Evidence¶
Core Idea¶
Evidence is the structural relation between an observable trace and a hypothesis about an unobservable state of affairs: a thing functions as evidence for a claim when it is more probable to occur if the claim is true than if it is false, and when its production is traceable enough that the inference can be challenged and defended. Three structural commitments travel with the relation across substrates. First, evidence is defeasible and weight-bearing: any single item licenses only a graded confidence in the conclusion, not a deduction, and additional information can raise, lower, or wholly defeat its weight. Second, evidence carries provenance: the chain of custody from the underlying event to the trace to its presentation is itself part of the evidence, because breaks in that chain are the standard route to defeating an inference. Third, evidence combines non-arithmetically: two independent corroborating items raise confidence more than one does, but two items from a common source raise it less than the independence picture would suggest, so the combination rule must care about dependence among traces.
What makes this a prime rather than the vocabulary of any one inquiry is that the same role-set — underlying state of affairs leading to a trace-producing mechanism, to a trace, through a chain of custody, via an inference rule, with aggregation across multiple traces and a set of defeaters — structures evidential reasoning in science, law, medicine, history, journalism, intelligence assessment, software debugging, and archaeology. The substrates differ entirely; the skeleton does not. The relation is inherently bound to inquiry practice — it carries an epistemic-normative load (burden of proof, standards of admissibility) and an institutional history in the legal rules of evidence — which is why it reads as framed even though its role-set is medium-neutral.
How would you explain it like I'm…
Clues That Point
Clues You Can Trust
Traces Behind a Claim
Structural Signature¶
the unobservable state of affairs — the trace-producing mechanism — the observable trace — the chain of custody from event to trace — the inference rule from trace to hypothesis — the aggregation rule across multiple traces — the set of defeaters
Evidence is present when each of the following holds:
- An unobservable state (the target hypothesis). A claim about a state of affairs not directly observed; a trace is evidence only relative to a specific such hypothesis.
- A trace-producing mechanism (the generator). The process by which the underlying state leaves an observable mark; its character (study design, taphonomy, instrument) governs how strongly the trace bears on the claim.
- A trace (the observable). The thing that functions as evidence — probative when it is more probable if the claim is true than if false.
- A chain of custody (the provenance relation). The traceable path from the underlying event to the trace to its presentation; breaks in this chain are the standard route to defeating an inference, so provenance is itself part of the evidence.
- An inference rule (the weight-bearing relation). A defeasible mapping from trace to graded confidence in the hypothesis — never a deduction; any single item licenses only weight, which further information can raise, lower, or defeat.
- An aggregation rule (the combination invariant). A rule combining multiple traces non-arithmetically: independent corroboration raises confidence more, common-source traces less, so the rule must care about dependence among traces.
- Defeaters (the defeasibility invariant). Conditions that would undercut the inference — confounders, custody breaks, alternative explanations — whose anticipation is part of evaluating the evidence.
The components compose into a role-set in which any disputed conclusion localizes to one handle — a provenance break, a wrong combination rule, an ignored defeater — making evidential reasoning auditable rather than holistic.
What It Is Not¶
- Not provenance.
provenanceis the chain of custody from event to trace — one component of the evidence relation, the handle that defeats inferences when broken. Evidence is the whole role-set; provenance is the lineage sub-relation within it. - Not the inference rule alone.
inductive_reasoning,abductive_reasoning, andbayesian_updatingare modes of inference from evidence; evidence is the trace-to-hypothesis relation those modes operate on. The rule combines traces; evidence is what is combined. - Not belief.
belief_formationis the resulting doxastic state; evidence is the trace that licenses graded confidence. One can have strong evidence and form no belief, or believe with no evidence. - Not data. Data are traces before a hypothesis is fixed; a trace becomes evidence only relative to a specific claim and inference rule. "Having data" is not "having evidence for a hypothesis" — the prime's central distinction.
- Not proof. Proof is deductive and indefeasible; evidence is defeasible and weight-bearing, licensing only graded confidence that further information can raise, lower, or defeat.
- Not a signal in the detection sense.
signal_detection_theoryconcerns discriminating signal from noise at a sensor; evidence is the broader relation in which a trace bears on an unobservable hypothesis through a custody chain and an aggregation rule. - Common misclassification. Gathering data, then finding the claim it best supports, and treating the fit as evidential weight (HARKing). Catch it by asking whether the hypothesis was specified before the trace was examined: if the claim was chosen to fit the data, the trace was never a likelihood-discriminating test of it and its weight is illusory.
Broad Use¶
Evidence is the working unit of every discipline that reasons from observable traces to unobservable states. In law, physical exhibits, testimony, documents, and expert opinion must be both probative and admissible, and juries weigh them against an explicit standard of proof; the Federal Rules of Evidence operationalize the structural commitments directly.[1] In science, data are evidence for or against hypotheses, with strength depending on study design (the trace-producing mechanism), replication (independent corroboration), and freedom from confounders (defeaters).[2] In medicine, signs and symptoms are evidence for diagnoses, and differential diagnosis is structurally a Bayesian-style competition among hypotheses, each finding shifting posterior weight.[3] In history, primary sources are evidence for past events, with provenance, internal-versus-external criticism, and triangulation across independent sources playing the chain-of-custody and combination roles.[4] In journalism and intelligence, sources, documents, and intercepts function as evidence for assessments, and tradecraft is precisely the discipline of provenance, source reliability, and corroboration. In software debugging, a stack trace, log entry, or reproducer is evidence for a hypothesis about a defect, with bisection and instrumentation as evidence-production strategies and flaky reproducers as defeaters.[5] And in archaeology, physical remains and stratigraphic context are evidence for inferences about past life, with taphonomy supplying the trace-producing-mechanism analysis.[6]
Clarity¶
Naming the structural relation moves the analysis from "we have lots of data" to a specific set of questions: evidence for what claim, by what trace-producing mechanism, with what chain of custody, in combination with what other traces, vulnerable to what defeaters? It clarifies the common confusion between having data and having evidence for a hypothesis — data become evidence only relative to a specific claim and a specific inference rule. It also separates the probative question (how strongly does this trace bear on the claim?) from the admissibility question (does our procedure permit us to use it?), a distinction law makes explicit and other domains often blur. The clarifying force is to dissolve a vague mass of "findings" into a small set of named handles, each of which can be examined and contested on its own, so that an argument about a conclusion becomes a set of sharper arguments about specific evidential components.
Manages Complexity¶
The evidence relation factors any inference into a small set of analytically separable handles: the trace-producing mechanism, the provenance chain, the inference rule, the combination rule, and the candidate defeaters. When a downstream conclusion is in dispute, the dispute localizes to one of those handles, which compresses the argumentative space dramatically. Forensic, scientific, and historical disputes that look like disagreements about the conclusion almost always turn out to be disagreements about one specific evidential handle — a provenance break, a wrong combination rule, an ignored defeater — and isolating which handle is at issue converts an unbounded quarrel into a tractable one. The reduction also makes evidential reasoning teachable and auditable: rather than cultivating an inarticulate sense of "what the evidence shows," a practitioner can walk each handle in turn, which is exactly what legal procedure, scientific peer review, and intelligence tradecraft institutionalize. The complexity managed is the combinatorial tangle of many traces bearing on many hypotheses; the handles give it a fixed, inspectable structure.
Abstract Reasoning¶
Evidence relations support reasoning under the explicit assumption that all inference from traces is defeasible and weight-bearing, which unlocks several second-order moves. Reasoning about the optimal next evidence to collect — experimental design, discovery requests, source cultivation, active learning — is uniformly an instance of asking which trace would most discriminate between the live hypotheses given the current posterior. Reasoning about evidence of absence versus absence of evidence turns on whether a positive finding was likely: a failed search where success was probable is evidence against the claim, while a failed search where success was improbable is not, and treating these symmetrically is a recurring error. Reasoning about correlated traces warns that two traces sharing a producing mechanism do not constitute independent corroboration, a pattern that recurs in news (outlets citing one source), in clinical findings (two tests on the same biomarker), and in intelligence (two reports from one asset). And reasoning about defeaters as predictions notes that a hypothesis which systematically generates defeaters of its own apparent evidence is in worse shape than one that does not, even before any single defeater is realized — a structural feature that adversarial testing deliberately exploits.
