Skip to content

Use Time Source Attribution Calibration

Before using a commingled memory, note, claim, trace, or generated output, classify where it came from and how certain that attribution is.

Disposition check

The target prime reality_monitoring was checked against accepted archetypes, aliases, components, mechanisms, reconciliation maps, and prior outputs through queue position 40. The closest accepted neighbors are useful boundaries rather than collapse targets. canonical_classification covers stable category systems; source_provenance_triangulation evaluates external accounts; appearance_vs_reality_distinction_audit audits claim warrant about underlying reality; state_estimation infers hidden state; and cautious_pattern_completion marks inferred gap-filling. None of them directly covers the target prime's use-time discrimination of a commingled item as internally generated, externally perceived, reported, copied, inferred, reconstructed, generated, or unknown.

How to use this archetype

Use this archetype when a system stores mixed-origin information and later decisions depend on knowing what kind of source each item came from. The core move is not simply adding metadata. It is creating a gate at the moment of use that asks: what source class is this item being treated as, what cues support that attribution, how confident are we, and what may this item do downstream at that confidence level?

Practical pattern

A practical implementation starts with the source classes that matter in the domain. It then preserves or reconstructs cues, attaches confidence to the source attribution, separates source confidence from content truth, and escalates low-confidence or high-stakes items for corroboration. The system should explicitly allow an unknown source state, because forcing a label often creates more damage than preserving uncertainty.

Parameter dimensions

Important parameters include source-class granularity, cue capture cost, confidence scale, action threshold by stakes, corroboration burden, privacy limits, label visibility, decay tolerance, automation level, error review cadence, and appeal/review path. More granular source classes help only when users can apply them consistently. Stronger corroboration improves reliability but can slow work and suppress legitimate hypotheses if applied indiscriminately.

Invariants to preserve

Preserve the separation between source status and content truth. Preserve uncertainty when source cues are weak. Preserve a labeled route for internal generation, inference, simulation, and AI generation. Preserve credit and responsibility for external sources. Preserve stronger thresholds when source class affects high-stakes action.

Tradeoffs and failure modes

The main tradeoff is speed versus warrant. The archetype slows some uses in exchange for avoiding source-confusion failures. Common failures include fluency-as-source reasoning, provenance label theater, unknown-source normalization, false originality, cue decay through summaries, and automated source confidence overtrust.

Neighbor distinctions

Canonical Classification creates stable classes. Use-Time Source Attribution Calibration classifies the source class and confidence of a particular retrieved item.

Source Provenance Triangulation evaluates external accounts and sources. Use-Time Source Attribution Calibration also handles internally generated, reconstructed, copied, and generated items inside a mixed store.

Appearance vs. Reality Distinction Audit asks what is warranted about underlying reality. This archetype asks where the item came from and how that source attribution should govern use.

Cautious Pattern Completion helps fill missing content while marking inference. This archetype prevents completed material from later being remembered or cited as perception.

State Estimation infers the world state. This archetype infers the source class of an information item.

Examples

A clinician separates patient observation from later inference. An AI-assisted research system prevents generated text from becoming evidence. A journalist distinguishes primary document, interview, prior article, and editorial inference. A legal team distinguishes witness observation, hearsay, expert inference, and attorney reconstruction. A product team distinguishes observed behavior from internal assumption or generated summary.

Non-examples

A taxonomy of source types without use-time confidence and action thresholds is not enough. A provenance log that nobody consults at the moment of use is not enough. A generic fact-check that asks only whether a claim is true is not enough. A ban on imagination, simulation, or generated content is not this archetype; the point is correct labeling and evidential weighting.

Quality notes

The target prime is included as a direct source prime. Proposed-prime handling is controlled: this draft introduces no proposed primes. The pre-draft disposition check found close neighbors but no accepted archetype, alias, component, mechanism, reconciliation-map entry, or prior queue output that should absorb the target.

Common Mechanisms

  • Borrowed Idea Attribution Scan — Sweeps a shared store of notes and ideas for material that arrived from someone else but now feels self-generated, and routes each item back to the source that deserves the credit.
  • Chain-of-Custody or Lineage Check — Reconstructs an item's unbroken trail back to its origin — every handoff and transformation logged beside the content — so its source class is established rather than assumed when it is used.
  • Generated Content Disclosure Gate — Holds internally- or model-generated content at the point of release until it carries a label saying it was generated and is phrased so a downstream reader can weight it as such.
  • Hallucination Intrusion Triage — Takes items already flagged as possible fabrications or memory intrusions and sorts them by how much rides on them, quarantining, escalating, or releasing each before it is trusted.
  • Memory Source Probe — Interrogates one recalled item at the moment of recall for its source cues, then applies a rule to classify where it actually came from.
  • Observation Recheck or Replication — Converts a decayed or doubtful memory back into first-hand evidence by going and observing the thing again, instead of trusting the stored trace.
  • Provenance Lookup Before Publication — A last-gate check that, claim by claim, traces a draft back to where each piece actually came from and credits anything borrowed before it goes public.
  • Reality Monitoring Checklist — A short cue-by-cue checklist run at the moment of recall to decide whether an item was actually perceived from the world or generated inside your own head.
  • Source Attribution Confidence Rubric — A graded scale that scores how sure you are of an item's source — separately from whether the content is true — and trips a corroboration gate when the grade is low and the stakes are high.
  • Source Attribution Training Set — A curated corpus of real items whose true source class is already known, held as the gold reference that calibrates and teaches an attribution judgment — human or model.
  • Source Confusion Matrix Review — A retrospective review that tabulates which source classes get mistaken for which — reading the off-diagonal cells to find systematic, directional misattributions and feed the fixes back.
  • Source-Label Preserving Summary Template — A summary format that forces each condensed statement to carry its source class through compression, so shortening a document can't quietly flatten observed, reported, and generated content into equally-confident prose.

