Evidence-to-Claim Matrix¶

Evidence-warranting tool — instantiates Context-Bounded Meaning Recovery

Lays each interpretive claim beside the marks that support, conflict with, or fail to appear for it, so a reading's narrative force can be told apart from its evidentiary warrant.

A persuasive reading and a well-warranted reading are not the same thing, and the gap between them is where projection hides. Evidence-to-Claim Matrix inverts the usual flow: instead of narrating a reading and letting its momentum carry the audience, it makes each claim a row and audits it against the substrate, sorting the evidence for that claim into supporting, conflicting, absent, and ambiguous. Its defining move is that the empty and conflicting cells are the payload — a claim whose row is all "absent" is carrying story force with no warrant behind it. Where a close reading records the marks and an alternative-readings review compares whole readings, this tool does one narrow, unglamorous thing: it separates what a claim asserts from what the marks actually support.

Example¶

An editor is reviewing an investigative feature arguing that a manufacturer knowingly shipped a defective part. The draft is gripping, and that is exactly the problem — nobody can yet say which sentences are load-bearing. The editor builds the matrix: each consequential claim becomes a row ("engineers warned management," "the warning was documented," "management overruled it"), and each is scored against the reporting. Most rows fill out with supporting and corroborating marks. But the emotional climax — management overruled it — has one ambiguous source and otherwise empty support and conflict columns.

The output is not a verdict on the story but a map of where its warrant is thin: "the strongest claim rests on the weakest row." That changes the edit. The climax is softened to what the evidence admits, a reporter is sent back for the missing corroboration, and the piece survives legal review not because it lost its force but because its force now sits on marks that exist.

How it works¶

The distinguishing move is organizing by claim rather than by source, then treating absence and conflict as first-class findings. Each claim is pinned to the specific marks that bear on it; each link is typed as support, conflict, absence, or ambiguity. Reading down a column exposes cherry-picking (a wall of support with no one having looked for conflict); reading across a row exposes a claim that persuades on narrative alone. What survives is a reading whose rhetorical weight and evidentiary weight have been forced apart and re-matched.

Tuning parameters¶

Claim granularity — whether a whole thesis is one row or is split into its sub-claims. Finer splitting catches buried leaps but bloats the matrix and can bury the load-bearing claim among trivia.
Evidence-status taxonomy — how many cell types you distinguish; adding "independently corroborated" or "single-sourced" sharpens the audit at the cost of speed.
Warrant threshold — how much support a claim must show before it is allowed to stand as asserted rather than hedged.
Evidence weighting — treating every mark equally versus weighting by source independence and quality, which changes which rows look thin.
Audit scope — every claim, or only the load-bearing ones the reading cannot survive without.

When it helps, and when it misleads¶

Its strength is catching the claim that convinces without support, and making cherry-picking visible by demanding the conflict and absence columns be filled as diligently as the support column. It converts "this reading feels solid" into a checkable ledger.

Its classic misuse is being run backwards: filled in after a conclusion is fixed, with only the supporting column populated, so the matrix decorates a foregone reading instead of testing it — the narrative fallacy given a spreadsheet.^[1] It also quietly over-weights whatever is easy to evidence and under-counts what resists documentation. The discipline that keeps it honest is to populate the conflict and absent cells first and hardest, and to treat a tidy support column as a prompt to go looking for what would break the claim.

How it implements the components¶

evidence_mark_trace — each claim is tied to the specific marks that bear on it; the matrix is that trace, made explicit and inspectable.
coherence_and_tension_test — the supporting-versus-conflicting columns test whether the marks cohere with a claim or pull against it, surfacing the tensions a smooth narrative would paper over.

It does not record the substrate's marks in the first place (that is Close Reading Table) or enumerate whole rival readings (that is Alternative Readings Review); it audits claims against marks that other mechanisms surface.

Instantiates: Context-Bounded Meaning Recovery — supplies the warrant check that keeps a reading bound to its evidence.
Consumes: Close Reading Table — the located marks it audits claims against.
Sibling mechanisms: Close Reading Table · Alternative Readings Review · Context Reconstruction Worksheet · Framework Selection Memo · Glossary or Code Key · Hermeneutic Loop Log · Ambiguity Register · Interpretation Brief · Interpretation Red Team · Pragmatic Force Walkthrough · Precedent Comparison Table · Stakeholder Meaning Check · Translation Backcheck

Notes¶

The matrix warrants individual claims; it says nothing about whether a better whole reading would fit the same marks. A reading can pass every row and still be the wrong frame — that judgment belongs to Alternative Readings Review. Keeping the two separate is what lets a team strengthen a claim's evidence without pretending it has also ruled out the rivals.

References¶

[1] The tendency to accept a coherent, compelling story as if the coherence were itself evidence — the narrative fallacy, a term popularized by Nassim Nicholas Taleb. Demanding the conflict and absence columns is the standard corrective, which is why they carry the weight here. ↩