Evidence Stream Matrix¶
Template — instantiates Independent Evidence Triangulation
A one-row-per-stream table recording each stream's claim coverage, origin, method, population, time, quality, uncertainty, and known dependencies before any synthesis begins.
Before you can weigh evidence, you have to lay it out honestly. The Evidence Stream Matrix is the template that makes the architecture of an evidence set explicit: one row per stream, and columns that pin down what each stream actually is — what claim it bears on, where it came from, what method produced it, which population and time window it covers, whether it clears the quality floor, how uncertain it is, and what it is known to depend on. Its defining trait is that it is stream-centric and pre-synthetic: it catalogs the raw material and freezes the roster so that convenient streams cannot be quietly added or awkward ones dropped once results are in. It says nothing yet about whether the streams agree — only about what they are.
Example¶
A conservation team needs to know whether a threatened salamander population in a watershed is declining — the decision is whether to petition for protected status. Five streams are on the table, and before anyone argues about the answer, they fill a matrix. Row one: a decade of standardized night surveys — method, transect counts; population, three headwater streams; time, 2015–2024; quality, high; uncertainty, wide in wet years; dependencies, none. Row two: environmental-DNA water sampling — different method, overlapping sites, shorter record, a known false-negative issue in silty flow. Row three: a citizen-science observation app — broad coverage, but self-selected effort and no calibration, flagged as below the quality floor for population trend and admitted only as presence data. Rows four and five: a habitat model and a museum-specimen record, each with its own coverage and blind spots.
Laid out this way, two things jump out that a pile of reports would hide: the night surveys and the eDNA sampling cover the same three streams in the same years — a dependency to watch — and the citizen app, the most voluminous source, is the weakest and cannot carry a trend claim on its own. The matrix has not decided anything; it has made the evidence set legible enough to decide with.
How it works¶
- Anchor every row to the same scoped claim. A column records what each stream can and cannot observe about that claim, so streams answering a neighboring question are visibly off-target.
- One row per stream, fixed before results. The roster is frozen; late additions or deletions are logged, not silent.
- A quality-floor column gates influence. Each stream is marked pass/fail against minimum provenance, validity, and completeness bars; sub-floor streams are demoted or admitted only for what they can support.
- An uncertainty column travels with each stream — its bias exposure, missingness, and scope limits kept per-row rather than blended away.
Tuning parameters¶
- Column set — the attributes tracked. More columns capture more structure but raise the filling cost and the temptation to leave cells blank.
- Quality-floor height — how demanding the minimum bar is. A high floor keeps weak signals out but can starve a low-stakes inquiry of usable streams.
- Stream granularity — whether a multi-site survey is one row or several. Splitting surfaces within-stream heterogeneity; lumping keeps the table readable.
- Dependency-column depth — a quick "shares site/instrument with row X" note versus a full lineage; the matrix flags dependencies, but deep tracing belongs elsewhere.
When it helps, and when it misleads¶
Its strength is that it turns "we looked at a lot of evidence" into an inspectable roster where coverage gaps, quality failures, and overlapping populations are visible before anyone starts arguing about the conclusion. Its quality-floor column is a disciplined echo of structured evidence-grading schemes[n1] and is the matrix's best guard against a flood of weak streams drowning one strong one.
Its failure mode is that a filled matrix feels like analysis when it is only bookkeeping. A tidy grid can lull a team into thinking the hard work is done, when the matrix has not yet weighed a single stream or resolved a single conflict. It can also entrench a bad roster: freezing the streams early is a virtue against cherry-picking but a liability if the initial inventory missed the one stream that mattered. The guarding discipline is to treat the matrix as the input to synthesis, not a substitute for it, and to allow roster changes only through a logged, justified amendment.
How it implements the components¶
claim_and_decision_scope— the matrix is built around one scoped claim, and each row's coverage column records what that stream observes of it, keeping off-target streams visible.evidence_stream_inventory— the row set is the explicit, frozen inventory of every source, method, and dataset in play.evidence_quality_floor— the pass/fail quality column operationalizes the minimum bar each stream must clear before it can influence synthesis.stream_specific_uncertainty_profile— the per-row uncertainty and scope-limit columns keep each stream's reasons-it-might-be-wrong from being averaged into a single band.
It catalogs streams but does not act on them: classifying whether sources independently corroborate a claim is source_diversity_plan, owned by Cross-Source Corroboration Table, and pricing the result into a confidence_calibration_statement is Confidence Update Worksheet's.
Related¶
- Instantiates: Independent Evidence Triangulation — supplies the frozen, quality-gated evidence inventory the whole lifecycle reads from.
- Sibling mechanisms: Cross-Source Corroboration Table · Confidence Update Worksheet · Source Dependency Graph · Multi-Method Study Design
Editorial Notes¶
Form Classification¶
Form family: Representation, Specification & Plan
Rationale: Evidence Stream Matrix operates as a non-executable information artifact that externalizes static or prospective structure because it a one-row-per-stream table recording each stream's claim coverage, origin, method, population, time, quality, uncertainty, and known dependencies before any synthesis begins.
Independent corroboration: The frozen evidence defines Evidence Stream Matrix as 'A one-row-per-stream table recording each stream's claim coverage, origin, method, population, time, quality, uncertainty, and known dependencies before any synthesis begins', so its operative form is Representation, Specification & Plan.
Review outcome: Independent reviewer agreement; high confidence.
Origin Attribution¶
Primary origin: Statistics & Experimental Design
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: Cataloguing method, population, time, quality, uncertainty, and dependence before synthesis comes from statistical study appraisal and research synthesis.
Related originating lineages:
- Data Science & Analytics — Modern analytics materially shaped the explicit cataloging of pipeline origin, data dependencies, and stream coverage.
- Medicine & Healthcare — Systematic-review evidence tables in evidence-based health practice materially shaped the one-row-per-study or stream form.
Review resolution: Both reviewers agree that statistics_experimental_design is primary. I retain medicine_healthcare, data_science only as formative origin lineages; cross_disciplinary_synthesis is appropriate because the final form materially combines the agreed primary with the retained formative lineages. Reach is multi_domain because the structure transfers across several fields but is not a near-universal human pattern, an applicability judgment kept separate from provenance. Encyclopedia synthesis is true because the exact generalized artifact is an encyclopedia-authored combination or refinement. No unresolved historical ambiguity remains after reconciling the secondary fields.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Reconciled after independent review; medium confidence.
Notes¶
[n1] GRADE (Grading of Recommendations, Assessment, Development, and Evaluations) is a standard framework in medicine for rating the certainty of each body of evidence before recommendations are drawn. The matrix's quality-floor column is a lightweight, general-purpose cousin: a minimum bar each stream must clear to influence the conclusion. ↩