The Good Is Not Its Measure

A relation gets re-specified as a quantity that proxies it; an optimization frame installs itself and answers the legitimacy question in its own terms before the question can be asked. The two instance-notes track this at the treasury and at the self. This draft asks whether the move is one move — and if so, what only becomes visible from above.

1. The move, stripped of locus§

Some goods are relations: non-rivalrous, not held as a stock, not maximizable as a number. Security is one. Coordination is one. Freedom is one.

Alongside each relation there is usually a quantity that proxies it — a measure that tracks the relation well enough, most of the time, to pass for it. Balance tracks treasury security. Wealth tracks an agent’s latitude. This correlation is exactly what makes the substitution seductive rather than obviously wrong: the proxy is not a lie, it is a usually-reliable measure — right up until it is optimized as an end, at which point it detaches from the relation it was standing in for.

Quantity substitution is the silent re-specification of the good from the relation to the proxy. Its operation is reification — a relation treated as a possessable stock — and its signature is that, once made, an optimization frame is entailed rather than chosen. Within that frame, maximizing the proxy simply is the work, and the question of whether the relation is still served becomes unaskable, because the frame has already answered it in the proxy’s terms. This is the foreclosure both instance-notes describe. It is the default output of competence, not of bad faith.

Two established neighbors sit downstream of this move, and locating them precisely sharpens the claim rather than threatening it. Goodhart’s law is the epistemic symptom: optimize the proxy and its correlation with the target collapses — the measure stops measuring. Campbell’s law is the causal-harm mechanism: use the proxy for high-stakes decisions and the underlying process it tracked is actively corrupted, not merely detached. Quantity substitution is the upstream ontological act neither isolates — the relation reified into the proxy in the first place, of which Goodhart’s detachment and Campbell’s corruption are the consequences. Their independently-established generality across testing, policing, and monetary policy is outside evidence that the substrate is domain-general; the neighbors corroborate the invariance rather than compete with it. Borrow their dynamics; keep the reification as the act and the derivation structural.

There is a further difference, and it is why the move is worse than either law alone. Goodhart and Campbell both presuppose you still hold the target as the good — which is what lets you, in principle, notice the degradation and correct. Quantity substitution adopts the proxy as the good, and in doing so severs the corrective signal: the relation is no longer the thing being sensed, so its detachment (Goodhart) and its corruption (Campbell) proceed unregistered. The move is a sensor-swap — it blinds the system to its own most important variable, which fails invisibly and is degraded at once. This is precisely the failure the distal-sensing and field-fitness work exists to catch; quantity substitution names its ontological cause.

2. What is invariant, and what varies§

The value of a parent is a schema: to instantiate the move, bind these slots. The invariant is the top row; the two known instances fill the rest.

The first four rows generalize cleanly. The last two do not — and that is the finding. (The table shows the two end-positions; a third — the externality-imposer — sits between them and is treated in §3.)

3. The asymmetry: scale-free diagnosis, scale-dependent correction§

Read down the last two columns and the parent earns its keep. The diagnosis is identical across loci: same reification, same entailed frame, same foreclosure, same disclosing structure. But the ought that converts diagnosis into correction is supplied per locus, and the two known loci supply different oughts — with opposite consequences for what may be done about it.

At the institution, the good is held in trust, and an institution owes fidelity to the purpose under which it was constituted. That ought licenses a correction with teeth: a sorting criterion, a burden shift onto the party proposing the substitution, and — where a protection guards an irremediable failure — a wall.

At the agent, the same move meets the same diagnosis, but self-determination forbids the corresponding remedy. The right to exit is the right to sell is the right to make even the self-undermining trade; wall it and the framework becomes the thing it opposes. So the correction is capped at legibility: the trade made clear enough that whoever chooses it is choosing it.