Knowledge Transfer¶
Because evidence is a role-set rather than a domain, an intervention discovered in one inquiry practice transfers to any other by re-identifying the roles, and the prime's reach lies in how routinely this works. Chain-of-custody discipline from law transfers to data lineage in engineering: the legal rule that admissibility depends on documented custody from collection to courtroom becomes the engineering rule that traces whose lineage is unrecorded cannot ground a downstream claim, and the intervention is identical — instrument the pipeline and attest each hand-off. Triangulation in history, corroboration in intelligence, and replication in science are the same move under three names — produce a structurally independent trace and check for convergence — so a historian's source-criticism and a scientist's replication study are recognizably the same upgrade to confidence. The legal notion of burden and standard of proof transfers to the statistical significance threshold and to the policy bar-for-action: choosing a threshold for acting on an evidence-weighted conclusion is present in every evidential domain, and the legal habit of making that threshold explicit and contestable is exportable to fields that usually leave it implicit. Adversarial examination in law transfers to red-teaming in intelligence and pre-mortems in organizations, all generic disciplines for surfacing defeaters before acting. And Bayesian updating supplies the canonical aggregation rule, so that recognizing an evidential problem suggests the Bayesian form even where practitioners lack the vocabulary — while the prime remains broader than any one rule, since Dempster-Shafer weighing, legal balancing, and narrative coherence in history all instantiate the same role-set with different aggregation functions.[7] In every transfer the practitioner runs the same diagnosis — identify the trace, its producing mechanism, its custody chain, the inference and combination rules, and the live defeaters — and the transfer holds because none of these roles is tied to a substrate: a bridge inspector reading a hairline crack and an intelligence analyst weighing two intercepts are performing the same structural inference, distinguished only by the material of the trace.
Examples¶
Formal/abstract¶
A diagnostic medical test interpreted by Bayes' rule is the cleanest formal instance, instantiating the inference and aggregation roles quantitatively. The unobservable state is whether the patient has a disease \(D\); the trace-producing mechanism is the assay, characterized by its sensitivity \(P(+\mid D)\) and specificity \(P(-\mid \neg D)\); the trace is the positive result.[3] The inference rule is Bayes' theorem: the posterior \(P(D\mid +) = \frac{P(+\mid D)\,P(D)}{P(+)}\), which makes the prime's defeasibility/weight-bearing invariant exact — a positive test does not deduce disease, it shifts a prior to a graded posterior.[8] The famous base-rate result is the prime's claim made vivid: for a rare disease (low prior \(P(D)\)), even a highly sensitive and specific test yields a posterior well below certainty, because most positives are false positives drawn from the large healthy population.[3] The aggregation rule's non-arithmetic, dependence-sensitive character is precise here too: a second independent test (different mechanism, e.g. imaging versus blood assay) multiplies the likelihood ratios and sharply raises the posterior, whereas re-running the same assay — a common-source trace — adds little, because its errors are correlated with the first.[3] The defeaters are explicit: a confounder that inflates the marker, a mislabeled sample (a custody break), an alternative condition producing the same trace. The reasoning-about-next-evidence move is the clinician's: order the test whose likelihood ratio most discriminates between the live hypotheses given the current posterior.
Mapped back: The diagnostic test instantiates every role — target hypothesis (disease), trace-producing mechanism (the assay with its sensitivity/specificity), trace (the result), defeasible inference (Bayesian update), dependence-aware aggregation (independent confirmatory test versus correlated re-test), and defeaters — and the base-rate and correlated-trace results are the prime's weight-bearing and non-arithmetic-combination invariants in exact form.
Applied/industry¶
Software debugging shows the identical role-set in an engineering substrate, with no epistemological vocabulary required. The unobservable state is the hypothesis about a defect — say, "a race condition in the cache-eviction path corrupts the record." The trace-producing mechanisms are the running system and the engineer's instrumentation: logging, a debugger, a stress harness. The traces are a stack trace, a log line, a failing reproducer. The chain of custody matters exactly as the prime says: a stack trace from an optimized build with inlined frames, or a log from a machine whose clock was skewed, has a broken provenance that defeats the inference — which is why engineers pin builds and record environments. The inference rule is defeasible: a reproducer raises confidence in the race hypothesis but does not prove it; a flaky reproducer (the prime's canonical defeater) that fails to reproduce on retry undercuts the weight. The aggregation invariant appears in the warning against correlated traces: two failures observed on the same misconfigured CI node are not independent corroboration, just as two news outlets citing one source are not.[5] The reasoning-about-next-evidence move is bisection — choosing the commit or the instrumentation point that most discriminates between the live defect hypotheses, the active-learning instance of "collect the trace that best splits the posterior." The transfer the prime promises is literal: chain-of-custody discipline imported from law becomes data-lineage instrumentation (attest each hand-off in the pipeline), and adversarial examination becomes the pre-mortem or red-team that surfaces defeaters before the fix is shipped.[9]
Mapped back: Debugging runs the prime end-to-end — defect hypothesis, instrumentation as trace-producing mechanism, traces with auditable custody (pinned builds), defeasible inference, dependence-aware aggregation (correlated CI failures), and defeaters (flaky reproducers) — and demonstrates the cross-domain transfer: the bridge inspector, the clinician, and the engineer run one structural inference from trace to hidden state, distinguished only by the material of the trace.
Structural Tensions¶
T1 — Data versus Evidence-for-a-Hypothesis (Relational Scope). A trace is evidence only relative to a specific hypothesis; "having data" is not "having evidence." The tension is that traces are collected before the hypothesis is fixed, tempting their reuse for whatever claim is convenient. The failure mode is hypothesis drift: gathering data, then finding the claim it best supports — the structure of HARKing and post-hoc storytelling — so the same trace appears to confirm a hypothesis it was never a likelihood-discriminating test of. Diagnostic: ask whether the hypothesis was specified before the trace was examined; if the claim was chosen to fit the data, the trace's evidential weight for it is illusory, because it was never more probable under that claim than its rivals.
T2 — Probative Weight versus Admissibility (Two Gates). Evidence must clear two distinct gates — how strongly it bears on the claim (probative) and whether the procedure permits its use (admissible) — and they are routinely blurred. The failure mode is conflating force with permission: discarding a highly probative trace because of a procedural defect as if it were uninformative, or admitting a procedurally clean trace as if admissibility implied weight. Diagnostic: ask the two questions separately — "how much does this shift my confidence?" and "may I use it under my rules?"; a trace can be strong but inadmissible, or admissible but weak, and treating either gate as the other mis-weights the conclusion.
T3 — Independent versus Correlated Traces (Aggregation Dependence). Evidence combines non-arithmetically: independent corroboration raises confidence sharply, but common-source traces raise it far less than their count suggests. The failure mode is double-counting a single source: treating two outlets citing one wire report, two tests of one biomarker, or two CI failures on one bad node as independent confirmation, inflating confidence on what is structurally one trace. Diagnostic: ask whether the traces share a producing mechanism or upstream source; if they do, their errors are correlated and the aggregation rule must discount them — counting correlated traces as independent is the most common route to false certainty.
T4 — Absence of Evidence versus Evidence of Absence (Sign Asymmetry). A failed search is evidence against a claim only when a positive finding was likely had the claim been true; otherwise it is uninformative. The tension is the asymmetry between these, which intuition collapses. The failure mode is symmetric treatment: reading "we found nothing" as disconfirmation regardless of whether the search could have found something, or dismissing a genuinely powerful null result as "just absence of evidence." Diagnostic: ask how probable the positive trace was under the hypothesis given the search's power; a null from a search that would likely have detected the effect is real evidence against, while a null from an underpowered search carries almost no weight, and the two must not be equated.
T5 — Chain of Custody versus Trace Content (Provenance Coupling). The trace's content and its provenance are separable handles, and a probative trace with a broken custody chain is defeated regardless of what it shows. The tension is that attention flows to the vivid content while the unglamorous lineage is where inferences actually break. The failure mode is content fixation: reasoning from what a trace appears to show while ignoring that its path from event to presentation is unrecorded or compromised (a skewed clock, an inlined build, an unverified source). Diagnostic: before weighing what the trace says, audit how it reached you; if any hand-off from the underlying event to the presented trace is undocumented or breakable, the provenance defeater applies and the content's apparent force is suspended until the chain is repaired.
T6 — Defeasible Weight versus Threshold for Action (Standard-of-Proof Locus). Evidence only ever licenses graded confidence, but action requires crossing a threshold, and the threshold is a separate, often implicit, choice. The tension is between the continuous weight evidence supplies and the binary decision it must feed. The failure mode is implicit-bar reasoning: treating "the evidence points this way" as automatically sufficient to act, with no explicit standard of proof, so the threshold floats to whatever the advocate prefers. Diagnostic: ask what confidence level is required to act here and whether the evidence clears it — make the standard explicit and contestable (as law does with "beyond reasonable doubt" versus "preponderance"); a conclusion presented as action-ready without a named threshold has hidden the most consequential choice in the inference.
Structural–Framed Character¶
Evidence sits in the mixed-framed band of the structural–framed spectrum, with a frontmatter aggregate of 0.6 driven above the midline by its inseparability from inquiry as a practice. The relational core is real and richly articulated — a target hypothesis, a trace-producing mechanism, a trace, a custody chain, an inference rule, an aggregation rule, and a set of defeaters — and that role-set is genuinely medium-neutral, which is why a clinician, a bridge inspector, and a software engineer run the same trace-to-hypothesis inference. But what tips the prime into the mixed-framed band is that the whole relation only makes sense relative to an agent conducting inquiry.
