ADR-058: A THEORY can corroborate a CLAIM: evidence the record links to its own claim without the claim resting on it is corroborated_by, not grounds

Context

ADR-044 added corroborates for the record’s own claims: one claim gives independent evidence for another “without that claim depending on it”, so retiring it costs the other support and not its footing. The relation runs from CLAIM to CLAIM. Readings, which are THEORYs, could reach a claim only through grounds, the claim’s evidence from outside the record, which the claim stands on.

The status review of 2026-10-27 (ADR-057) listed 38 edges where an Active CLAIM or ARG grounds on, or rests on, a Proposed entry. Three readers reviewed 27 of the dependents. In most of them the Proposed reading was cited as follows:

• As later evidence. CLAIM-079’s note of 2026-10-12, “evidence now linked”, added THEORY-142’s stylized simulation.
• As an instance. CLAIM-164’s note says of THEORY-078 and THEORY-084: “This claim does not need them: its core is Davis–Kahan.”
• For an attribution. CLAIM-434 says: “the finite case of what Monderer and Shapley call … (THEORY-896), proved here because their paper is unread.”

In each case the claim did not need the reading. Several of the claims’ own directives said “nothing here rests on it”. grounds said otherwise.

Decision

• corroborated_by on a CLAIM may name a THEORY, and a THEORY has corroborates, its converse. It means what ADR-044’s relation means: independent evidence, and no dependence.
• grounds keeps its meaning, a reading the claim stands on. A claim whose argument uses a reading’s contested thesis keeps it in grounds. If the reading is Proposed and the claim Active, that is an inversion, and a status review decides it.
• Attribution and quotation take no edge. Where a reading is cited only to say who said something, or to quote a definition, the body’s citation is the link. A definition the claim depends on is better carried by a TERM, or by an Active THEORY split off to state it, as THEORY-1096 does for Blackwell’s definitions.

Alternatives considered

• Drop the edges. This removes the inversion and also the record of which evidence was found for the claim. That record is what “evidence now linked” notes were made to keep.
• A general cites relation. A body citation already records that a claim cites a document, and the rendered page links it. A field would be a second copy (DP-3).
• Let complements take a THEORY. complements relates two claims that sit beside each other in an argument, not evidence for one of them.

Consequences

• An Active claim corroborated by a Proposed reading is not an inversion, so the worklist’s count falls to what is really unresolved.
• Rejecting a THEORY that corroborates a claim leaves the claim’s status alone, and a reviewer may still want to look at the claim. The THEORY’s page lists what it corroborates.
• The review that adopted this decision moved the edges it found. Other grounds edges added as evidence remain, and a status review that meets one moves it.