The method, rendered live

1 Define the Problem

“What decision am I really trying to make?”

Framing is the highest-leverage step — most bad decisions are well-run answers to the wrong question. Stating the decision as a question whose answer IS the decision keeps everything after it honest. (Smart Choices, Hammond/Keeney/Raiffa, Ch. 2.)

Questions the agent will probe

• What triggered this, and is the trigger the real problem or a symptom?
• Which constraints are real vs. self-imposed?
• What are the essential elements of the problem?
• What other decisions does this hinge on or affect?
• Is the scope right — broad enough to be worth solving, narrow enough to act?
• How would someone else (or my own future self) frame this?
• Am I still working on the right problem?

"Done enough" — the checklist read back to you

• Read cold by a stranger, does the statement capture what you are actually deciding?
• Is this the root decision, or a symptom/trigger of something bigger?
• Is the scope right — broad enough to matter, narrow enough to act on?
• Are the constraints we listed real, or assumed (and worth challenging)?

Traps this step watches for

• Framing — The way the question is worded is steering the answer — a yes/no frame, a loss-vs-gain frame, a borrowed frame.
• Anchoring — The first number, option, or idea on the table is pulling every later estimate toward it.

Where you can go from here

• forward Clarify Objectives — The decision is framed — clarify what matters (the objectives) for THIS decision.

2 Clarify Objectives

“What do I really want out of this decision?”

You separated ends from means — the crux of good objectives. The fundamentals are the ends you'll actually weigh; the means are routes to them. Keeping them apart stops a means from masquerading as a goal. (Smart Choices, Ch. 3.)

Questions the agent will probe

• What would make me really happy here (the wish list)?
• What do I most want to avoid (the worst outcome)?
• What's the impact on others — what do I want for them?
• Why is that important? (ask ~5×, to separate ends from means)
• What do I really mean by this objective, so I can tell if it is met?
• Could I live with the choice these objectives produce — and explain it to someone?

"Done enough" — the checklist read back to you

• Is there an objective for everything that matters — including the unglamorous ones (cost, time, risk, stress)?
• Have we separated true ends (what you ultimately care about) from means (instrumental)?
• Did we run the "why does that matter?" ladder at least once, down to a fundamental?
• Would you be comfortable having the decision judged by these objectives — nothing important left off?

Traps this step watches for

• Status-quo — The current / default option is getting a pass the alternatives don't — "let's just keep doing this."
• Framing — The way the question is worded is steering the answer — a yes/no frame, a loss-vs-gain frame, a borrowed frame.

Where you can go from here

• forward Generate Alternatives — Objectives are clear — generate real options that could achieve them.
• loopback Define the Problem — The work exposed that the wrong problem was framed — restate the decision before continuing.

3 Generate Alternatives

“What are my real options for pursuing my objectives?”

Three or more genuinely different options are on the table — the quality of a decision is capped by the quality of its alternatives, so widening the set past the obvious one is exactly the move. (Smart Choices, Ch. 4.)

Questions the agent will probe

• How could I achieve each objective (turn means into options)?
• Which constraints could I assume away?
• What would I do if my target were far beyond reach (high aspirations)?
• What have others done — and whom, even outside the field, could I ask?
• Is the deadline real, or is waiting itself an option?
• Stop-test: do I have a genuine RANGE, with some options distinctly different?

"Done enough" — the checklist read back to you

• Are these genuinely different options, not variations of one idea?
• Is the status quo (staying / doing nothing) explicitly on the table?
• Did we ask "what else?" and "what if the obvious option were off the table?"
• Is there at least one option you had not considered before this conversation?

Traps this step watches for

• Status-quo — The current / default option is getting a pass the alternatives don't — "let's just keep doing this."
• Anchoring — The first number, option, or idea on the table is pulling every later estimate toward it.
• Confirming-evidence — You're gathering support for an option you've already half-picked, and discounting what cuts against it.

Where you can go from here

• forward Evaluate Consequences — Options exist — describe how each performs on each objective (the consequences table).
• loopback Clarify Objectives — Generating a real option surfaced a value you had not named — add the objective before you score.
• loopback Define the Problem — The work exposed that the wrong problem was framed — restate the decision before continuing.

4 Evaluate Consequences

“How well does each option actually perform on each objective?”

You described how each option actually performs on each objective — concrete consequences, not gut impressions. That table is what makes the tradeoffs visible instead of hand-waved. (Smart Choices, Ch. 5.)

Questions the agent will probe

• If I'd already chosen this option, what is daily life actually like?
• What's the consequence, in the words or numbers that capture it?
• Do my descriptions cover ALL objectives — and do any imply an objective I had not named?
• Is each consequence accurate, complete, and precise?
• What scale truly captures this objective, even the intangibles?
• Can I knock out any option as clearly inferior?

