The Instruments

Citation Latticeமேற்கோள் வலை

Promotion requires an independent path to a higher-tier source, not additional citations at the current tier. Formally: tier(c) = max over independent source paths, not a count.

Measures

The evidentiary state of every load-bearing claim in the archive, expressed as a discrete tier derived from the status of the source institution.

Cannot measure

Whether a claim is true. The lattice measures the standing of the body asserting it, which is a proxy and is stated as one.

Failure mode

Tier inflation. A widely repeated claim accumulates citations from sources that are all downstream of one original, and the volume of references is mistaken for independence.

Falsified by

A sampled audit finds that claims at the highest tier are corrected at a rate indistinguishable from the tier below. If tiers do not separate error rates, they encode nothing.

Opinion Spaceகருத்து வெளி

PCA to two components over the response matrix; k-means over the projection; Laplace noise added to all published counts at ε ≤ 1.0; no cohort published below a floor of 25 participants.

Measures

The structure of agreement across a self-selected participant set: how many distinct positions exist, and which statements are shared across them.

Cannot measure

Anything about the population. Participation is self-selected and unweighted, so the output describes the participants and nobody else. It is not a poll, not a referendum, not a census, and not a mandate.

Failure mode

Brigading. A coordinated group enters at volume, forms an artificial cluster, and the map reports a manufactured position as a natural one.

Falsified by

Two independent participant sets drawn under the same conditions produce cluster structures that do not correspond. If the geometry is not reproducible, it is an artefact of who arrived.

Return Index

A ratio over two published series, reported with the series named, the reporting period, and the definitional caveats of both.

Measures

The observed relationship, in published data, between educational output in a region and the retention of that output within the region.

Cannot measure

Cause. It reports a gap between qualification and local absorption; it does not identify why, and it cannot distinguish between opportunity, choice, and coercion.

Failure mode

Denominator drift. Administrative categories change between reporting periods and the index reads a definitional change as a real movement.

Falsified by

The index moves in the opposite direction to the underlying published series when both are recomputed from source. Any divergence from the primary series falsifies the index, not the series.

Refusal Engineமறுப்பு இயந்திரம்

Measures

Whether every operational refusal in the archive is bound to a named external instrument — a statute, a treaty article, a professional standard, or a published safeguarding rule.

Cannot measure

Whether the refusal is wise. It checks that a refusal has an external anchor, not that the anchor was correctly chosen.

Failure mode

Anchor drift. An instrument is amended or superseded and the refusal continues to cite the old version, so the refusal looks anchored while floating free.

Falsified by

An audit finds a refusal in force whose cited instrument does not, on reading, support it. One such finding invalidates the engine's guarantee and requires a full re-audit.

Gate Ladder

Measures

Which capabilities are unlocked, expressed as conditions rather than dates. Each gate is a stated, checkable condition; a capability is available only when its gate is open.

Cannot measure

When a gate will open. The ladder deliberately contains no time dimension.

Failure mode

Gate erosion. A condition is quietly reinterpreted so that a gate reads as open without the underlying condition changing.

Falsified by

A capability is found in use while its gate is, on the published condition, closed. This is a governance failure and is recorded as one in the changelog.

Continuity Orrery

Survival is treated as a function of independent replicas, not of storage durability at any one replica.

Measures

The archive's survival surface: how many independent copies exist, across how many custodians, in how many jurisdictions, and how long since each was verified.

Cannot measure

Whether any copy is faithful. It counts copies and verification events; fidelity is established by hash comparison, which is a separate check.

Failure mode

Correlated custody. Several copies appear independent but share a hosting provider, a jurisdiction, or a single responsible person, so one event removes them together.

Falsified by

A simulated loss of the largest single custodian leaves the record unreadable. If removing one node breaks the archive, the orrery has been reporting redundancy that does not exist.

Decay Clock

Measures

For every cited claim, the elapsed time since its source was last confirmed reachable and unchanged. The archive's own rot, measured against itself.

Cannot measure

Whether an unreachable source was correct when cited. A dead link is a continuity failure, not a truth failure, and the two are reported separately.

Failure mode

Verification theatre. An automated reachability check passes because a URL resolves, while the document behind it has been silently replaced.

Falsified by

A manual sample finds sources marked verified whose content has materially changed. The clock then measures nothing and must be rebuilt around content hashing.