The being-informed → engagement test (§6) surfaced a third position the original two instances happened to bracket. A platform optimizing engagement looks institution-like, but fidelity-to-constituted-purpose does not ground the correction: informedness may be a good the platform never took as its charge, so it is not betraying a purpose it was constituted to hold. The ought that does the work there is neither trust nor autonomy but externality — the substitution imposes a cost on parties outside the transaction (the polity’s epistemic health) — and that gates a third kind of remedy again: accountability and transparency obligations, neither the treasury-style wall nor the agent-style legibility-only. The gate is therefore not binary. It is keyed to the structural position of the substituting party — trustee of a constituted purpose, autonomous self, or imposer of an externality — and each position supplies a different ought and licenses a different remedy. The original two instances sat at the ends; the test revealed the middle.

The parent’s one non-obvious claim is therefore this: the recognizability of quantity substitution does not travel with a license to correct it. The diagnostic generalizes; the remedy is gated by the structural position of the substituting party and the rights or duties that position carries. Anywhere this move is found, the diagnosis transfers for free — and the question of what may be done about it must be re-derived from the position and the rights in force there.

4. The seam§

Because of §3, this parent does not own a single is/ought crossing. It owns a slot where the crossing goes, filled per instance.

Everything in §§1–2 is descriptive: what relations are, how proxies track them, what reification does, what the frame forecloses. That layer is shared and carries no imported ought. The crossing into correction happens only when a locus supplies its governing ought — and the two instances cross on different commitments (fidelity-to-purpose; self-determination-as-limit). Marking it this way keeps the parent honest: it does not smuggle a universal norm under the cover of a universal diagnosis. The diagnosis is universal. The norm is local, and named where it is used.

5. Held tension§

Inherited from both children and not resolved here: the proxy is a legitimate instrument. Balance genuinely serves security; wealth genuinely serves latitude. The move becomes a substitution only when the proxy is optimized as the terminal good and detaches from the relation. The line between “quantity serving the relation” and “quantity as end” is the same margin both instances leave open, and the parent generalizes the margin rather than closing it. Collapsing it — declaring quantities suspect as such — would be the same foreclosure running from the opposite direction.

6. Promotion conditions and test log§

Hub status is earned, not presumed, so the third-locus test was run rather than assumed — against two loci not chosen to flatter it.

Being-informed → engagement. The relation is informed attention; the proxy is engagement (views, watch-time, shares); the frame is engagement-maximization; the disclosing moment is the viral falsehood, where the metric peaks exactly where the relation is worst. Passes §§1–3, and forced the three-position refinement now recorded there — a platform’s remedy is gated by externality, not by trust or autonomy.

Governance participation → turnout. The relation is meaningful, consequential inclusion; the proxy is turnout, activated stake, DRep headcount; the frame is turnout-maximization. Passes §§1–3 — and did so before this note existed: the endorsed voting rationale records the Catalyst finding that incentivizing reviewers produced less net deliberative capacity, as non-deliberative agents flooded the space to collect the incentive. That is a Campbell instance sitting in the project’s own source material, and the rationale already prescribes the trustee-position remedy the parent predicts — no incentives until it is assured they will not cause a net capacity loss, a burden-shift onto the party proposing the proxy.

Both loci pass, with corroboration from two directions the original instances could not supply: external (Goodhart’s and Campbell’s established cross-domain generality) and internal (a pre-existing instance and its matching remedy already written into the rationale). On the stated condition, this warrants moving toward earned hub status.

Two honesty-notes are logged rather than smoothed over. First, the platform sub-locus strained the fidelity-to-purpose ought and is what forced the three-position gate; the schema held only after being refined, which is a real edit, not a clean pass. Second, every instance so far is a proxy-for-a-relation in a coordination-adjacent domain; a genuinely distant locus — biological, or purely private and non-social — is the harder test, and would be required before claiming the move beyond commons contexts.

Status is left at CANDIDATE: the test supports promotion, but the promotion decision itself is deferred to the author rather than taken here.