automatically more useful. The missing fact matters when plausible answers would lead to different choices under the team's already agreed criteria. NASA's decision-analysis guidance supplies that distinction; the scenario adapts it to software rather than claiming NASA's project process is a software standard. [S2]

The analytic documentation principles come from ODNI's analytic standards. This lesson borrows their distinction between assumptions and evidence. The cost-estimate confidence example uses the Infrastructure and Projects Authority's input-maturity rule. It does not apply intelligence-agency policy to a software team. A confidence label is a statement about support for the assessment, not a made-up numerical probability. If a scenario supplies a probability for arithmetic, that input is explicitly an assumption. [S1]