"Done enough" — the checklist read back to you

• Is every option scored honestly on every objective — including where your preferred option looks bad?
• Does any objective score the same across all options? If so it is not discriminating — revisit it.
• Are the scores what you actually expect, not what you hope?
• Did any consequence reveal an objective we had not named?

Traps this step watches for

• Overconfidence — Estimates and ranges are suspiciously tight — "it'll take about two weeks," full stop.
• Recallability — A vivid, recent, or dramatic event is dominating the estimate out of proportion to how often it really happens.
• Confirming-evidence — You're gathering support for an option you've already half-picked, and discounting what cuts against it.

Where you can go from here

• forward Resolve Tradeoffs — The table is scored — resolve the tradeoffs down to a recommendation.
• loopback Clarify Objectives — A row where every alternative scores about the same — that objective is vague or redundant; sharpen it or drop it, then re-score.
• loopback Define the Problem — The work exposed that the wrong problem was framed — restate the decision before continuing.

5 Resolve Tradeoffs

“What will I give up on one objective to gain on another?”

You made the tradeoff explicit instead of papering over it — swapping real amounts between objectives, not arguing abstract importance. That's the difference between a real decision and a rationalization. (Smart Choices, Ch. 6.)

Questions the agent will probe

• Can I eliminate any option by dominance (better on some, worse on none)?
• By practical dominance (only a small, lone advantage left)?
• What change would cancel an objective between two options (even-swap, step 1)?
• What change elsewhere compensates for it (the real value judgment)?
• Did making the swap create a new dominance?
• Am I weighing the AMOUNTS, not the abstract "importance" of an objective?
• Are my swaps consistent (A>B, B>C ⇒ A>C)?

"Done enough" — the checklist read back to you

• Have we made the hard trade explicit — what you would give up on one objective to gain on another?
• Does the recommendation trace back to the objectives and the consequence table?
• Is this YOUR call, reflecting your priorities — not just what the table mechanically implies?
• Did we prune clearly-dominated options before swapping?

Traps this step watches for

• Sunk-cost — A past investment of time / money / effort is being used to justify continuing.
• Confirming-evidence — You're gathering support for an option you've already half-picked, and discounting what cuts against it.
• Framing — The way the question is worded is steering the answer — a yes/no frame, a loss-vs-gain frame, a borrowed frame.

Where you can go from here

• loopback Generate Alternatives — No alternative is good enough — decision quality is capped by the best option you considered; go invent a better one.
• loopback Evaluate Consequences — The scores are too coarse to swap on — refine the consequence table before trading off.
• loopback Define the Problem — The work exposed that the wrong problem was framed — restate the decision before continuing.
• skip Assess Uncertainty — Include Uncertainty when the outcome hinges on unknowns whose resolution could change which option you pick (the leverage filter). Otherwise skip it.
• forward Commit to Action — The tradeoffs are resolved into a recommendation — ask the user to accept it and commit: the first action, who owns it, and when to review.

6 Assess Uncertainty optional

Where you can go from here

• skip Evaluate Risk Tolerance — Include Risk Tolerance when a worst-case scenario could be intolerable or hard to reverse — can you survive it?
• forward Commit to Action — The risk profile is sketched and the recommendation stands — move to commitment.

7 Evaluate Risk Tolerance optional

Where you can go from here

• skip Identify Linked Decisions — Include Linked Decisions when this choice constrains, or is constrained by, future decisions.
• forward Commit to Action — The worst case is survivable (or the recommendation was adjusted) — move to commitment.

8 Identify Linked Decisions optional

Where you can go from here

• forward Commit to Action — Downstream decisions are named and staged — move to commitment.

9 Commit to Action

“Is this decision going to get acted on — and how will we know?”

A recommendation without commitment is a study, not a decision — the decision-quality chain is only as strong as its weakest link, and link 6 is commitment to action. Recording the first action, its owner, and a review date is what turns the analysis into something the world will notice. (Decision Quality, Spetzler/Winter/Meyer — the six links.)

Questions the agent will probe

• Did the user say yes, explicitly, in their own words — or did I assume it?
• Is the first action concrete enough to do within ~72 hours?
• Does the named owner actually own it (can they do it without permission we have not got)?
• Is the review date tied to when reality will have something to say?
• If the user hesitated — what is the hesitation telling us about the tradeoffs?

"Done enough" — the checklist read back to you

• Read the commitment back cold: does it sound like the user deciding, or the agent concluding?
• Could the owner do the first action this week without anything else happening first?
• When the review date arrives, will you be able to tell whether the decision held?

Traps this step watches for

• Status-quo — The current / default option is getting a pass the alternatives don't — "let's just keep doing this."
• Overconfidence — Estimates and ranges are suspiciously tight — "it'll take about two weeks," full stop.

Where you can go from here

• loopback Resolve Tradeoffs — The user hesitates to commit — the recommendation has not really survived contact with reality. Reopen the tradeoffs rather than recording a hollow commitment.