Leading with a result and stating the assumption is a third path between asking and running

What it is

There is no need to ask “decline or accept” first, and no need to silently write the whole letter as a decline. Hand over a decline draft, and write beside it: “I wrote this as a polite no; say if it is a yes.” What the person sees is material plus a premise they can overturn — not a question, and not a fait accompli. That is result-plus-assumption, between asking and running: action has started, interpretation still sits with the person.

The third path holds when the assumption is visible and switchable in one beat. Hide the assumption inside the generation, and it collapses back to execute-by-assumption. State the assumption and give no material, and it collapses back to a follow-up.

Why it happens

People are better at judging a concrete draft than at answering an abstract multiple-choice about a thing that does not exist yet. “Is this decline letter right” carries more information than “do you want to decline,” because the draft already spreads tone, length, and whether a next meeting is mentioned. Writing the assumption as a clickable tag keeps switch cost on the tag: “make it an accept” is one regeneration under a new assumption, not a round of Q&A first.

The path depends on the result being discardable. If the first result has already been sent or already charged, a clearly written assumption is still late. So it serves retractable generation and previewable plans, not world acts that have already landed.

Studying it

Three arms on the same vague instruction: ask then act, act on the default with no labelled assumption, draft first with the assumption labelled. Dependent variables: turns to an acceptable result, rate at which the assumption is overturned, whether overturn uses the tag, mis-send or mis-execute rate. Also measure “I know what it just guessed” — only the labelled arm should sit well above the other two.

A high overturn rate is not failure. High overturn plus low total turns means the third path is working: people use the draft as a probe and the tag as a gear change. Low overturn plus high mis-send means the tag was not read.

Where it stops holding

Acts on a world boundary cannot replace confirmation with “do it then label it”; a tag is not a gate. When the assumption space is larger than two and the readings are mutually incompatible, one draft is not enough — show two side by side, or fall back to asking that one dimension. A tag that looks like a disclaimer (tiny grey type, a long sentence) will not be read, and the third path does not exist visually. When the user has asked to be asked first, do not break their control rhythm with a draft.

Applying it

• For previewable generation, default to one draft, with the assumption as a short tag above it that can be clicked to the other reading.
• The tag must be human speech naming a concrete reading (“written as a decline”), not “based on current understanding.” A click to switch should keep the parts that are already right and change only what the assumption touches.
• The moment an act would leave preview (send, pay), step off the third path and spread the reading before execution.
• Check: on “reply to this mail,” the UI hands over a decline draft plus tag. Count how many people click the switch, how many send the draft as final, how many ignore the tag and rewrite. High ignore-and-rewrite means the tag does not behave like a control. Make the tag two clickable readings; switches should rise and rewrites should fall.

Related

• Same group: L2.15.1 Assuming one reading and running with it is faster, and it also enlarges the cost of being wrong · L2.15.2 A follow-up is worth asking only when the ambiguity would lead to clearly different outcomes · L2.15.3 Too many follow-ups turn the exchange into a questionnaire and cancel the efficiency natural language was for · L2.15.5 Weigh the cost of clarifying against the cost of undo; if undo is cheap, do not ask
• Nearby: L3.01 Generating Multiple Options · L3.07 Multi-option Generation and Side-by-side Comparison · L2.13 Control Granularity and Editability
• Search terms: result-plus-assumption · assumption made visible · draft as probe