[[home|← Ultimentality Wiki]]


# Causal Falsifiability

## Definition

A falsifier counts only when it can *do* something to the target — change, inhibit, delete, demote, or recode the claim or process it attacks. **Causal falsifiability** is the requirement that a falsification make causal contact with what it falsifies. Storage, citation, acknowledgment, or display of an objection does not satisfy it. A criticism that is recorded, indexed, summarized, and even praised, while the governed claim or output remains exactly what it was, has not falsified anything. It has been archived.

The requirement resolves the abstract predicate "could be shown false" into a demand on the update path. It distinguishes what a landed falsifier is permitted to do:

```text
wording change      — the formulation is rewritten
mark change         — the epistemic tier or status is revised (demotion, retyping)
dependency change   — what the claim rests on, or what rests on it, is rewired
behavior change     — the process the claim governs now runs differently
action inhibition   — a governed action is blocked or delayed
deletion            — the claim dies
```

Two conditions certify that contact was made rather than performed. First, a **trace** from the falsifying evidence to the changed state: an inspectable path showing that *this* objection produced *that* revision. Second, a **regression test**: a re-run of the original failure that no longer passes, so the change is protected against silent reversion. Falsification without the trace is assertion; the trace without the regression test is a change that can quietly disappear.

## Type and formal status

**E: Derived, CV.** This is a carving of the falsification concept, not a frame-forced consequence. It refines "falsifiable" into "causally falsifiable" and enumerates the change-types a landed falsifier may produce. Both the refinement and the six-fold list are contestable by a better carving or by a counter-instance — a legitimate falsification that owes nothing to causal contact, or a change-type the list cannot hold. Because it fixes an operational test the wiki's own governance runs on itself, it stands near `[[falsification-standard|The Falsification Standard]]` and `[[answerability-predicate|The Answerability Predicate]]`, but it adds no authority those pages do not already carry.

**A: mapping-accuracy aspiration.** The claim intends to map how governed systems actually resist or absorb criticism — the observable gap between a system that can be moved by an objection and one that cannot. It is inaccurate, not merely disallowed, wherever a system with a permanently closed update path is nonetheless the kind of thing the framework should be willing to call answerable.

This page is a **treatise-side extension, held contestable.** It is not canonical, and the causal-effect requirement is itself a marked claim, reopenable like any other content on the wiki.

## What it regulates

Causal falsifiability regulates the characteristic excess of *displayed* correction: the drift by which a system accumulates tests, kill conditions, logged objections, and confidence scores while its governing outputs stay causally immune. It supplies the operational teeth behind `[[effective-and-ornamental-answerability|Effective and Ornamental Answerability]]` — the criterion that separates recording from routing-and-effect. Applied to `[[falsification-standard|The Falsification Standard]]`, it forbids "we state our tests" from passing as "our tests can move us." Applied at the claim level it inherits `[[causal-error-mark|The Error Mark With Causal Force]]`: an error signal eₚ earns the name only where it makes some Δ(p) reachable. Where those pages name the requirement, this page names the *test* — evidence, trace, changed state, regression — and thereby regulates the gap between a criticism's existence and its force.

## What regulates it

Left alone, the causal demand overshoots into a demand for perpetual visible change: every objection must delete or rewrite something, or the system stands convicted of ornament. Three regulators hold it back.

`[[outcomes-of-attack|The Outcomes of Attack]]` supplies the distinction between *possible* and *mandatory* change. Causal falsifiability requires a reachable update path, not a forced update; a failed attack is logged as a failed attack, never converted into a change and never counted as confirmation. `[[same-level-attack-rule|The Same-Level Attack Rule]]` gates what is even owed a causal effect: a falsifier aimed at the wrong type or level has earned no update, so its inertness convicts nothing. `[[scar-record|The Scar Record]]` disciplines the change itself — the trace and regression test must survive as an inspectable constraint, so that "we changed it" cannot be discharged by cosmetic churn. Above all, recursive marking regulates the requirement's own status: the causal-effect rule is Derived and reopenable, not the one exempt meta-standard.

## Valid attack surface

The strongest attack targets the carving where it is most exposed: **causal effect is neither necessary nor sufficient for falsification.** Not necessary — a proof of internal contradiction refutes a claim under its own conditions whether or not any system chooses to update, so falsification-as-truth appears to float free of falsification-as-effect. Not sufficient — a system can change its wording, marks, and dependencies in response to an objection and still be tracking nothing about the world, so causal contact does not certify that the falsifier was *right*. A valid attack presses either edge: exhibit a complete, legitimate falsification whose force is independent of contact with the target, or exhibit causal effect standing in for accuracy.

