Communications artefact · DEC-30 · 2026-07-15

ASSAY ↔ doctrine

Two things an SME should have in hand before exploring the demonstrator: the crosswalk — which JIPOE and joint-planning products ASSAY computes, and how faithfully — and the divergence register — the small number of places ASSAY departs from written doctrine on purpose, each with its defence stated up front.

Read the crosswalk first. ASSAY is best understood not as a claim to be doubted but as an automation of products the staff already owes: MCOO, event matrix, synchronisation record, running estimate, the CCIR loop. Where a row says gap, the gap is admitted and tracked; where the register says divergence, the departure is deliberate and defended. Nothing here is discovered by exploring — it is stated.

1 · The crosswalk — what maps, what's absent

Sources: JP 2-01.3 (JIPOE), ATP 2-01.3, JP 5-0, FM 5-0. Since SPEC-21, the step mapping is machine-carried: every Meridian knowledge object names its originating JIPOE step on its provenance chip (jipoe_step), pinned by a fixture test — the "doctrinally shaped" claim is auditable, not asserted.

JIPOE / JPP artefactASSAY analogueFidelity
Step 1 — define/bound the OE; significant characteristics; gaps → collection requirements Absent — the demonstrator starts at "here are the fourteen questions that matter" (forward derivation is the horizon programme, issue #43) gap — admitted
Step 2 — describe effects; MCOO overlays, mobility classes; weather Sparse channels (default + RegionOverride, DEC-29), explicitly derived from MCOO doctrine; K5/K9 weather with validity windows strong
Step 3 — evaluate adversary; ORBAT, capabilities; HVT list; COG analysis Red ORBAT is vignette prose, not typed objects; no HVT/COG shapes — a deliberate deferral (thesis G), stated here rather than discovered partial — by design
Step 4 — adversary COAs; situation/event templates; event matrix; MLCOA/MDCOA R1–R3 as first-class ScenarioCOAs; expected_answers = the event matrix (DEC-18); minimax = the MDCOA discipline generalised strong on products — hand-authored on derivation (#43)
Indicators / NAIs / collection Open questions with collection options, banded cost, earliest_result; the discrimination ranking strong conceptually — NAI geography untyped ("where to look" is textual)
PIR / CCIR tied to decisions A committed decision re-prioritises verification of the knowledge under it — the CCIR loop, computed (walkthrough beat 5) strong — the single best C2 moment in the system
Decision support template / matrix (DP × NAI × TAI × LTIOV) Ingredients all present (commitments, verdicts, validity windows, earliest_result, discrimination); the assembled artefact is absent the gap most worth closing — queued as SPEC-24
COA validity: feasible / acceptable / suitable / distinguishable / complete Hard constraints / commitments / C1-as-mission / the handful + distinct_because / completeness not addressed (plans are routes + windows, DEC-20) strong — distinguishability operationalised, the criterion staffs most often fake
COA analysis — wargaming (action–reaction–counteraction) Absent; the scorer evaluates static compiled worlds. ASSAY should instrument the wargame, not replace it gap with an honest bridge
COA comparison — weighted decision matrix Deliberately deleted; replaced by the verdict matrix + banded non-domination + stated tie-breaks divergence №1 — defended below
Synchronization matrix / wargame worksheet The verdict matrix + delta log are a computed, always-current synchronisation record strong — previously unclaimed; claimed here
Running estimates Staleness flags = the machine-checkable running estimate: is the plan still valid, and exactly where not strong; novel — doctrine has no mechanism linking a specific stale fact to specific verdicts
Assessment (MOE/MOP), execution Out of scope — no execution model divergence №5 — stated

2 · The divergence register — deliberate departures, defended

ASSAY departs from written doctrine in five places. Each departure is a decision with a defence, recorded here so it is a statement, never a discovery.

№1 — No weighted COA decision matrix

Doctrine says
JP 5-0 offers the weighted matrix as a sample COA-comparison technique.
ASSAY does
No numeric weights exist anywhere (ASSAY-DEC-19). Comparison is the honest verdict matrix, banded non-domination for the handful, and tie-breaks stated in prose.
Defence
The MCDM-critique literature has never rehabilitated the technique: subjective weights project false rigour, ordinal scales get treated as cardinal, winners flip under tiny weight changes, and naturalistic-decision-making evidence shows the matrix is routinely back-filled to justify a choice already made. ASSAY delivers the matrix's legitimate function — organised comparison, distinct_because — without the illegitimate one.

№2 — No scenario-probability arithmetic

Doctrine says
Rank-order adversary COAs by likelihood; planning often proceeds MLCOA-first.
ASSAY does
Holds the likelihood ordering as banded knowledge (K14a–c, scenario_weight) and refuses to compute with it: weights order attention and reporting, never compile into constraint or cost. Robustness is worst-case (minimax) across the COA set — the MDCOA discipline — never a weighted blend.
Defence
Aligns with the most-dangerous-COA requirement, with robust decision-making / assumption-based planning / info-gap methods, and with ICD 203 / PHIA's stance that likelihood is a human judgement. ASSAY leaves the judgement where doctrine puts it — with the analyst and the commander.

№3 — No adversarial reaction dynamics

Doctrine says
COA analysis is the wargame: action, reaction, counteraction.
ASSAY does
Supports COA comparison honestly over static compiled worlds; it does not simulate a reacting adversary.
Defence
The wargame remains a human activity ASSAY should instrument — honest adjudication support, traceable verdicts on demand — not replace. A computed reactive red (thesis H) is deliberately deferred until the human-in-the-loop version has taught the project what it wants.

№4 — No civil-harm quantification

Doctrine says
Acceptability weighs cost against advantage; LOAC proportionality weighs anticipated harm.
ASSAY does
C3 is a geometric red line — does FE-FALCON fire into the populated district — never a weighted harm metric (ASSAY-DEC-35). The civil-population band (K3) is compiled and displayed but no verdict arithmetic reads it.
Defence
Whether firing into 55,000 people is "worse" than 5,000 is a commander's value judgement the system refuses to pre-compute (DEC-19). The system declines the weighing; the commander is not spared it — the trade is surfaced, banded, and left on the commander's desk, which is exactly where JP 5-0's acceptability criterion puts it.

№5 — No execution or assessment model

Doctrine says
Planning continues into execution: MOEs, MOPs, assessment working groups.
ASSAY does
Stops at plan selection with a recorded rationale. No execution model, no MOE/MOP shapes.
Defence
Scope, stated: the demonstrator's theses concern knowledge → compile → score → analyse. The staleness machinery (thesis F) is the seed a future assessment loop would grow from.

Mission-command-compatible automation. The common thread: ASSAY accelerates the auditable parts — compile, score, trace, staleness, at machine speed — and refuses to accelerate the judgement parts — likelihood, value trade-offs, selection, at commander speed. The refusal posture is a C2 position, not an anti-automation one.