The decisive criterion is human-practice-boundedness, scored the full 1.0: evidence is inherently an inquiry-practice category. A trace is evidence only relative to a hypothesis being entertained and a question being asked, so the relation presupposes an inquirer — there is no evidence-for-a-claim in a physical substrate that asks nothing. That single maxed criterion is the heaviest contributor to the 0.6. The remaining four each sit at 0.5, recording genuine but partial framing. The vocabulary half-travels (vocab_travels 0.5): the role-set ports from law to science to debugging, but the epistemic idiom — probative, defeasible, burden of proof — travels with it. It carries epistemic-normative load (evaluative_weight 0.5): standards of proof, admissibility, and burden are normatively charged, though the bare likelihood relation (more probable if the claim is true than if false) is neutral. Its origin is partly institutional (institutional_origin 0.5): epistemology is formal, but the legal rules of evidence are a concrete institutional history the term carries. And invoking it half-imports (import_vs_recognize 0.5): one partly recognizes a trace-to-hypothesis relation, partly imports the inquiry frame of weighing and contesting.
The honest reading is a prime whose role-set is substrate-neutral enough to support real cross-domain transfer, but whose existence presupposes the practice of inquiry — the maxed human-practice score, with the four supporting halves, correctly places it in the mixed-framed band without pushing it to the institutional extreme of certification.
Substrate Independence¶
Evidence is a strongly substrate-independent prime — composite 4 / 5 on the substrate-independence scale. Its breadth is at the ceiling (domain breadth 5): the trace-to-hypothesis relation — with its provenance, defeasibility, and non-arithmetic combination — recurs with the same force across law (exhibits, testimony, standards of proof), science (data weighed by study design and replication), medicine (signs and symptoms in a differential), history (primary sources with internal and external criticism), journalism and intelligence (sources and corroboration as tradecraft), software debugging (stack traces and reproducers, with flaky reproducers as defeaters), and archaeology (remains in stratigraphic context). Transfer evidence is likewise at the ceiling (5): the role-set ports so cleanly that a clinician, a bridge inspector, and a software engineer run the same trace-to-hypothesis inference, and the chain-of-custody, defeater, and aggregation handles carry unchanged. Structural abstraction is a notch lower at 4 because the relation, though medium-neutral in its role-set, is inherently an inquiry-practice category: a trace is evidence only relative to a hypothesis being entertained and a question being asked, so it presupposes an inquirer and has no purely physical substrate that asks nothing. The composite of 4 records a role-set recognized across nearly every discipline that reasons from traces, held just short of 5 by its dependence on the practice of inquiry — the same human-practice centroid that places it on the framed side.
- Composite substrate independence — 4 / 5
- Domain breadth — 5 / 5
- Structural abstraction — 4 / 5
- Transfer evidence — 5 / 5
Relationships to Other Abstractions¶
Current abstraction Evidence Prime
Parents (1) — more general patterns this builds on
-
Evidence is part of Provenance Prime
Evidence contains provenance as the chain-of-custody relation connecting an underlying event to its observable trace and presentation.The live evidence identity requires a trace-producing mechanism and a traceable path from event to trace to presentation. Provenance is one of its fixed audit handles: a break supplies a defeater. The inherited record's prose already says provenance is a component of evidence but points the edge from component to whole and encodes part_of under decompose.
Children (19) — more specific cases that build on this
-
Code Smell Domain-specific is a kind of Evidence
A code smell is defeasible evidence specialized to observable source-code features that support a hypothesis about a deeper design or maintainability problem.It inherits an observable trace, a supported but revisable hypothesis, provenance and context, and possible defeat by contrary evidence. The child fixes the traces to code structure, the hidden states to design and maintainability problems, and the intervention probe to behavior-preserving refactoring.
-
Combinatorial method (linguistics) Domain-specific is a kind of Evidence
The proposed strict upward parent is
prime:evidence.prime:evidence is the nearest broader Prime; the source domain and invariant supply the residual. This is a proposal-only workspace relationship: the accepted Prime supplies a genuinely instantiated structural prerequisite or superclass, while Combinatorial method (linguistics) adds domain-specific constraints. The entry does not collapse into that parent because the domain-specific identity determined by a proposed reading is supported by converging structural, archaeological, and contextual distributions and generates testable predictions across texts It also declines a nearby thematic catalog node: the neighbor does not literally subsume the constitutive identity of Combinatorial method (linguistics). This explicit assert-and-decline pattern keeps the proposed DAG narrow and prevents a merely thematic edge. The prospective workspace queue contains one strict upward edge toprime:evidence. No live DAG mutation is authorized. -
Evidence (law) Domain-specific is a kind of Evidence
The proposed strict upward parent is
prime:evidence.The legal framework is organized around trace-to-proposition evidentiary relations; jurisdictional admissibility, foundation, allocation, and burden rules provide the autonomous legal specialization. The edge is proposal-only and points to a frozen prior-baseline Prime. The entry does not collapse into the parent because the authoritative procedural framework connecting proffered proof to legally permissible fact determination, not evidence in the general epistemic sense, one item admitted in a case, or a universal list of reliable information A thematic neighbor is declined whenever it does not literally subsume that rule. The prospective workspace queue contains one strict upward edge toprime:evidence. No live DAG mutation is authorized.
- Eyewitness testimony Domain-specific is a kind of Evidence
The proposed strict upward parent is `prime:evidence`.Eyewitness testimony is a form of evidence whose reliability depends on acquisition and reconstruction; legal-memory mechanisms supply the residual. This is a proposal-only workspace relationship: the accepted Prime supplies a genuinely instantiated structural prerequisite or superclass, while Eyewitness testimony adds domain-specific constraints. The entry does not collapse into that parent because legally presented reconstructive memory evidence and the separation of witness sincerity, confidence, and identification accuracy It also declines a nearby thematic catalog node: the neighbor does not literally subsume the constitutive identity of Eyewitness testimony. This explicit assert-and-decline pattern keeps the proposed DAG narrow and prevents a merely thematic edge. The prospective workspace queue contains one strict upward edge to `prime:evidence`. No live DAG mutation is authorized.
- Flat spot (reflection seismology) Domain-specific is a kind of Evidence
Flat Spot instantiates Evidence because an observable seismic trace bears defeasibly on an unobserved fluid-contact state through an explicit wave-physics generator and a named set of defeaters.The prospective workspace queue contains one strict upward edge to `prime:evidence`. No live DAG mutation is authorized.
- Frequency analysis Domain-specific is a kind of Evidence
The proposed strict upward parent is `prime:evidence`.prime:evidence is the nearest broader Prime; the source domain and invariant supply the residual. This is a proposal-only workspace relationship: the accepted Prime supplies a genuinely instantiated structural prerequisite or superclass, while Frequency analysis adds domain-specific constraints. The entry does not collapse into that parent because the domain-specific identity determined by the cipher leaks stable plaintext-derived frequency structure and the inferred mapping explains multiple independent ciphertext statistics It also declines a nearby thematic catalog node: the neighbor does not literally subsume the constitutive identity of Frequency analysis. This explicit assert-and-decline pattern keeps the proposed DAG narrow and prevents a merely thematic edge. The prospective workspace queue contains one strict upward edge to `prime:evidence`. No live DAG mutation is authorized.
- Learning pyramid Domain-specific is a kind of Evidence
The proposed strict upward parent is `prime:evidence`.prime:evidence is the nearest broader Prime; the source domain and invariant supply the residual. This is a proposal-only workspace relationship: the accepted Prime supplies a genuinely instantiated structural prerequisite or superclass, while Learning pyramid adds domain-specific constraints. The entry does not collapse into that parent because the domain-specific identity determined by the identity is the recognizable retention-percentage hierarchy and its status is explicitly recorded as a debunked or unsupported claim It also declines a nearby thematic catalog node: the neighbor does not literally subsume the constitutive identity of Learning pyramid. This explicit assert-and-decline pattern keeps the proposed DAG narrow and prevents a merely thematic edge. The prospective workspace queue contains one strict upward edge to `prime:evidence`. No live DAG mutation is authorized.
- Likelihood principle Domain-specific is a kind of Evidence
The proposed strict upward parent is `prime:evidence`.prime:evidence is the nearest broader Prime; the source domain and invariant supply the autonomous residual. This is a proposal-only workspace relationship: the accepted Prime supplies a genuinely instantiated structural prerequisite or superclass, while Likelihood principle adds domain-specific constraints. The entry does not collapse into that parent because the domain-specific identity determined by the model and parameter are fixed, likelihoods are based on the observed data, proportionality is parameter-independent, and the claim concerns evidential equivalence rather than a complete decision rule It also declines a nearby thematic catalog node: the neighbor does not literally subsume the constitutive identity of Likelihood principle. This explicit assert-and-decline pattern keeps the proposed DAG narrow and prevents a merely thematic edge. The prospective workspace queue contains one strict upward edge to `prime:evidence`. No live DAG mutation is authorized.