Compression statement

Use-Time Source Attribution Calibration applies when information items from different source classes share the same store, workspace, model, memory, or record system, but downstream action depends on whether a given item was perceived, inferred, imagined, copied, generated, observed, reported, or reconstructed. The archetype builds a use-time attribution gate: recover source-correlated cues, compare them against source-class criteria, attach confidence, preserve uncertainty, and route high-stakes or low-confidence items to corroboration before they are treated as evidence.

Canonical formula: commingled_item_store + source_cues + use_time_source_classification + confidence_annotation + corroboration_gate -> source_appropriate_weighting; fluent_trace - source_discrimination -> hallucination_intrusion_or_borrowed-source_misattribution

Abstractions this archetype builds on — directly (a source ingredient) or as a related pattern. Links follow the typed catalog namespace.

Built directly on (6)

  • Classification: Sorting entities into discrete categories by explicit rules, turning unbounded variation into a finite, reusable map for downstream reasoning and action.
  • Confidence Annotation: A graded warrant marker attached to a claim that travels with it and tells downstream consumers how much weight to place on it.
  • Production Signature: A production process involuntarily imprints stable regularities on its output, letting an analyst attribute the output to its source.
  • Provenance: A documented, traceable record of an entity's origin and successive custody transfers that establishes authenticity and assigns accountability by linking present state back to first known state.
  • Reality Monitoring: A system holding items from multiple source classes must, at use-time, attribute each item to its source — most basically internally-generated versus externally-perceived — with characteristic errors when the cues mislead.
  • Signature-Borne Provenance: A conservative property carries origin through transport, read back without a custody chain.

Also references 20 related abstractions

  • Calibration: Aligning a system's output to a trusted reference by measuring deviation, adjusting to reduce it, and monitoring for drift.
  • Correlated-Source Attribution Failure: When combined sources share underlying variation, joint inference stays strong while attribution to any individual source becomes unstable, sign-flipping, or arbitrary.
  • Data Integrity: Accuracy and consistency preserved.
  • Evidence: A defeasible, provenance-bearing relation between an observable trace and a hypothesis about an unobservable state.
  • Evidence-Fidelity Decay: Delay between event and capture lets backfill silently fuse observation, inference, and reconstruction into one uniform record.
  • Measurement Uncertainty and Observational Noise: Measurement noise arises from instrument and observation limits.
  • Memory Consolidation: Converting a newly encoded trace from a fragile, overwritable form into a durable, interference-resistant one through a slow post-encoding stabilization process.
  • Monitoring: Continuously observing a system's state to detect deviation from expected behavior and trigger a response, separating genuine signal from routine noise.
  • Observability: Infer internal state externally.
  • Pattern Recognition: Identify regularities.

Variants

Narrower or domain-specific specializations that share this archetype's core structure. Recognized variants are established; candidate variants are provisional.

Internal Content Externalization Guardrail · risk or failure variant · recognized

Prevents imagined, simulated, inferred, or model-completed material from being used as if it were externally observed evidence.

  • Distinct from parent: Narrower than the parent because it handles one directional error: internal-to-external misattribution.
  • Use when: Generated or internally reconstructed material enters evidence channels; Fluency or vividness can be mistaken for observation.

Borrowed Source Recognition Check · affective or cognitive variant · recognized

Detects when externally supplied ideas, phrases, or judgments are being experienced or presented as self-generated.

  • Distinct from parent: Narrower than the parent because the relevant source-class distinction is external-origin versus self-origin.
  • Use when: Authorship, originality, credit, or responsibility depends on recognizing an external source; Repeated exposure makes another source feel self-generated.

Generated Output Reality Monitoring · domain variant · recognized

Marks and checks AI-generated or automated content before it is treated as retrieved fact, human analysis, or external evidence.

  • Distinct from parent: Domain-specific to AI and automation contexts.
  • Use when: Model outputs are blended into notes, reports, decisions, or memory stores; Users may confuse generation, retrieval, and observation.

Reconstructed Record Source Check · temporal variant · recognized

Differentiates original observations from later backfill, summary, reconstruction, or interpretation when records are used.

  • Distinct from parent: Narrower because its key cue problem is time-delayed record reconstruction.
  • Use when: Delay or transformation has degraded evidence fidelity; Reports or records contain mixtures of observation and reconstruction.

Source-Unknown Holding Variant · implementation variant · recognized

Preserves a usable but non-evidence-grade holding state for items whose source class cannot yet be attributed confidently.

  • Distinct from parent: Implementation variant focused on the unknown-source state and action permissions.
  • Use when: Forcing a source label would create false certainty; The item may still be useful for hypothesis generation or exploration.

Near names: Reality Monitoring Design, Source Discrimination Gate, Internal/External Source Check, Provenance at Use Time.