Framework Layer

The Framework transforms a bounded Problem Space into explicit decision variables, rules, legitimacy tests, and validation criteria.

Transformation Contract

• What transforms into this layer? A problem space containing ambiguity, constraints, candidate variables, and a decision requirement.
• What does this layer produce? A structured decision model: relevant variables, relationships, legitimacy rules, assumptions, and tests.
• How does it improve future decisions? Reusable rules reduce framing errors and make missing variables or invalid assumptions easier to detect.

Closed-Loop Role

• Depends on: Problem Space definition and feedback from Reports and Cognitive Assets.
• Improves: decision structure, variable extraction, consistency, and testability.
• Modifies upstream: It can expose that the original Problem Space was incomplete, misframed, or not a legitimate decision problem.

Reports refine the Framework when empirical evidence contradicts a rule, reveals a missing variable, or changes a validation threshold. Cognitive Assets preserve those revisions so future problems do not restart from zero.

Decision Legitimacy

A legitimate decision requires more than a useful output. It requires a valid authority source, an accountable responsibility chain, a known automation boundary, and review conditions appropriate to the consequence of the action.

Output, Recommendation, Action, Decision

An output is generated material. A recommendation proposes a possible course of action. An action changes the state of the world. A decision determines that an action should be taken, delayed, changed, or refused under identifiable authority.

Confusing these forms creates judgment risk. AI can generate outputs and recommendations, and automated systems can execute actions. Neither fact alone establishes a legitimate decision.

Core Dimensions

• authority
• responsibility
• reversibility
• explicit delegation
• human review
• automation boundary

Framework Essays

• What Makes a Decision Legitimate
• When a Decision Is Not a Decision
• Limits of Automated Judgment
• Decision Without Explicit Authority Is Not Legitimate
• Automatic Execution With Irreversible Effects Is Not Legitimate
• AI-Generated Advice Does Not Transfer Decision Responsibility
• Responsibility Cannot Be Shifted by Default System Rules

No Layer Is Terminal

The Framework is revised by evidence and can revise the Problem Space upstream. A framework that cannot be challenged by outcomes is not part of the closed-loop system.