- Oral history Domain-specific is a kind of Evidence
The proposed strict upward parent is `prime:evidence`.Oral histories create contextualized testimonial evidence; memory and archival ethics supply the residual. This is a proposal-only workspace relationship: the accepted Prime supplies a genuinely instantiated structural prerequisite or superclass, while Oral history adds domain-specific constraints. The entry does not collapse into that parent because co-created recorded memory as both historical source and event of narration, including ethical stewardship and interpretive reflexivity It also declines a nearby thematic catalog node: the neighbor does not literally subsume the constitutive identity of Oral history. This explicit assert-and-decline pattern keeps the proposed DAG narrow and prevents a merely thematic edge. The prospective workspace queue contains one strict upward edge to `prime:evidence`. No live DAG mutation is authorized.
- Phenomenal conservatism Domain-specific is a kind of Evidence
The proposed strict upward parent is `prime:evidence`.The principle treats conscious seemings as defeasible evidence; epistemic defeat conditions supply the residual. This is a proposal-only workspace relationship: the accepted Prime supplies a genuinely instantiated structural prerequisite or superclass, while Phenomenal conservatism adds domain-specific constraints. The entry does not collapse into that parent because seeming-based default justification with defeasibility, extending across perceptual, memorial, introspective, and inferential appearances It also declines a nearby thematic catalog node: the neighbor does not literally subsume the constitutive identity of Phenomenal conservatism. This explicit assert-and-decline pattern keeps the proposed DAG narrow and prevents a merely thematic edge. The prospective workspace queue contains one strict upward edge to `prime:evidence`. No live DAG mutation is authorized.
- Textual criticism Domain-specific is a kind of Evidence
Textual Criticism instantiates Evidence because attested readings and material witness features bear defeasibly on unobserved textual states and transmission histories through provenance-explicit hypotheses and defeaters.The prospective workspace queue contains one strict upward edge to `prime:evidence`. No live DAG mutation is authorized.
- Tracing (Law) Domain-specific is a kind of Evidence
**Evidence** is the proposed immediate parent.Provenance, Transformation, Identity, Substitution, Ownership, Aggregation, and Priority are related. The modern unitary account emphasizes tracing's neutrality between legal and equitable rights. The prospective queue contains one strict edge to `prime:evidence`. No live DAG mutation is authorized.
- Absence as Information Prime is a kind of Evidence
Absence as information is evidence carried by an expected but missing trace under a modeled absence-generating process.The child specializes evidence to a non-occurrence whose likelihood under competing hypotheses is discriminating. A modeled expectation and an understood absence-generating process are required; uninstrumented silence does not satisfy the genus.
- Production Signature Prime is a kind of Evidence
A production signature is the evidence species in which an observable output trace bears defeasibly on the hypothesis that a particular process made it.It inherits the hidden state, trace-producing mechanism, observable trace, inference rule, dependence-sensitive aggregation, and defeaters. It narrows the hypothesis to source identity and the trace to involuntary, stable, cross-output regularities surviving content normalization.
- Congruence Bias Domain-specific presupposes Evidence
Congruence Bias presupposes an evidential trace-to-hypothesis relation because a probe is congruent or non-diagnostic only by comparing how likely its trace is under the favored hypothesis and its rivals.The child occurs before an outcome is interpreted, but its identity is still relational: a selected probe is defective because its trace-producing mechanism would yield nearly the same outcome under competing states, making the trace weak evidence between them. The live Evidence prime already names the target hypothesis, trace-producing mechanism, comparative probability, defeaters, and next-trace discrimination roles. Evidence is not a piece of the cognitive bias; it is the inquiry relation that must already obtain for the selection to be diagnosably congruent rather than an arbitrary choice.
- Hempel's Paradox Domain-specific presupposes Evidence
The paradox presupposes an evidence-to-hypothesis relation whose logical relevance and quantitative weight can come apart.Black ravens and white shoes matter only as observable traces assessed against a universal hypothesis. Remove that defeasible trace-to-hypothesis relation and neither confirmation nor the dispute over its degree remains. Evidence supplies the relation; the child exposes a formal tension between binary confirmation and informativeness within it.
- Sourcing Ladder Domain-specific presupposes Evidence
A Sourcing Ladder presupposes Evidence because its ranks have meaning only relative to how strongly sources support a particular claim.Source prestige alone is explicitly nonportable across claims; the ladder is selected by the evidentiary architecture of the claim at hand. Without a claim-support relation to weight, there is no higher-rung-wins rule or confidence ceiling to apply.
- Aviation accident analysis Domain-specific is a decomposition of Evidence
The accepted reference-grade review places Aviation accident analysis under Evidence because the child instantiates or depends on the parent's broader structure while retaining its own constitutive identity.A structured post-accident and historical-data process for identifying causal errors and converting findings into preventive aviation safety action. The parent is defined more broadly: A defeasible, provenance-bearing relation between an observable trace and a hypothesis about an unobservable state.
- Brodie–Trendelenburg percussion test Domain-specific is a decomposition of Evidence
The accepted reference-grade review places Brodie–Trendelenburg percussion test under Evidence because the child instantiates or depends on the parent's broader structure while retaining its own constitutive identity.A bedside percussion maneuver that infers superficial venous valve incompetence when a proximally generated impulse is transmitted to a distal palpating finger. The parent is defined more broadly: A defeasible, provenance-bearing relation between an observable trace and a hypothesis about an unobservable state.
Hierarchy paths (4) — routes to 4 parentless roots
- Evidence → Provenance → Traceability → Observability
- Evidence → Provenance → Attestation → Authentication
- Evidence → Provenance → Traceability → Transformation → Function (Mapping)
- Evidence → Provenance → Custody Transfer → State and State Transition → Phase Space
Neighborhood in Abstraction Space¶
Evidence sits among the more crowded primes in the catalog (18th percentile for distinctiveness): several abstractions describe nearly the same structure, so a description that fits it will tend to fit its neighbors too — transporting it usually means disambiguating within this family rather than landing on it exactly.
Family — Accumulation, Decay & Maintenance at Interfaces (16 primes)
Nearest neighbors
- Cross-Level Inference — 0.74
- Proof By Contradiction — 0.74
- Provenance — 0.74
- Falsifiability — 0.74
- Bijectivity — 0.73
Computed from structural-signature embeddings · 2026-09-10
Not to Be Confused With¶
The nearest neighbor (similarity 0.93) is provenance, and the relation is part/whole: provenance is one component of the evidence relation, not a synonym for it. Provenance is the chain of custody — the traceable path from the underlying event to the trace to its presentation — and it is the handle that defeats an inference when broken (a skewed clock, an inlined build, an unverified source). But evidence is the full role-set: target hypothesis, trace-producing mechanism, trace, inference rule, aggregation rule, and defeaters, of which provenance is just the custody sub-relation. A trace can have impeccable provenance and still be weak evidence (it bears little on the hypothesis), or strong content yet be defeated by a custody break. Conflating the two collapses the separable "how strongly does this bear on the claim?" question into the "where did this come from?" question, when the prime's whole value is keeping the handles distinct so a dispute can localize to one of them. Provenance answers lineage; evidence answers the entire trace-to-hypothesis inference, lineage included.
A second genuine confusion is with the inductive_reasoning / abductive_reasoning / bayesian_updating family — the inference rules that operate on evidence. These name how one moves from traces to graded confidence: induction generalizes from instances, abduction infers the best explanation, Bayesian updating multiplies likelihood ratios into a posterior. Evidence is not any of these rules; it is the relation between trace and hypothesis that the rules consume and combine. The distinction matters because the same evidence can be processed by different rules (legal balancing, Dempster-Shafer weighing, Bayesian updating, narrative coherence) yielding different combination behavior, and because errors localize differently: a wrong inference rule is a different defect from a defeated trace or a broken custody chain. Treating evidence as identical to Bayesian updating, in particular, over-narrows the prime — evidence is broader than any one aggregation function, and the role-set survives even where practitioners lack the Bayesian vocabulary. The rule is the operation; evidence is the operand.
A third confusion is with belief_formation. Belief is the doxastic output — the state of holding a hypothesis as true; evidence is the input that licenses graded confidence in it. The two are routinely run together ("the evidence" treated as identical to "what we believe"), but they dissociate in both directions: one can possess strong evidence and withhold belief (suspending judgment pending a defeater check), or hold a belief with no evidence at all (faith, bias, motivated reasoning). The prime insists that evidence only ever licenses weight, never compels belief, and that crossing from weight to a held conclusion requires a separate, often implicit, threshold choice. Confusing evidence with belief hides that threshold — the most consequential and most contestable step — by treating the trace's weight as if it automatically settled the doxastic question. Evidence shifts confidence; belief is the state confidence may or may not be allowed to produce.
