BC-DS · Solo Decision Architecture · 3.0

Solo Decision Architecture.

The model of decision architecture for a consequential decision you open for yourself, on your own authority, and answer for alone. Solo means accountable alone, not working alone.

01 · What SDA is for
A decision held by one person

Distortion can form inside one person’s thinking before it reaches anyone else.

Solo Decision Architecture (SDA) is for a consequential decision that one person structures from their own perspective and answers for alone, such as whether to move a family to another city. Which model governs a case depends on how it is opened, not on its size: a case you open for yourself, on your own authority, is a Solo case.

Solo does not mean working alone. The people who matter to the decision are named, and their input is recorded. The decider is simply the one who commits and answers for it.

AI is often present at the very moment a person frames a decision, and it builds fluently on whatever frame it is given. SDA captures the decider’s own account before any AI takes part, and keeps the act of deciding with the decider.

“Write down where you want to arrive before examining the present or asking AI for analysis.”

Your account comes first, in your own words, before any AI.

Your words

Where you want to arrive is described before the present is read.

Your destination

Each assumption the plan depends on is tested, and what remains uncertain is named.

Your assumptions

One recorded commitment, with an owner, a date and the conditions that would reopen it.

Your commitment
A hypothetical example

A family move. Someone is deciding whether to move their family to another city. The belief that the family can relocate is an assertion. Confirmed school places for two children, an agreement that lets a partner work remotely and six months of housing costs set aside are facts that can be verified.

What the session does. It gives a structure for testing each assumption the plan depends on, recording what is verified and naming what remains uncertain.

02 · Backward from the destination
The Dual Lens Principle

Describe where you want to arrive. Then read where you stand.

SDA applies the discipline’s foundational principle: the future you want is described first, and the present is read against it. The two are compared until neither changes. Both stay hypotheses until the commitment is carried out, but after the crossing the destination changes only through a second recorded act.

Deciding forward

Everything at once.

You start from today and weigh every factor together. None comes with a clear priority, so bias sets the priorities without anyone choosing them.

“Should we move? There are the schools, the job, my mother, the money...”

Deciding backward

One destination first.

You start from the situation you want to create. Every factor still matters, but now each has a priority: how does it help reach the destination?

“The family settled in the new city, my career moving again, the children at schools they like. What has to be true for that?”

01
The desired future state

A concrete description of the situation you want to create. It stays a draft until you fix it at the crossing.

02
The finish line

A marker set in advance, with a date, to recognize when you have arrived.

03
Guardrails

The things you refuse to lose on any route.

04
The default future

What is likely to happen if you decide nothing. Waiting is also a choice.

03 · AI in the session
AI proposes. You decide.

What AI may do, and when.

SDA does not exclude AI. It sets when AI takes part and what it may do. Every part AI offers arrives beside your own words, never in place of them, and nothing enters the record unless you elect it.

1
Before AI · You alone

Your account and your sorting

No AI at all

Your account is captured in your own words. You sort it yourself before any AI sees it, so the first statement of what you said is yours.

  • Your account. Kept as given, word for word.
  • Your sorting. Destination, what drives the decision, the pressure you feel and today.
  • A note. Whether you asked an AI about the decision before the session is recorded.
2
After you · AI beside you

Proposals you elect or reject

AI adds nothing you did not say

Once you have sorted your account, AI may offer its own sorting beside yours, list the people you named as candidates for the seats and hold any solution you named too early, unevaluated.

  • Its sorting. Beside yours, never instead of it.
  • The other side. The strongest argument against the route, always asked for.
  • Marked and dated. Every output labeled as AI’s.
The own-words rule
A blank field is correct. An invented field is a failure.

No part of the sorting may contain anything you did not say. Your election is the only route by which anything enters the record.

AI may
  • Suggest how to sort your words, after you have sorted them
  • List the people you named as candidates for the seats
  • Hold a solution you named early, unevaluated
  • Make the strongest case against the route
  • Return a tension as a question, not a verdict
  • From phase 3, ask a question that carries a sourced fact, shown beside the record until you answer it
AI may not
  • Take part before your account is captured
  • Draft the destination
  • Put anything into the record that you did not say or confirm
  • Hold a seat or pass the commitment gate

You give the account. AI may propose and sort. You make the decision.

04 · The session
Nine phases, shortened for a light case

From your own account to a recorded commitment.

The SDA framework follows a session of nine phases. A light case uses a shorter session.

1

Your account

You record or type everything about the decision, in your own words and order. No questions, no fields and no AI.

2

Your sorting

You sort your own account first: where you want to be, what is driving the decision, the pressure you feel and today as you see it. Only then does AI offer its sorting beside yours, adding nothing you did not say.

3

The finish line and the reading

You state the finish line, with its date, and the things you refuse to lose. Only then is the present read against them, each assumption stated so it can be tested and the default future projected.

4

The route

You build the route backward from the destination to today, one milestone at a time, each a state you could photograph: a signed document, a published page. Then you write your strongest case against it, before AI writes its own.

5

The seats

You name who has standing at the commitment, and list everyone the decision affects, including those who will bear the consequences.

6

The cost

A cost statement, drafted for you as a proposal, names who gains and who bears the cost. You confirm or correct it and accept responsibility for the cost. If you hesitate, the hesitation is treated as information.

7

Convergence

Two readings, forward and backward, test the destination, today and the route against each other, until comparing them changes none of them.

8

The crossing

You fix the destination in a recorded act. After it, the destination changes only through a second recorded act, with a written reason.

9

The commitment gate

The check before you commit: the cost still accepted, the seats filled, the needed input on record. You commit, decline or hold.

A decision that looks made but was postponed, settled too early or announced with nothing to make it binding is not a commitment.
A model for decisions you own

Write down where you want to arrive.

The Decisiontect beta is coming. Decisiontect is BC-DS’s commercial app. SDA itself is open: you can apply it without the app. Meanwhile, write down where you want to arrive before you read the present.

Built from tested parts

SDA combines established methods: working backward from a destination, known as backcasting (Robinson, 1990); defining values before options (Keeney, 1992); independent judgments before group discussion; structured dissent; and the premortem (Klein, 2007). Whether the combination leads to better decisions than the parts used separately has not yet been tested; the papers state it as a hypothesis.

SDA is set out in its paper, version 3.0, which also names the parts it does not yet cover: doi.org/10.5281/zenodo.23626329.