Decision Premise Register¶

A standing register — instantiates Premise–Action Decoupling

A standing ledger that pins each premise a decision rests on to a named owner and to the downstream choices that would have to be reopened if the premise falls.

Decision Premise Register is the durable ledger where a decision's premises live after they have been surfaced — one row per premise, carried forward in time. Its job is not to find assumptions or grade them; it is to keep them from going feral. Each premise gets two things it otherwise never has: a named owner who is accountable for revisiting it, and a recorded link to the decisions that depend on it, so that when a premise is revised, everything built on top of it can be found and reopened. This is the archetype's memory: it is what stops an "apparently settled" consent, jurisdiction, or forecast from being silently inherited downstream as fact, because the register holds it as a premise-with-a-warrant-and-an-owner, not a closed question. Where the Assumption Audit is the act of examining premises, the register is the artifact that outlives the act.

Example¶

An investment committee approves a leveraged acquisition. The memo reads as a decision; the register rewrites the decision as the premises holding it up. Three go on the ledger: management will stay post-close, regulatory approval is a formality, the segment grows ~8% a year. Each row carries an owner — the deal partner owns retention, the general counsel owns the regulatory read, the lead analyst owns the growth estimate — and a warrant grade carried over from the audit. Crucially, each row also records what rests on it: the retention premise underwrites the price paid; the regulatory premise underwrites the deal timeline and the bridge financing.

Eight months on, a competitor's blocked deal signals the regulator is tightening. Because the register backlinks the "regulatory is a formality" premise to the timeline and the financing, the moment its owner downgrades that premise, the two decisions that depended on it light up as needing to be reopened — and it is clear whose call each is. Without the register, the weakened premise would have quietly stayed in the deal's foundation, its failure discovered only when the financing came due.

How it works¶

Open a row per premise — record the premise as a flat claim, its current warrant grade (imported from the audit), and the date and source of that grade.
Assign an owner — name the single person accountable for watching and revisiting each premise; an unowned premise is one no one will notice going stale.
Backlink the dependents — record which decisions, commitments, or numbers rest on the premise, so its revision has a traceable blast radius.
Log revisions, don't overwrite — when a premise's warrant changes, append the change and flag the linked decisions for review; the history is the point.

The distinguishing discipline is persistence and traceability: the register's value is entirely in outlasting the meeting and preserving the premise→decision links, not in analysis. It records; it does not adjudicate.

Tuning parameters¶

Inclusion bar — how material a premise must be to earn a row. Too low and the register bloats into an unmaintainable list no owner reads; too high and a quietly critical premise is never tracked.
Ownership granularity — one owner per premise versus a shared or role-based owner. Single named owners create real accountability but can bottleneck; shared ownership spreads load but diffuses it.
Review cadence — how often owners must re-attest their premises. Frequent review catches drift early but taxes attention; rare review lets premises quietly expire.
Backlink depth — whether you record only the immediate decision on a premise or the full chain downstream. Deeper links make revision consequences visible but are heavier to maintain.

When it helps, and when it misleads¶

Its strength is turning "we assumed that at the time" into an addressable, owned record with a traceable set of dependents — so a premise can be revised without re-litigating the whole decision, and a revision can't quietly orphan the choices built on it. It is what makes a decision's foundations legible months after everyone has moved on.

It misleads when it is mistaken for judgment. A register can log a premise as "warranted" and thereby lend it a false solidity it never earned — the record is only as good as the audit feeding it, and a tidy row invites treating a stale grade as a live one. It also decays silently: an unread register with absentee owners is worse than none, because it advertises control that isn't there. And it is easily run backwards — populated after the fact to make a decision look more considered than it was. The discipline that keeps it honest is live ownership with a real review cadence, and a bright line between what the register records (owner, dependents, history) and what it must not pretend to do (decide whether a premise is true).^[1]

How it implements the components¶

Decision Premise Register realizes the record-and-accountability components — the durable, standing side of the archetype:

assumption_owner — it assigns each logged premise a single named owner accountable for watching and revisiting it, so no premise is left ownerless and therefore unmonitored.
consequence_record — it holds the persistent, append-only record linking each premise to the decisions that depend on it and logging every revision, so a premise's change has a findable blast radius.

It does not test warrant or triage which premises matter (evidence_test, materiality_threshold, alternative_assumption) — that is the Assumption Audit, whose graded findings this register imports; and it does not surface premises from a message (embedded_premise) — that is Literal Content Parsing.

Instantiates: Premise–Action Decoupling — the register is the standing memory that keeps a decision's premises revisable instead of silently inherited.
Consumes: Assumption Audit supplies the surfaced premises and their warrant grades that populate the ledger.
Sibling mechanisms: Assumption Audit · Burden-of-Proof Rule · Literal Content Parsing · Context-Condition Testing · Consent Confirmation Form

Notes¶

The register deliberately holds two fields it does not compute: the warrant grade (owned by the audit) and the premise text (owned by parsing). Its integrity depends on that division — the moment a register starts deciding a premise is settled rather than recording someone's dated judgment that it is, it stops being a memory and becomes a laundering step, the very thing the archetype exists to prevent.

References¶

[1] The assumption log (or assumption and constraint register) is a standard project-management artifact — a maintained list of the assumptions a plan depends on, each with an owner and status. Its discipline is exactly this one: an assumption you cannot see, own, or trace is an assumption you cannot revise safely. ↩