For a practitioner these distinctions make inference auditable. Confusing evidence with provenance reduces a multi-handle relation to a custody check, missing weak-but-clean and strong-but-defeated traces. Confusing it with the inference rules conflates the operand with the operation, mislocating defects. Confusing it with belief hides the standard-of-proof choice that turns graded weight into action. The unifying discipline is the prime's handle-by-handle audit: name the hypothesis, the trace-producing mechanism, the custody chain, the inference and combination rules, the live defeaters, and the threshold — and contest each separately rather than arguing about "the evidence" as an undifferentiated whole.
Solution Archetypes¶
Solution archetypes in the catalog that build on this prime — directly (this prime is a source ingredient) or as a related prime.
Built directly on this prime (19)
- Abductive Explanation Selection: Turn a surprising observation into a ranked, provisional best explanation, while keeping rivals, uncertainty, and revision triggers visible.▸ Mechanisms (8)
- Abduction Log — A running record of explanation changes, retired rivals, new evidence, and revision triggers.
- Anomaly-to-Hypothesis Workshop — A facilitated session that converts anomalies into candidate hypotheses while preserving dissent and uncertainty.
- Differential Diagnosis Workup — Lists candidate conditions or causes, weighs symptoms and tests, and narrows to a working diagnosis.
- Disconfirming Probe Plan — Specifies the observations or tests most likely to overturn the current best explanation.
- Explanatory Case Memo — Records the observation, candidate explanations, comparison rationale, confidence, and next probes for audit or handoff.
- Forensic Scenario Reconstruction — Builds and compares scenarios that could have produced observed traces while preserving uncertainty about alternatives.
- Inference-to-Best-Explanation Matrix — Compares candidate explanations against explanatory fit criteria and records the provisional winner.
- Model-Debugging Hypothesis Loop — Uses surprising model behavior to form, test, and revise explanations about data, architecture, prompts, or deployment context.
- Associative Transfer Warrant Audit: Do not let contact, co-membership, resemblance, endorsement, or proximity carry trust, blame, risk, quality, or credibility unless the link has a valid transfer warrant.▸ Mechanisms (8)
- Association-to-Evidence Matrix — Lists each association, the property allegedly transferred, the warrant channel claimed, available evidence, counterevidence, and permitted decision use in one reviewable grid.
- Associative Claim Red Team — Challenges a proposed property transfer by asking what would have to be true for the association to be warrant-bearing and what evidence would refute it.
- Category-Membership Attribution Audit — Tests whether a property has been attributed to an individual or item merely because it belongs to a category, class, cluster, or population.
- Contact/Contagion Warrant Test — Distinguishes actual transmission or shared exposure from symbolic contact, mere proximity, or imagined contamination — and bounds any real channel by distance and decay.
- Endorsement Scope Checklist — Determines whether association with a person, brand, institution, funder, publisher, or platform is an endorsement, a neutral conduit, or an irrelevant co-location — and scopes what a real endorsement covers.
- Guilt-by-Association Review — Checks whether blame, risk, or disqualification is being assigned because of contact, affiliation, or co-membership without a valid transfer rule.
- Halo and Taint Decomposition Table — Separates positive halo, negative taint, aesthetic appeal, status borrowing, contamination fear, and evidence-bearing relations into distinct rated columns.
- Trust Transitivity Breakpoint Review — Identifies where trust, reputation, or certification stops transferring across a dependency or intermediary chain and renewed verification is required.
- Conditioned Probability Frame Specification: State what is being taken as given before interpreting, comparing, or acting on a probability.▸ Mechanisms (10)
- Base-Rate Check
- Conditional Probability Annotation — Attaches the conditioning frame to a probability value as a machine-readable label so the context travels with the number instead of being stripped downstream.
- Frame Compatibility Review — A gate run before two probabilities are compared or pooled, checking that their events, denominators, time windows, and sampling rules are actually commensurable.
- Given-That Clause — A sentence template that forces every stated probability to name its event and its 'given that' condition in the same breath, so no number is spoken naked.
- Likelihood-Ratio Frame — Separates the strength of the evidence from the probability of the hypothesis by expressing what a signal says as a ratio that updates a base rate rather than replaces it.
- Probability Tree — Draws sequential conditions as branching paths, multiplying along each branch so nested 'given that' steps stay in order and the denominator narrows one condition at a time.
- Reference Population Note — A short written note pinning exactly which population a rate is computed over, and where that denominator stops being valid, so the number can't drift onto a different base.
- Scenario Condition Card — One card per named scenario that fixes the full assumption set a forecast is conditioned on, so a scenario-conditioned probability can never be read as unconditional.
- Stratified Rate Table — Splits an aggregate rate into subgroup rows, each with its own denominator, so subgroup-conditioned probabilities are compared side by side without the marginal hiding them.
- Two-by-Two Probability Table — Lays two binary variables into four joint cells so P(A given B) and P(B given A) are computed from the same grid and can never be confused for each other.
- Construct–Proxy–Signal Validity Alignment: Make a measurement earn its interpretation by tracing the claim from construct to proxy to signal and requiring evidence that the signal captures the intended construct rather than a correlated surrogate.▸ Mechanisms (10)
- Cognitive Interview or Response-Process Probe — Watches respondents actually answer — thinking aloud — to check that the mental process generating the signal matches the construct, not a shortcut or a misreading.
- Construct Validity Argument — Assembles the reasoned case that a score deserves its interpretation — marshalling every strand of validity evidence into an explicit argument with a scoped claim and stated limits.
- Construct-to-Proxy Traceability Table — A row-per-claim table that traces each construct dimension down to the proxy and signal standing for it — and marks explicitly what each proxy leaves out.
- Content-Domain Review Panel — A panel of domain experts that fixes what the construct includes and excludes and judges whether the items representatively cover that domain — content validity by expert judgment.
- Factor-Structure or Latent-Model Check — Fits a latent-variable model to item responses to test whether their internal structure matches the construct's theorized dimensions — internal-structure evidence.
- Known-Groups or Contrast-Case Test — Checks that the measure separates groups already known to differ on the construct — and that the separation isn't explained by a confound the groups also differ on.
- Measurement Invariance Audit — Tests whether the measure means the same thing across subgroups — so a score gap reflects a real construct difference, not the instrument behaving differently by group.
- Multi-Trait Multi-Method Matrix — Crosses several traits with several measurement methods so convergent and discriminant validity can be read off — and method variance separated from true trait variance.
- Proxy Drift and Goodhart Audit — Periodically re-checks whether a proxy still tracks its construct once people are optimizing it — catching the moment a measure-turned-target decouples and needs revision.
- Validity Limitation Memo — A short written statement travelling with the measure that fixes what its scores may and may not be used to claim, for whom, and what harms to watch when it's used.
- Context-Preserved Meaning Capture: Record what happened together with the contextual cues, meanings, roles, and observer notes that make the event interpretable later.▸ Mechanisms (9)
- Analytic Memo — Captures the researcher's evolving interpretation in dated memos that keep observed evidence, participant meaning, and observer inference visibly separate.
- Artifact Metadata Capture — Attaches provenance and setting to a captured artifact so the object stays interpretable once it leaves the moment it documents.
- Codebook Loss Note — Records what a coding scheme discards when local categories are translated into external codes, so the loss travels with the coded data.
- Contextual Inquiry Note — Captures situated work by observing a person in their own environment and asking why in the moment, preserving the roles and reasoning behind what they do.
- Ethnographic Fieldnote Protocol — A disciplined routine for turning fleeting field observation into durable written record before memory flattens the scene.
- Interpretation Recovery Review — Hands the record to someone who was not there and checks whether they can reconstruct its meaning without inventing the missing context.
- Member Check Session — Takes provisional interpretations back to the people who were studied and asks whether the account rings true from the inside.
- Observation Vignette Template — A fixed template for rendering one observed scene as a compact standalone narrative organized around why it mattered.
- Scene-Setting Ledger — A structured checklist that inventories the setting of an observation — time, place, material environment, who was present, and what was notably absent.
- Deception Blowback Containment: When misleading signals are deliberately introduced, contain them with explicit audience boundaries, truth anchors, provenance markings, expiry rules, and re-entry monitors so the deception cannot boomerang into friendly decisions.▸ Mechanisms (12)
- After-Action Truth Reconciliation — After a bounded deception ends, sweeps up its residual artifacts, restores the record to ground truth, debriefs everyone it touched, and tallies what the operation actually cost.
- Audience-Channel Matrix — Maps every intended and unintended receiver of a planted signal across each channel, then scores where it is most likely to be exposed, re-enter friendly systems, or be trusted by one's own side.
- Bounded Correction Protocol — When a planted signal reaches friendly decision-makers, corrects or re-contextualizes it just enough to stop it steering their decisions — without disclosing more of the deception than necessary.
