Narrator Reliability Matrix¶
Review matrix — instantiates Source Distortion Modeling
Scores or describes reliability by claim type, evidence base, motive, vantage, consistency, and corroboration.
The Narrator Reliability Matrix is a grid. Down one axis run the classes of claim a single source makes — sensory detail, sequence, quoted numbers, attributed motive, causal interpretation. Across the other run the reliability dimensions — evidence base, vantage, internal consistency, independent corroboration, stake. Each cell holds a scoped grade, and no cell holds a verdict about the whole source. Its defining move — the thing that separates it from a yes/no trust call — is its refusal to grade the source as a whole: it insists that the honest answer to "can we trust them?" is a table, not a number, because one source can be pin-sharp on what they saw and worthless on why it happened.
Example¶
A city-desk editor gets a detailed tip from a single anonymous source inside a housing authority: names, dates, a dollar figure, and an account of who ordered a demolition and why. Rather than deciding "credible" or "not," the editor builds a matrix. The rows are the source's claim classes — the eyewitness account of the meeting, the document references, the dollar figure, the named decision-maker, the imputed motive. The columns ask: first-hand vantage? internally consistent? independently corroborated? what is the source's stake? The grid fills in unevenly. The source was in the room (high vantage on what was said); the dollar figure matches a filed budget line (corroborated); but the claim that a specific official "ordered" it rests on inference, and the source holds a grudge against that official (stake high, vantage low). The story runs the corroborated facts, attributes the motive claim as the source's own characterization, and drops one number nothing else supports. The matrix turned a single tip into five differently-trusted claims.
How it works¶
- Partition first, grade second. Break the account into claim classes before scoring; a single global grade is exactly what the matrix exists to refuse.
- Score each cell on scoped dimensions — vantage, consistency, corroboration, stake — never one blended "trust" figure.
- Read the pattern, not just the cells. A source strong on detail and weak on causation is a recognizable, informative shape.
- Carry the per-cell grade downstream so a strong cell is not dragged down by a weak neighbor, or a weak one flattered by a strong one.
Tuning parameters¶
- Partition granularity — coarse (four claim types) versus fine (a dozen). Finer catches "reliable for X, not Y" splits but costs effort and invites false precision.
- Grade vocabulary — numeric weights versus qualitative bands (accept / attribute / bracket / reject). Numbers aggregate cleanly but imply a precision the evidence rarely has.
- Dimension set — which columns you score on; adding a corroboration column makes the matrix depend on trace-gathering it does not itself perform.
- Aggregation rule — whether and how cells combine into a claim-level call, or stay deliberately disaggregated.
When it helps, and when it misleads¶
Its strength is that it kills the most common source error — arguing about whether to trust a source as a whole instead of partitioning claim-level reliability. It makes "trustworthy for what?" the operative question and exposes the informative shape of a source: great on sequence, hopeless on motive. Intelligence practice codified exactly this split long ago — the Admiralty (NATO) source-evaluation system grades source reliability and information credibility on two separate scales precisely so the two never collapse into one letter.[n1]
Its failure mode is that the grades are only as good as their inputs, and a tidy grid lends false authority to what are really estimates; worse, a matrix can be filled in to rationalize a conclusion already reached, its columns chosen to flatter the desired claim. The classic misuse is treating a high-corroboration cell as settled when the "corroboration" came from sources sharing the tipster's upstream. The guarding discipline is to grade the evidence you actually have, mark empty cells as unknown rather than average them away, and never let the matrix stand in for the independence check it merely cites.
How it implements the components¶
claim_type_reliability_partition— the matrix's rows are the partition; splitting the account into claim classes is its first and definitional act.reliability_weighting_rule— each cell carries a scoped weight or band, so reliability is assigned per claim class rather than to the source as a whole.
It does not gather the independent evidence its columns cite (corroboration_and_countertrace_set) — that is the Corroboration Ladder — nor decide what may be released downstream (distortion_corrected_reading, sympathetic_vs_skeptical_reading_pair) — that is the Claim Release Gate. The matrix scores; its siblings gather and gate.
Related¶
- Instantiates: Source Distortion Modeling — supplies the claim-level reliability partition the archetype demands.
- Consumes: Corroboration Ladder — its corroboration column reads off the ladder's rungs.
- Sibling mechanisms: Distortion Model Card · Account/Event Reconstruction Table · Vantage-Bias Interview Protocol · Corroboration Ladder · Contradiction Timeline · Motive-Opportunity-Bias Analysis · Claim Release Gate
Editorial Notes¶
Form Classification¶
Form family: Assessment, Review & Assurance
Rationale: Narrator Reliability Matrix operates as a bounded evaluation of existing evidence or work that produces a finding or disposition because it scores or describes reliability by claim type, evidence base, motive, vantage, consistency, and corroboration.
Independent corroboration: The frozen evidence defines Narrator Reliability Matrix as 'Scores or describes reliability by claim type, evidence base, motive, vantage, consistency, and corroboration', so its operative form is Assessment, Review & Assurance.
Nearest alternative: Analysis, Modeling & Optimization — Narrator Reliability Matrix includes features of an analytical, modeling, inference, comparison, or optimization procedure that derives insight or a solution, but its defining operation is a bounded evaluation of existing evidence or work that produces a finding or disposition.
Review outcome: Independent reviewer agreement; medium confidence.
Origin Attribution¶
Primary origin: Literature & Literary Theory
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: The concept of an unreliable narrator and attention to motive, vantage, and selective telling shaped its narrator-specific framing.
Related originating lineages:
- Communication & Media Studies — Journalistic source evaluation contributes claim-specific credibility and evidence-base assessment.
- History & Historiography — Historical source criticism contributes corroboration, vantage, motive, and consistency tests.
- Security Studies & Intelligence Analysis — Intelligence source-evaluation systems, especially the Admiralty/NATO separation of source reliability from report credibility, supply the matrix's disciplinary core.
Review resolution: Authoritative-source research resolves the primary-origin disagreement. Narratorial unreliability is a named narratological concept; converting it into a scored source-reliability matrix synthesizes intelligence and historical source criticism. Origin breadth is limited to formative lineages; present-day applicability is recorded separately as domain_reach=multi_domain.
Attribution caveat: The term comes from narratology, while the scored matrix behaves more like source-criticism practice.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Researched adjudication after independent review; medium confidence.
Sources consulted:
Notes¶
[n1] The Admiralty Code (also called the NATO System) rates a source's reliability on a letter scale (A–F) and the credibility of each individual report on a separate numeric scale (1–6); keeping the two axes independent is the whole point — a usually-reliable source can still relay a doubtful report, and an untested source can relay one that is independently confirmed. ↩