Lifecycle Overview

An intelligence result advances only when the candidate cohort, evidence snapshot, challenge record, scenario set, and decision policy are explicit. The default endpoint is advisory decision support; enforcement requires a separate, attributable promotion.

stateDiagram-v2
    [*] --> CandidateSet
    CandidateSet --> Challenged: evidence posture bound
    Challenged --> Evaluated: scenarios and policy applied
    Evaluated --> Advisory: recommendation supported
    Evaluated --> Hold: uncertainty or conflict dominates
    Evaluated --> Refused: governed prerequisite fails
    Advisory --> Enforced: explicit policy promotion
    Advisory --> Superseded: evidence or policy changes
    Hold --> CandidateSet: new evidence arrives
    Refused --> CandidateSet: missing evidence supplied

Evaluation lifecycle

Candidate validation and fingerprinting establish the comparison set. Evidence posture then records support, provenance, gaps, and readiness. Contradiction and falsifier passes challenge the favorable account before scenarios are evaluated.

Scenario consensus is not forced. Conflicting actions, high hold pressure, or a wide confidence spread trigger human escalation. Unresolved questions remain attached to the result. Evidence gates can downgrade a recommendation or force a hold, and the downgrade chain records why.

Authority lifecycle

IntelligenceDecisionSupportEnvelope is advisory by default. Promotion to enforced policy requires a policy identifier, named promoter, and rationale. That transition grants operational authority outside the reasoning calculation; it does not make the underlying evidence stronger.

When evidence, candidates, thresholds, weights, or scenario policy changes, the previous result remains a historical decision record. A new evaluation supersedes it rather than rewriting its basis. Learning can propose a policy adaptation after outcomes are reviewed, but existing recommendations retain the exact policy under which they were produced.

Decision dossier by state

State	Required contents	Permitted exit
candidate set	included and excluded candidates, eligibility policy, rejection reasons, and cohort fingerprint	bind one evidence revision
challenged	evidence posture, contradictions, falsifiers, missing support, calibration context, and review requirements	evaluate or refuse
evaluated	component scores, orientation, weights, thresholds, ties, sensitivity, alternatives, scenarios, and regret	advisory, hold, or refusal
advisory	preferred posture, rationale, uncertainty, downgrade path, unresolved questions, and human-review state	remain advisory, supersede, or promote explicitly
enforced	policy identity, promoter, rationale, time, target authority, and source advisory record	execute through the responsible downstream owner
superseded	replacement decision and changed evidence, candidate, or policy identities	remain immutable for audit

A dossier is incomplete when it retains only the preferred candidate. Rejected alternatives and refusal pressure are part of the explanation because they show whether the outcome came from evidence, policy, or an artificially narrow choice set.

Supersession and feedback

Observed outcomes return as new evidence, not as edits to the old decision. Before learning can affect later policy, the outcome record must identify the original recommendation, actual action, operational deviations, observation, eligibility for learning, and reviewer disposition.

Changed input	Required response
evidence record, freshness, or contradiction	issue a new evidence-bound evaluation
candidate inclusion or hard filter	issue a new cohort fingerprint and decision
score orientation, weight, threshold, or tie rule	issue a new policy identity and comparison
calibration corpus or confidence meaning	re-evaluate confidence language and affected decisions
observed consequence	append a reconciliation record; preserve the original rationale
promotion authority	append a new authority event; do not relabel the advisory calculation