ATTRIBUTE · DESK VIEW · ADMIRALTY-RATED

AZIMUTH — attribution engine

AZIMUTH turns a detected behavior into a characterized, rated attribution — mapped to DISARM, structured in ABCDE, and assessed only to the level the evidence supports. Confidence cannot outrun its evidence tier; at least two hypotheses stay on the board; and an honest “Unknown actor” is a normal, frequent verdict here, not a failure. Desk view — Admiralty two-axis (source A–F / info 1–6) only; never rendered alongside the public three-state standard.

Attribution case queue

BY CONFIDENCE BAND · LINKED NARRATIVE
Switch cases here or from the queue — useful for comparing an “Unknown actor” verdict against an attributed one.
≥2 hypotheses — always onEvery case keeps at least two hypotheses visible. The assessed actor must equal the hypothesis rated best_supported.
Confidence ceilingThe band cannot exceed the strongest rated evidence tier present. Raising it further is blocked in the UI unless logged with a justification.
Nonpartisan by constructionActors are a CLASS (state / commercial / domestic / autonomous / unknown) + behavior — never a party, campaign, or candidate. See the lint demo below.

ABCDE structure

PER-FIELD ADMIRALTY

DISARM technique mapping

VERSION-PINNED

Alternative hypotheses

KEPT OPEN · ≥2 REQUIRED

Confidence ladder

HONEST CEILING
In-memory demo only — nothing here is persisted.

Evidence basis

RATED · EACH WITH A FALSIFICATION CONDITION

Nonpartisan lint — try an actor label

DEMO · BLOCKS PARTY / CANDIDATE NAMES

AZIMUTH will not accept a party, campaign, or candidate name as an assessed_actor. Actors are described by class + behavior. Type a label below and check it.

Blocklist shown is illustrative for this prototype; production AZIMUTH lints against a maintained party/campaign/candidate list at save time, per house convention §4.