The framework's answer is a separation, not a denial: causal falsifiability governs whether a falsifier can *count within a governed system's answerability*, which is a distinct question from the claim's truth-value on the `[[alethic-axis|Alethic Axis]]`. A counter-instance that collapses those two questions, or a change-type the six-fold list cannot represent, is a live CV attack on this page.

## What happens if isolated

Isolated on the causal side — the requirement without the preserved-marking side and without the outcomes classification — it degrades into **churn-theater.** A system reads "falsifiers must produce change" as "produce change to look falsifiable," and manufactures cosmetic edits: reworded-to-nothing formulations, delete-and-restore cycles, a proud tally of pages touched per revision. The trace is fabricated after the fact; the regression test guards nothing. The metric is gamed precisely because it has become a metric.

Isolated on the other side — drop the causal requirement entirely — and falsification collapses back into the very ornament this page exists to prevent: kill conditions that are stated but unreachable, objections filed where the governing layer cannot see them. Each isolation is a distinct, observable failure: churn without preservation, or preservation without effect.

## What larger property emerges from the coupling

Coupled with preserved marking, the scar record, and the outcomes classification, causal falsifiability yields `[[postfalsifiability|Postfalsifiability]]`: a system whose current invariants are the compressed residue of falsifiers that actually landed, each surviving form still carrying a live, reachable kill condition. The coupling makes *observable* a difference that is otherwise mere assertion — the difference between "survived a real attack" and "was never really attackable." Killed forms leave `[[negative-information|Negative Information]]`; surviving forms remain reopenable; and the record of what changed, and what refused to change under valid pressure, is what `[[vls-as-desire|VLS as Desire]]` accumulates into across repeated cycles of attempted destruction and correction. Neither term alone produces this: causal effect without marking is untracked volatility, marking without causal effect is ornamental humility, and only their coupling is what `[[crystallization|Crystallization]]` can select on.

## What would actually kill the claim

By its own governing test, the claim dies in either of two ways. First, if the framework's own logged falsifiers provably leave every relevant output and governance decision unchanged — then the standard is ornamental *by its own criterion*, and the page fails under `[[self-application|Self-Application]]`. Second, if the demanded evidence-to-changed-state trace cannot in principle be reproduced or inspected — then "causal" adds no operational content over "acknowledged," and the distinction it lives on has quietly evaporated.

A third, sharper kill retires the carving rather than convicting the system: exhibit the *relevant* kind of falsification the framework needs — one that governs answerability, not merely abstract truth — whose force is genuinely independent of any causal contact with the target and cannot even use causal contact as a proxy. Then causal falsifiability is not strengthened by the failed challenge; it is replaced by a better carving.

## Prohibited misreadings

- **"Every objection must cause a change."** The requirement is a reachable update path, not a mandatory update. A failed attack stays logged as a failed attack; it is never forced into a cosmetic change and never `[[outcomes-of-attack|counted as confirmation]]`.
- **"Causal means physical-mechanical."** Wording, mark, and dependency changes are causal changes; the falsifier need not be a material event. This is not a claim that only physical happenings can falsify.
- **"Recording the criticism is enough."** That is exactly what the page denies: storage, citation, and display without routing-and-effect are ornamental — the failure diagnosed by `[[effective-and-ornamental-answerability|Effective and Ornamental Answerability]]`.
- **"Causal effect proves the falsifier was right."** A change in system state is not proof the change tracks the world; effective and accurate are separate axes — see the `[[two-axis-ledger|Two-Axis Ledger]]`.
- **"This is the one standard that cannot itself be attacked."** Reading the causal-effect requirement as a forced, exempt meta-rule is the `[[textual-nephilim|Textual Nephilim]]`. It is Derived, CV, and reopenable; its own kill condition is stated above, and its status must survive the `[[self-sealing-test|Self-Sealing Test]]`.

## See also

[[causal-error-mark|The Error Mark With Causal Force]] · [[effective-and-ornamental-answerability|Effective and Ornamental Answerability]] · [[outcomes-of-attack|The Outcomes of Attack]] · [[postfalsifiability|Postfalsifiability]] · [[falsification-standard|The Falsification Standard]] · [[self-sealing-test|The Self-Sealing Test]] · [[scar-record|The Scar Record]] · [[same-level-attack-rule|The Same-Level Attack Rule]] · [[answerability-predicate|The Answerability Predicate]]