Home / The Outcomes of Attack

The Outcomes of Attack

Definition

When a valid attack lands on a content-bearing claim, the framework recognizes exactly four things that can happen to the claim — and one thing that is not among them. The four standardized outcomes are:

Deletion. The claim dies. It is removed from the corpus and its dependents are rerouted or fall with it.
Revision. The claim changes. Its wording, scope, dependency, or formalization is rewritten, and the earlier form is retained as the reason the new one exists.
Demotion or retyping. The claim survives at a different level or status. Its mark moves (for instance FT → CV), its abstraction level shifts, or its tier changes, without the underlying relation being destroyed.
Invariant recognition. The attack — run as a genuine removal test — shows that deleting the relation destroys the function it regulates. The claim survives as load-bearing, and what is recorded is the function that would be lost, not a verdict of proof.

The excluded fifth is confirmation. A valid attack that does not land is logged as a failed attack — a record type, never an outcome, and never evidence that the claim it failed against is therefore true. Counting a survived attack as support is the drift toward a forced tier — the Textual Nephilim — and is prohibited across the frame.

Every resolution — whichever of the four it reaches — is recorded as a fixed object, not a narrative:

resolution:
  target:                   <page or claim under attack>
  attack_type:              <frame declension | counter-instance | removal test | ...>
  evidence:                 <what was shown, causally traced>
  outcome:                  deletion | revision | demotion/retyping | invariant-recognition
  changed_content:          <what text or relation changed, or "none">
  changed_marks:            <E/A mark transitions, or "none">
  changed_links:            <edges added, removed, or rerouted, or "none">
  remaining_kill_condition: <what would still defeat the surviving claim>

A resolution missing these fields is not a resolution; it is an anecdote. The record is the operative object, and it is precisely what Crystallization and the scar record later read.

Outcome	The claim	What moves	What the record fixes
Deletion	dies	it leaves the corpus; dependents rerouted or dropped	evidence, the removed content, downstream link repairs
Revision	changes	wording, scope, dependency, or formalization	old form, new form, why the old failed, the surviving kill condition
Demotion / retyping	survives at a new level or status	its E/A mark, tier, or abstraction level	old mark, new mark, the counter-instance that forced the move
Invariant recognition	survives as load-bearing	nothing in the claim; the understanding of it	the attempted removal, the function that would be lost, the still-live kill condition

Type and formal status

E (epistemic): Derived, CV. This is a classification — a carving of the outcome-space — and by the invariant that every count is a contestable carving, the count of four is contestable by a counter-instance (a genuine fifth outcome) or by a better carving that merges or re-cuts the set. It is not FT: the partition does not follow from the definition of attack alone; it is a proposed exhaustive-and-disjoint cover of what a landed attack does, and a cover can be wrong.

A (alethic): the classification aspires to map the actual behavior of a claim under a valid attack. Its accuracy claim is that these four exhaust the ways a claim is moved and that "confirmation" names no motion at all. Exhibiting a real movement outside the four would make the carving inaccurate, not merely unstylish — the axes do not predict each other, so contestable here never means probably wrong.

This page is a treatise-side extension, held contestable. It fixes no authority and licenses no exempt or forced status for anything, including itself.

What it regulates

This page supplies the outcome vocabulary that Postfalsifiability presupposes: a postfalsifiable corpus is one whose invariants are the residue of attacks that were killed, revised, retyped, or recognized as invariants — and that residue is only legible if each of those verbs is defined and recorded. Without a fixed outcome set, "the claim survived falsification" degrades into an unaudited boast.

It regulates one characteristic excess directly: confirmation creep — the drift by which a survived attack is silently reclassified as support and, from there, the claim is nudged toward exemption. The Falsification Standard forbids this in principle; this page operationalizes the forbidding by giving survival no cell labeled confirmed.

It regulates the failed-attack / invariant-recognition boundary, the very seam its own kill condition targets. Invariant recognition is a positive result — a removal test that produced a specified functional loss. A failed attack is an absence — an attack that never reached the seam. Collapsing the second into the first is how "you couldn't defeat it" becomes "it is necessary," which is the Textual Nephilim.

Finally it regulates attack bookkeeping: the eight recorded fields force every resolution to name its target, type, evidence, and remaining kill condition, so that no attack can dissolve into a story about who prevailed.

What regulates it

Causal Falsifiability is the primary regulator. An outcome counts only when its motion is a traceable state-change: a deletion that removes nothing, a revision that alters no wording, mark, dependency, or behavior, is not an outcome but a display. Causal falsifiability supplies the requirement that each of the four names a real change; this classification supplies the names. Neither is sufficient alone.

The Same-Level Attack Rule and Attack-Type Matching sit upstream. They decide whether an attack is valid at all — whether it contests the claim at the type and abstraction level where the claim is made. An ill-typed attack has no outcome on this page; it is rejected before classification, not slotted into a default cell.

