Evidence Relation Matrix¶
Template — instantiates Evidentiary Trace Warranting
Maps traces to claims with provenance, relevance, link type, weight, scope, and defeaters.
An Evidence Relation Matrix is a many-row cross-tabulation in which each row is one trace-to-claim relation and each column is a property of that relation — origin, relevance rule, weight, scope. Its defining trait, and the thing no single-argument artifact can do, is the column-wise operation across rows: it exists to hold many relations side by side and combine them into one auditable, still-decomposable judgment. Where a single card makes one inference legible, the matrix answers the questions that only appear at scale — do these five traces converge or conflict, which claim is best supported once everything is on the table, and does the aggregate rest on independent sources or on one source counted five times? The matrix is the archetype's synthesis surface: it aggregates without laundering, because every row keeps its provenance and scope even after the columns are combined.
Example¶
A due-diligence team is auditing a target company's claim that revenue is recurring and durable. No single document settles it, so the lead builds an evidence relation matrix. Each row is one trace against one sub-claim: the signed contracts (row 1, sub-claim "customers are under multi-year terms"), the billing-system export (row 2, "revenue actually recurs monthly"), management's cohort chart (row 3, "churn is low"), and three reference calls (rows 4-6, "customers intend to renew").
The columns force each row to carry its own reliability. The contracts have strong provenance (executed originals) but narrow scope (they cover 40% of revenue). The cohort chart is highly relevant but weak on provenance — it is management-prepared and unaudited. The reference calls are independent but the sample is self-selected. Read row by row, each looks supportive. Read down the columns, the synthesis interface surfaces the real picture: the durable-revenue claim is well-supported for the contracted 40% and thinly supported for the rest, and two of the three "independent" signals trace back to management's own numbers — so they are not as independent as the row count suggests. That aggregate, not any one row, is what the deal memo can defensibly stand on.
How it works¶
The matrix is a table with one row per trace-claim relation and a fixed column set:
- Trace and Claim identifiers (referenced, not re-authored — the relation content lives on cards).
- Provenance — where the trace came from and how strong that origin is.
- Relevance — the stated rule by which the trace bears on the claim.
- Weight and scope — carried per row and kept attached through aggregation, so a locally strong row is never read as globally strong; link type and defeaters are referenced from the mechanisms that own them.
The distinctive step is the synthesis pass: reading down each column to detect convergence, conflict, double-counting, and coverage gaps, then producing a combined judgment per claim that carries its scope with it. A well-built matrix never collapses to a single number — it aggregates while keeping each row's limits visible, so a reviewer can reopen any cell that drove the conclusion.
Tuning parameters¶
- Granularity — one row per (trace, claim) pair versus per trace. Finer rows expose double-counting but grow the table quickly.
- Column set — how many relation properties are tracked. More columns are more auditable but raise the cost of every added row and tempt teams to leave cells blank.
- Independence bookkeeping — whether the matrix explicitly tags shared sources so correlated evidence is not summed as if independent. Strict tagging prevents inflated confidence but is laborious.
- Aggregation rule — how column-wise reads combine (narrative synthesis, weighted tally, best-evidence). Formal rules are repeatable; narrative ones are flexible but harder to audit.
- Coverage view — whether the matrix also reports which sub-claims have no strong rows, turning silence into a visible gap.
When it helps, and when it misleads¶
Its strength is that it is the only sibling built for aggregation without laundering. A pile of individually-fine traces can still add up to a bad conclusion through double-counting, hidden correlation, or uneven coverage; the matrix makes those cross-row pathologies visible where a stack of separate arguments would hide them. It is closest in spirit to the evidence tables used in systematic reviews and to auditing's requirement for sufficient, appropriate evidence — enough of it, and of the right kind, before an opinion is issued.[n1]
Its failure mode is the tidy grid itself: a full, well-formatted matrix radiates a completeness it may not have, and its uniform rows invite treating a strong-looking cell as settled when it is merely filled. The classic misuse is summing a column as if the rows were independent — five dashboards fed by one pipeline are one piece of evidence, not five. The guarding discipline is to audit for shared provenance before aggregating and to keep the combined judgment decomposable, so any surprising conclusion can be traced back to the exact rows that produced it.
How it implements the components¶
provenance_chain— a dedicated column records each trace's origin and its strength, kept with the row through aggregation.relevance_rule— each row states the rule by which its trace bears on its claim, so relevance is explicit and comparable across rows.scoped_evidential_weight— the matrix carries per-row weight with its scope attached, preventing a locally strong row from being read as globally strong.evidence_synthesis_interface— its column-wise synthesis pass is the combine-many-relations operation, detecting convergence, conflict, and double-counting to yield an auditable aggregate.
It does not author the individual trace, claim, or reasoning bridge of any single relation (observable_trace_record, hypothesis_or_claim_anchor, inferential_link_model) — that one-argument work belongs to the Claim-Evidence-Reasoning Card, its nearest twin. The card makes one inference legible; the matrix holds and combines many.
Related¶
- Instantiates: Evidentiary Trace Warranting — the matrix is the archetype's general synthesis surface, holding many relations at once.
- Consumes: Claim-Evidence-Reasoning Card — each row references one such single-relation record, which the matrix then aggregates.
- Sibling mechanisms: Admissibility or Relevance Gate · Claim-Evidence-Reasoning Card · Defeater Register · Evidence Strength Ladder · Evidence Update Review · Relevance and Alternative Explanation Check · Trace-to-Claim Diagram
Editorial Notes¶
Form Classification¶
Form family: Representation, Specification & Plan
Rationale: Evidence Relation Matrix operates as a non-executable information artifact that externalizes static or prospective structure because it maps traces to claims with provenance, relevance, link type, weight, scope, and defeaters.
Independent corroboration: The frozen evidence defines Evidence Relation Matrix as 'Maps traces to claims with provenance, relevance, link type, weight, scope, and defeaters', so its operative form is Representation, Specification & Plan.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Accounting & Auditing
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: Auditing formalized itemized evaluation of evidence provenance, relevance, reliability, sufficiency, appropriateness, and its aggregation into a supportable conclusion.
Related originating lineages:
- Law & Governance — Mapping traces to claims by relevance, weight, scope, and defeaters most directly extends legal evidence analysis.
- Medicine & Healthcare — Systematic-review evidence tables materially developed row-wise appraisal and synthesis across heterogeneous studies.
- Rhetoric — Argumentation theory materially contributes explicit claim-warrant and rebuttal relations.
Review resolution: ISA 500's sufficient-and-appropriate evidence framework is the strongest professional analogue to the trace-to-claim matrix. Legal and argumentative warranting add link types and defeaters to the encyclopedia synthesis.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Researched adjudication after independent review; high confidence.
Sources consulted:
Notes¶
[n1] Auditing standards (e.g., ISA 500, Audit Evidence) require an auditor to obtain sufficient appropriate evidence — sufficient in quantity, appropriate in relevance and reliability — before forming an opinion. The standard's insistence on relevance and reliability per item, aggregated to a conclusion, is the professional analog of what an evidence relation matrix operationalizes. ↩