- Compartmented Briefing — Reads in only the friendly actors with a genuine need to know that a signal is planted — and how to recognize, route, or disregard it — so the deception stays effective while the few who must not be fooled are immunized.
- Contaminated Record Quarantine — Flags and isolates records known to carry planted or synthetic artifacts so they cannot flow into analytics, training data, audits, or official memory as if they were genuine.
- Deception Blowback Register — A standing ledger that records every authorized planted signal — its purpose, owner, audience, truth-anchor location, expiry, and cleanup duty — so nothing is deployed unrecorded and nothing is later mistaken for genuine.
- Friendly Reliance Probe — Samples the organization's own decisions, reports, and metrics for signs that friendly actors have started treating the planted signal as genuine evidence.
- Re-Entry Red-Team Review — Before a misleading signal is released, an independent reviewer war-games every way it could loop back through friendly systems and hardens the containment until those paths are closed.
- Sunset and Debrief Trigger — A pre-committed condition — an end date, an operation's close, or a risk threshold — that automatically fires cleanup: correction, debrief, declassification, deletion, or quarantine.
- Synthetic or Exercise Marker — Stamps every artificial, decoy, or exercise artifact with a durable, machine-readable label that says 'not real — exclude me' so authorized systems can recognize and filter it later.
- Training-Data Exclusion List — A standing denylist that stops marked synthetic or planted artifacts from being ingested into models, dashboards, search indexes, and decision-support datasets.
- Truth Anchor Memo — The authoritative internal record of what is real, what was planted to mislead, who is briefed, and which decisions must be made from the truth rather than the cover.
- Emic-Etic Dual-Account Interpretation: Preserve insider and outsider descriptions as separately governed accounts, then use their mismatch as evidence instead of forcing premature translation into one frame.▸ Mechanisms (10)
- Category Loss Annotation — Tags each mapping from a local term to an external category with the exact distinction that is dropped, so category flattening becomes a visible, typed mark rather than a silent default.
- Co-Interpretation Workshop
- Divergent Account Brief — A decision-facing brief that sorts every finding into emic, etic, convergent, or divergent, states what each may and may not license, and carries the unresolved residue forward instead of smoothing it into one story.
- Dual Codebook Audit — A periodic conformance check on two separately maintained codebooks — one emic, one etic — that verifies their closure rules stayed distinct and no local category was quietly laundered into an analytic one.
- Emic-Etic Contrast Matrix — A grid that places the same cases under both the emic and the etic account at once, so correspondences, partial matches, and outright breaks between the two become visible cell by cell.
- Etic Analytic Coding Frame — A structured, theory-derived codebook that builds the outsider account explicitly — naming the analytic framework, its constructs, and the sampling and competence conditions under which its codes may be applied.
- Guided Emic Interview Protocol
- Insider Review Panel — A standing panel of community insiders with real authority to validate the emic account, contest an outsider framing, and withhold consent for uses that would expose local knowledge to harm.
- Member Checking Session — A working session that takes draft interpretations back to the very participants who supplied the data, so they can confirm, correct, or contest whether the analyst's reading still means what they meant.
- Reflexive Positionality Memo — A first-person memo in which the analyst records who they are relative to the phenomenon — their standpoint, stakes, and power — and traces how that position shapes what each account can and cannot see.
- Evidence-Bound Authentication: Grant trust, access, or evidential weight only after an asserted identity or origin is bound to admissible evidence and returned as a scoped authentication verdict.▸ Mechanisms (12)
- Authentication Broker — Sits between clients and the capability, verifies who is asking, and issues a scoped, short-lived credential that grants exactly the access the request needs — and no more.
- Certificate Chain Validation — Establishes that a certificate's key really belongs to the named subject by walking the signed chain from leaf to a trusted root, enforcing each hop's validity, scope, and revocation, and returning a graded verdict.
- Chain-of-Custody Evidence Review — Authenticates an artifact by reconstructing its unbroken, documented custody trail — proving the thing in hand is the same one collected at origin, handled intact and untampered.
- Challenge-Response Authentication — Proves a claimant holds a secret by having them compute the correct answer to a fresh, unpredictable challenge — so possession is demonstrated without the secret crossing the wire, and a captured exchange can't be replayed.
- Credential Verification Workflow — Runs an asserted eligibility or identity claim through a defined sequence of primary-source evidence checks — with a staffed exception lane for the cases the rules can't settle — before a protected offer or access is granted.
- Digital Signature Verification — Confirms that a specific message was signed by the holder of a given key and hasn't changed since — binding content to its signer's key, with tamper-evidence built in, while deliberately leaving key-to-identity trust to a separate step.
- Federated Identity Assertion — Lets one system grant access on the strength of a signed identity assertion from a trusted external identity provider — accepting the provider's authentication verdict instead of re-authenticating the user itself.
- Liveness or Presence Check — Proves a real, live, present subject is producing the evidence right now — so a photo, recording, mask, or deepfake cannot stand in for a genuine presence.
- Multi-Factor Authentication — Requires evidence from two or more independent categories — something you know, have, and are — so compromising any single factor is not enough to pass.
- Provenance Chain Review — Authenticates an artifact by tracing its origin and every handoff — accepting it only when the lineage back to a trusted source is complete and unbroken.
- Revocation Status Check — Checks whether an otherwise-valid credential has been revoked, expired, or pushed out of scope since it was issued — so trust is withdrawn the moment its basis changes.
- Zero-Knowledge Authentication Protocol — Proves possession of a secret, or the truth of a claim, while revealing nothing beyond the claim itself — authentication without handing over the underlying data.
- Evidentiary Trace Warranting: Treat evidence as a defeasible relation between a trace and a claim, not as raw data or free-floating support.▸ Mechanisms (9)
- Admissibility or Relevance Gate — Prevents traces below provenance, quality, or relevance thresholds from being used in high-stakes reasoning.
- Claim-Evidence-Reasoning Card — Captures a compact claim, evidence trace, and reasoning bridge for review.
- Defeater Register — Lists facts, conditions, and counterevidence that would weaken or defeat an evidence relation.
- Evidence Provenance Log
- Evidence Relation Matrix — Maps traces to claims with provenance, relevance, link type, weight, scope, and defeaters.
- Evidence Strength Ladder — Labels evidence strength while preserving scope, uncertainty, and defeasibility.
- Evidence Update Review — Revisits evidence relations when sources, context, measurement, or rival explanations change.
- Relevance and Alternative Explanation Check — Tests whether a trace actually discriminates among hypotheses or is also expected under alternatives.
- Trace-to-Claim Diagram — Visualizes trace, source, inference bridge, hypothesis, and defeaters.
- Expected-Absence Signal Interpretation: Treat a missing expected event as evidence only after verifying that it was expected, observable, producible, timely, and unlikely to be missing for benign reasons.▸ Mechanisms (9)
- Absence Likelihood Dashboard — Tracks missed-event rates, latency distributions, false absences, confirmed failures, and response outcomes so silence has a measured base rate instead of a gut feeling.
- Confirmation Probe Request — Sends a low-cost, bounded follow-up before treating absence as strong evidence or triggering severe action.
- Detection Opportunity Audit — Checks whether the observer, sensor, search, or communication channel actually could have detected the expected event.
- Exception-Lag Review Workflow — Reviews recurring benign lags and exceptions so thresholds and calendars stay realistic instead of firing on ordinary delay.
- Expected Event Register — Lists expected events, due windows, owners, channels, exception conditions, and interpretation rules as the shared source of truth for what should have happened.
- Missing Heartbeat Monitor — Detects missed keepalives, check-ins, reports, or scheduled signals and routes them through false-absence checks before declaring failure.
- No-Response Escalation Protocol — Defines when and how to follow up, reroute, escalate, or pause after an expected response does not arrive.
- Null-Result Power Check — Estimates whether a failed search or null observation had enough sensitivity to count as evidence of absence.
- Silence Signal Review Board — Reviews high-stakes interpretations of silence or nonresponse where power, consent, safety, or exclusion risks are present.
- Independent Convergence Evidence Appraisal: Treat repeated independent arrival at the same solution-shape as evidence of fit only after auditing independence, shared pressures, abstraction level, and alternative explanations for the convergence.▸ Mechanisms (7)
- Blind Pattern Comparison Round — Has independent judges rate whether candidate solution-shapes truly match, with lineage identities and the convergence claim masked, so the similarity score is not manufactured by expectation.
- Common-Cause and Copying Audit — Runs a fixed checklist of shared sources, incentives, regulations, vendors, and social proof to see whether any of them explains the repeated solution better than independent fit.
- Convergence Evidence Matrix — Tabulates lineages, pressures, solution-shapes, independence evidence, performance, and caveats into one grid so convergence can be read across rows instead of asserted.
- Convergence Warrant Memo — Writes the final calibrated inference as a short defeasible memo — confidence, the scope it applies to, the action it implies, and the conditions that would overturn it.
- Lineage Traceback Protocol — Reconstructs the origin and transmission history of each case to distinguish independent discovery from diffusion or copying.
