Skip to main content

Minimal Set to Decide and Handling of the Domain Not Decided

Context​

In many contexts, a state where "decisions cannot be made because there is insufficient context" has become normalized.

However in reality, the problem is often not that necessary information is lacking, but that assumptions needed for decisions are not fixed.

Minimal Context is not an attempt to reduce the amount of information needed for decisions. It is a framework for defining the types of assumptions that must be fixed for decisions to hold.

Objective​

The purpose of this topic is not complete understanding or comprehensive grasp.

It lies in differentiating assumptions that should be fixed and domains intentionally not fixed to create a state where decisions can be taken on even under incomplete situations.

Minimal Inputs​

Minimal Context is composed of the following four inputs.

1. SoT (Source of Truth)​

Fix what is treated as the "source of truth."

  • Location of the source of truth for data
  • Ultimate basis for decisions
  • Priority when contradictions arise

If SoT remains ambiguous, all discussions and corrections become relative.

2. Constraints​

Make explicit the conditions that limit choices.

  • Time, contracts, regulations
  • Organizational structure or responsibility distribution
  • Technically unavoidable assumptions

Constraints are not things that bind decisions, but boundaries that make decisions hold.

3. Non-Changeables​

Clarify assumptions excluded from discussion or consideration.

  • Policies agreed not to change
  • Design decisions not to touch this time
  • Conditions accepted as external factors

If these are not made explicit, everything is treated as implicit variable items.

4. Uncertainties (Intentionally Leave Uncertainty)​

Explicitly leave as uncertainty what cannot be determined at this time.

  • Items undetermined due to lack of information
  • Hypotheses with high verification cost
  • Decisions entrusted to future learning

If uncertainty is not made explicit and one proceeds, assumptions tend to be rewritten later.

Working Model​

Minimal Context includes a working model for handling incomplete hypotheses.

  • Which assumptions are hypotheses
  • What constitutes confirmation
  • Until when hypotheses are allowed

If hypotheses have no deadline, uncertainty behaves as fixed assumptions.

Tactics​

Intervention to make Minimal Context function is not increasing information.

  • Separate discussions unnecessary for decisions
  • Make fixed points explicit and limit variable domains
  • Treat undetermined items as hypotheses

These are not procedures, but structural interventions to make decisions hold.

Risks​

Minimal Context also has risks.

  • Fixed points become excessive and exploration room is lost
  • Uncertainty is neglected and not reconsidered
  • "Not deciding" is used as evasion of responsibility

These appear as failures in the opposite direction from Context Erosion.

Interaction with Failure Patterns​

Minimal Context strongly relates to the following Failure Patterns.

AI Collaboration​

When using AI as decision support, Minimal Context becomes essential.

Not whether information passed to AI is abundant, but whether fixed assumptions are included determines output quality.

  • SoT: What to treat as the source of truth
  • Constraints: Boundaries not to cross
  • Non-Changeables: Assumptions not changeable
  • Uncertainties: Domains where speculation is allowed and not allowed

AI output tends to lack consistency.

Last updated on by Hiroshi Kurabayashi