The Two-Mark System regulates the record's changed_marks field: demotion and retyping are motions within the E/A ledger, and this page cannot invent a mark the ledger does not carry.

And the CV mark on its own four-count regulates the page from inside. The classification is contestable by the same moves it classifies; a better carving would revise it, and a demonstrated fifth outcome would enlarge it.

Valid attack surface

A valid attack on this page must do one of the following, at the level of the classification itself:

Exhibit a genuine outcome of a landed valid attack that is irreducible to deletion, revision, demotion/retyping, or invariant recognition — a fifth motion the four cannot represent. (A candidate must be a motion of the claim, not a fact about the attacker or the audience.)
Show that two of the four collapse — for instance, that demotion is always a species of revision, or that invariant recognition is only deletion-of-the-attack under another name — so the carving is redundant.
Show that the classification cannot in practice separate a failed attack from invariant recognition, i.e. that any survival can be filed under either cell at will.

What is not a valid attack: objecting that the count "four" is arbitrary (all counts are CV; arbitrariness of a number is not a counter-instance), or offering a failed attack as though its failure were an argument against the page. A failed attack against this page is logged, per the page's own rule, as a failed attack.

What happens if isolated

Isolated from causal falsifiability, the four labels become bookkeeping theater: resolutions are written, outcomes are stamped, and nothing in the governed corpus moves. "Deletion" and "revision" without a traced state-change are ornamental — the failure the causal error mark names, transposed to the ledger.

Isolated from the failed-attack guard, the classification self-seals. With no cell for failed and a permissive reading of invariant recognition, every survived attack finds a home under "the removal test showed the relation is necessary," and from there the claim is quietly promoted toward the exempt tier. Classification without the guard is a confirmation engine wearing a falsification costume — exactly what the Self-Sealing Test exists to catch.

Isolated from the upstream validity rules, outcomes get assigned to attacks that never earned an outcome: a process offered against an operator, a substrate difference offered against a function, is mis-typed but still "resolved," and the ledger fills with resolutions of non-attacks.

What larger property emerges from the coupling

Coupled with causal falsifiability and read into the postfalsifiability ledger, this classification yields an auditable falsification record: a corpus whose surviving invariants are the compressed residue of logged, causally effective defeats, each survival recorded honestly and none mislabeled as proof.

The coupling is specific. The classification alone gives outcome names but cannot guarantee they correspond to anything. Causal falsifiability alone demands a real state-change but does not say which states count as which motion. Wired together — each outcome name bound to a required traced change — they produce a ledger that Crystallization can consume: high structural compression achieved because unstable arrangements were eliminated and the eliminations are on the record, distinguishing a crystal (still answerable) from a fossil (a dead form merely preserved). Semantic natural selection operates over exactly this record; without honest outcome-accounting it would mistake mere persistence for survival-by-selection.

What would actually kill the claim

The claim dies if the system cannot distinguish a failed attack from invariant recognition — if there is no operable criterion by which "the attack didn't connect" is separated from "a removal test demonstrated a specific functional loss," so that any survival can be booked as either. Then the classification collapses into the confirmation engine it was built to prevent.

It also dies if a critic exhibits a real fifth outcome of a landed valid attack, irreducible to the four, or shows that two of the four are never distinct in practice. Either result revises or re-carves the page and is logged, on this very page, as a revision or retyping — with a new kill condition attached.

And it is undercut if the eight record fields can be filled without any causally traceable change — if a complete resolution can be written for a motion that never occurred. That routes the failure to Causal Falsifiability and to the Kill-Table, where the corresponding row already stands: postfalsifiability treats all failures as support.

Prohibited misreadings

Invariant recognition read as confirmation, proof, or a forced tier. Recognizing that removing a relation destroys a regulated function does not lift the claim out of contestability. The claim stays Derived, keeps its E/A marks, and keeps a live kill condition. Reading load-bearingness as forcedness is the Textual Nephilim and the local face of absolutization.
A failed attack read as strengthening the claim. Failure to defeat is not evidence for. A failed attack is logged as a failed attack and contributes nothing to the truth of what it failed against. This is the framework-wide rule that a failed attack is never counted as confirmation.
The four outcomes read as a fixed, exhaustive, exempt set. The count is CV, like every count. It invites a better carving; it does not close.
Demotion or retyping read as deletion — or as promotion. Survival at another level or status is neither death nor victory. It is a recorded move within the two-mark ledger, with the counter-instance that forced it on the record.
Classification read as guaranteeing an outcome. Validity is decided upstream by the Same-Level Attack Rule and Attack-Type Matching. An invalid attack yields no outcome, not a default one.
The resolution record read as optional narrative. The eight fields are the operative object of the resolution, not a summary of it. Without them there is no logged outcome, only a claim about who won — which the framework does not accept.