- Negative-Case Scan — Actively hunts the comparable cases where the solution did not emerge or did not work, and lets those failures recalibrate how strong the convergence really is.
- Pressure–Solution Fit Rubric — Scores, case by case, how well the specific solution-shape answers the specific pressure — the mechanism story that says why this form solves this problem.
- Informal Fallacy Diagnosis and Repair: Repair arguments that can look formally valid but fail because their premises, context, relevance, or category moves are defective.▸ Mechanisms (10)
- Burden Shift Detection — Catches when the obligation to prove quietly moves from the claimant to the critic, so 'you can't disprove it' stops passing as evidence.
- Category Boundary Stability Check — Holds a key term or category to one fixed meaning across the whole argument, catching definitions that drift to dodge counterexamples.
- Counterexample Probe — Attacks a general claim by manufacturing the case that would make it false, then reports whether the claim survives, narrows, or breaks.
- Fallacy Label Justification Note — Ties a named fallacy to the exact defective inference move and states it as a reasoning correction rather than a stigmatizing label.
- Fallacy Triage Matrix — Sorts a pile of suspected reasoning defects by fallacy family and by how much each actually damages the inference, then routes each one.
- Feasible Alternative Comparison — Scores a real option against the best achievable alternative instead of an idealized one, exposing rejections driven by the nirvana fallacy.
- Formal vs Material Error Gate — Decides whether an argument's defect lives in its logical form or in its content, so formal invalidity is not mislabeled an informal fallacy.
- Option Space Expansion Prompt — Breaks a false binary by regenerating the live options a two-way framing suppressed.
- Premise Relevance & Acceptability Test — Screens each premise for whether it is acceptable in its own right and actually relevant to the conclusion it is offered to support.
- Steelman then Refute Protocol — Rebuilds an argument in its strongest fair form before attacking it, so any refutation lands on the position itself and not a caricature.
- Intrinsic Signature Provenance: Preserve or read an intrinsic, stable origin signature so provenance travels with the thing itself, even when external records are missing or distrusted.▸ Mechanisms (10)
- Blind Proficiency Test — Feeds a laboratory known-origin samples disguised as ordinary casework to measure — blind — how often its whole attribution pipeline gets the source right.
- Chemical Taggant Program — Embeds a covert, coded chemical marker into a product at manufacture so its batch and maker can be decoded after the fact — and can't be easily counterfeited.
- Digital Watermark or Content Fingerprint — Makes a digital file self-identifying — by hiding an imperceptible mark inside it or deriving a robust fingerprint from it — so its source survives copying, compression, and re-encoding.
- DNA or Biological Barcode — Reads an organism's own standardized DNA region to attribute a biological sample to a species or population, matched against a reference barcode library.
- Isotopic Fingerprint Analysis — Measures the stable-isotope ratios carried in a material to place its origin in the geography and geology those ratios record.
- Likelihood-Ratio Attribution Report — Turns a signature comparison into a calibrated likelihood ratio — how much more the evidence favors one origin than a stated alternative — with scope and limits attached.
- Manufacturing Toolmark Analysis — Reads the microscopic marks a tool or machine imprints on what it makes or touches, matching an object back to the individual tool that shaped it.
- Reference Library Match — Looks a query signature up against a governed library of known-origin references and returns scored candidate matches — only as trustworthy as the library is current and representative.
- Spectral Signature Matching — Measures a material's full spectrum and matches its shape against a reference spectral library to identify what it is — and thereby where or when it could have come from.
- Trace-Element Profile Matching — Fuses the concentrations of many trace elements into one multivariate profile and matches it to a specific source deposit or batch.
- Procedural Objectivity Warranting: Make a public claim objective by licensing it through separated verification, traceable evidence, calibrated sourcing, disciplined framing, and accountable correction rather than through the preferences of interested parties.▸ Mechanisms (10)
- Adversarial Editor or Red-Team Review — Assigns a reviewer whose job is to attack the claim — hunting the leaked interest, the buried counter-evidence, and the framing that makes a preferred conclusion look inevitable — before it reaches the audience.
- Affected-Party Right-of-Response Workflow — Gives the people a claim is about a bounded, documented window to contest facts and add context — without granting them veto, preclearance, or control over the framing.
- Attribution and Sourcing Standard — The house rules for naming and characterizing sources — when a source may be unnamed, how their proximity and incentive must be described, and what the audience is told about how the information was obtained.
- Blind or Masked Review Path
- Conflict-of-Interest Screening and Recusal
- Correction Policy and Change Log — A standing rule and public record that any error found after publication is fixed visibly, dated, and explained — with the affected claim restated and its status downgraded rather than quietly overwritten.
- Evidence-to-Claim Matrix — 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.
- Fact-Checking Checklist — A pre-publication pass that walks every checkable assertion back to a source that supports it, flags what cannot be verified, and separates verified fact from interpretation and opinion.
- Headline and Lead Claim Consistency Review — Checks that the headline and opening lines assert no more than the body can warrant — pulling back verbs, certainty, and scope until a skimming reader's takeaway matches the evidence.
- Source Triangulation Matrix — Arrays each claim against its sources to test whether apparent corroboration is genuinely independent or just one interested origin echoed — and whether the source mix is balanced enough to trust.
- Process-Imprint Source Attribution: Use stable, involuntary marks left by a production process to infer where an output came from, with controls for confounders, spoofing, and over-attribution.▸ Mechanisms (10)
- Chain-of-Custody Cross-Check — Compares process-imprint attribution against explicit provenance or custody evidence without conflating them.
- Chemical & Isotopic Signature Test — Uses material composition or isotopic ratios as an origin-imprint evidence channel.
- Manufacturing Batch Trace Analysis — Links outputs to production batches using repeated defects, residues, material composition, or tolerance profiles.
- Model-Output Signature Probe — Tests whether a model, generator, or pipeline leaves recurrent statistical artifacts.
- Negative-Control Signature Panel — Challenges candidate marks against non-source exemplars and shared-process controls.
- Sensor Fingerprint Analysis — Detects device-specific noise, calibration, dead-pixel, acoustic, or timing patterns.
- Signature Likelihood Report — Documents features, exemplars, controls, confidence language, alternative sources, and limits.
- Spoofing & Counter-Forensic Challenge — Attempts to imitate, suppress, transfer, or plant signature features before accepting attribution.
- Stylometric Attribution Model — Estimates source likelihood from stable linguistic, formatting, rhythm, or choice-pattern features.
- Toolmark Comparison Protocol — Compares striations, impressions, wear marks, or tooling defects across known and disputed artifacts.
- Salience-Significance Decoupling: Separate what got attention from what deserves weight.▸ Mechanisms (12)
- Attention-Capture Inference Test — Traces why an item captured attention — which channel, design, or sponsor made it prominent — and tests whether that reason has anything to do with why it would matter.
- Base-Rate Visibility Panel — Places the base rate and its denominator beside a vivid instance, so a striking case cannot be read as representative.
- Counterexample Surface Scan — Deliberately hunts the disconfirming cases a vivid story leaves unshown, so the counterexamples get weighed too.
- Dashboard Salience Calibration — Re-tunes a dashboard so visual prominence tracks significance rather than default, vendor, or recency — and publishes a key so viewers can tell the difference.
- Display Reason Label — Tags each shown item with the reason it is shown — sponsored, recommended, trending — so viewers can discount prominence that comes from the channel rather than importance.
- Evidence Weighting Rubric — Scores evidence against explicit significance criteria fixed before the evidence is seen, so vividness cannot smuggle in weight it has not earned.
- Notification Priority Review — Re-examines an alerting system so that what pages a human is governed by significance and escalation criteria, not by how loud or how often an alert happens to fire.
- Ranking Semantics Legend — A published key that states what a ranking's order actually means — the sort key behind it — so 'at the top' is never quietly read as 'most important.'
- Salience Red Team — A standing adversarial group chartered to ask what the loudest items are crowding out and who engineered their prominence.
- Salience-Significance Matrix — Scores each item twice — how much attention it grabs and how much it actually matters — so the loud-but-trivial and the quiet-but-critical sort into different corners.
- Sample Frame Reconstruction — Rebuilds the population and the selection filter a visible sample was drawn through, so 'the cases I can see' stops standing in for 'the cases that matter.'
- Shown-vs-Unshown Audit — Sets a display's visible items beside the relevant ones it leaves out, so the gap between what is shown and the full field becomes something you have to look at.
- Source Distortion Modeling: Treat a report from a systematically distorted source as a biased channel to be modeled, not as either transparent truth or useless noise.▸ Mechanisms (8)
- Account/Event Reconstruction Table — Separates what the source says from the reconstructed event sequence, inferred omissions, and uncertain intervals.
- Claim Release Gate — Prevents downstream publication, decision, or automation until claims from a distorted account have scoped confidence and corroboration.
- Contradiction Timeline — Places inconsistent statements, records, and observed events on a timeline to distinguish memory, framing, drift, and strategic revision.
- Corroboration Ladder — Orders independent traces from weak consistency checks to strong external confirmation and contradiction.
- Distortion Model Card — Documents the assumed distortion pattern, supporting evidence, scope, counterevidence, and expiry conditions.
- Motive-Opportunity-Bias Analysis — Checks whether a proposed distortion pattern is plausible given the source's incentives, opportunity, and known bias profile.
- Narrator Reliability Matrix — Scores or describes reliability by claim type, evidence base, motive, vantage, consistency, and corroboration.
- Vantage-Bias Interview Protocol — Elicits what the source could know, why they framed it as they did, and what pressures shaped the account.
- Textual Close-Reading Mode: Suspend premature paraphrase, inspect the artifact at its meaningful grain, and bind every larger interpretation to exact marks, relations, repetitions, placements, and omissions.▸ Mechanisms (13)
- Close Reading Table — Slows a reading to the level of the single mark — annotating each word, image detail, or gesture for what it does and the convention it invokes — before any larger framework is imported.
- Contextual Reentry Memo — Reintroduces genre, period, audience, authority, and use after the grain has been inspected — so context meets an artifact-grounded reading instead of quietly deciding it in advance.
- Counterreading Stress Review — Builds the strongest rival reading on purpose and makes the favored one either explain the marks it resists or narrow its own claim.
- Diction, Syntax, and Punctuation Trace — Follows the load-bearing formal features — word choice, grammatical attachment, modality, negation, punctuation, and scope — wherever a small difference could change duty, agency, timing, or certainty.
- Edition or Variant Collation Table — Aligns the versions, translations, amendments, and witnesses side by side so a reading can't mistake version-contingent wording for a stable property of the text.
- Interpretive Saturation Log — Tracks whether each further pass still changes a scoped claim, and calls the reading closed — with its open questions logged — once the passes stop paying off.
- Line-by-Line Annotation Protocol — Walks a bounded passage one line at a time, recording every material mark, observation, and question in sequence — without letting any of them collapse into a summary yet.
- Micro-to-Macro Claim Matrix — For each consequential claim, lays out the ladder from located marks up through observation, pattern, and scope to implication, so a reviewer can accept or reject each rung independently.
- Motif Concordance and Recurrence Map — Indexes every occurrence of a chosen term or form and maps how its sense shifts from one appearance to the next, so recurrence is read as movement rather than counted as frequency.
- Omission and Negative-Evidence Audit — Tests whether something absent is meaningfully missing by first establishing why it should have been present, then ruling out innocent explanations for the gap.
- Parallel-Passage Comparison — Lines a passage up against comparable passages to learn whether a word, formula, or omission is conventional or exceptional before any weight is placed on it.
- Paraphrase-Difference Test — Holds a finished summary against the evidence map and flags what the compression dropped — modality, exception, speaker, ambiguity, sequence, or relation — before the summary is allowed to stand in for the source.
- Sequence and Placement Map — Charts the order, adjacency, nesting, and framing of a text's parts so that meaning carried by where something sits — not just what it says — becomes visible and arguable.
- Warranted Belief Formation: Turn a proposition into a responsible belief only after clarifying its meaning, warrant, confidence, scope, action consequences, and conditions for revision.▸ Mechanisms (8)
- Belief Adoption Checklist — A run-once pass/fail gate that blocks a claim from becoming an action-guiding belief until proposition, warrant, confidence, scope, action implication, and a revision trigger are all in hand.
- Belief Premise Register — A standing ledger of the propositions a decision currently rests on, each with its confidence and scope, an owner, an adoption state, and a recheck date.
- Bias and Pressure Prompt — A short self-administered set of questions that surfaces the non-warrant forces — fluency, fear, authority, identity, incentive — that may be doing the real work behind a belief, and asks who is harmed if it is wrong.
- Claim-Warrant Matrix — A side-by-side grid that puts many competing claims on rows and their evidence, source quality, and counter-evidence on columns, so warrant can be compared across rivals before any one is believed.
- Confidence and Scope Label Template — A fixed grammar for tagging a belief with a calibrated confidence level and the bounded conditions under which it holds, so the caveat travels with the claim.
- Doxastic Commitment Ladder — A named ladder of graded belief states — from heard, to plausible, to provisional, to action-guiding — with a confidence band and an action license for each rung.
- Falsification Trigger Card — A one-belief tripwire sheet naming the specific observations that would weaken, suspend, or overturn it, who watches for them, and when to look again.
- Reflective Belief Dialogue — A structured multi-person conversation in which a peer actively challenges why a claim should be believed, what would change it, and what ethical limits bound acting on it.
Also a related prime in 24 archetypes
- Adaptive Precision-Weighted Signal Fusion: Combine imperfect signals by how reliable they are now, not by treating every input as equal or permanently trustworthy.
- Context-Bounded Meaning Recovery: Make interpretation accountable by explicitly binding a reading to a substrate, a context, a framework, evidence marks, and a boundary around plausible alternatives.
- Data-Control Boundary Inertization: Keep untrusted content inert until a structural boundary, validation rule, and authority gate explicitly permit it to become control.
- Dispute-Question Alignment: Stop arguing over answers until the parties have identified which kind of question they are actually contesting.
- Editorial Independence Firewall: Protect the evaluator’s judgment from affected-party control by separating authority, incentives, access, correction rights, and accountability channels.
- Enacted-Control Verification and Closure: Verify controls as enacted, not merely as documented, and close the gap when paper controls and real operating practice diverge.
- Independent Convergence Recognition and Transfer Design: Use independently repeated solutions as evidence of shared pressures or constraints while checking that the repetition is not copying, common ancestry, or false similarity.
- Latent Constraint Preservation Audit: Treat a persistent structure as possible evidence of a hidden constraint: understand its function, dependencies, and failure-prevention role before removing or simplifying it.
- Minimal-Disclosure Verification: Make a verifier confident that a bounded claim is true without handing over the underlying witness, record, identity attributes, or computation trace.
- Noise-Bounded Measurement Interpretation: Treat every measurement as a noisy observation with a bounded claim, not as a direct copy of reality.
References¶
[1] United States. Federal Rules of Evidence. 2023 edition. Codifies relevance (probative value), admissibility, and authentication (chain of custody), operationalizing the structural commitments of the evidence relation in law. registry ↩
[2] Shadish, William R., Thomas D. Cook, and Donald T. Campbell. Experimental and Quasi-Experimental Designs for Generalized Causal Inference. Houghton Mifflin, 2002. Treats the strength of causal evidence as governed by study design and the systematic ruling-out of threats to validity (confounders) — the trace-producing-mechanism and defeater roles in science. registry ↩
[3] Sackett, David L., Sharon E. Straus, W. Scott Richardson, William Rosenberg, and R. Brian Haynes. Evidence-Based Medicine: How to Practice and Teach EBM. 2nd ed. Churchill Livingstone, 2000. Treats signs and tests as evidence updating diagnostic probability via likelihood ratios and Bayes' rule, with sensitivity/specificity and the base-rate effect (a positive test on a rare disease yielding a low posterior because most positives are false positives). registry ↩a ↩b ↩c ↩d
[4] Howell, Martha C., and Walter Prevenier. From Reliable Sources: An Introduction to Historical Methods. Cornell University Press, 2001. Sets out the historian's source criticism — provenance/authentication, internal versus external criticism, and triangulation across independent sources — the chain-of-custody and combination roles in history. registry ↩
[5] Zeller, Andreas. Why Programs Fail: A Guide to Systematic Debugging. 2nd ed. Morgan Kaufmann, 2009. Treats debugging as scientific inference from observable traces (logs, stack traces, reproducers) to a hypothesis about a defect, with delta-debugging/bisection as evidence-production. registry ↩a ↩b
[6] Renfrew, Colin, and Paul Bahn. Archaeology: Theories, Methods and Practice. Thames & Hudson. Treats inference from physical remains and stratigraphic context, with formation processes and taphonomy supplying the trace-producing-mechanism analysis for archaeological evidence. registry ↩
[7] Shafer, Glenn. A Mathematical Theory of Evidence. Princeton University Press, 1976. Develops Dempster–Shafer belief functions as a non-Bayesian aggregation rule over evidence, illustrating that the evidence role-set admits different combination functions. registry ↩
[8] Bayes, Thomas. "An Essay towards Solving a Problem in the Doctrine of Chances". Philosophical Transactions of the Royal Society of London, vol. 53 (1763): 370–418. Original statement of the inverse-probability rule underlying Bayesian updating of a hypothesis on observed evidence. registry ↩
[9] Klein, Gary. Sources of Power: How People Make Decisions. MIT Press, 1998. Treats pre-mortem and adversarial review as disciplines for surfacing defeaters before acting, the engineering/organizational analogue of cross-examination. registry ↩