P191 — AIEP — Evidence Temporal Alignment Protocol

Publication Date: 2026-03-27 Status: Open Source Prior Art Disclosure Licence: Apache License 2.0 Author/Organisation: Phatfella Ltd Schema: AIEP_OS_SPEC_TEMPLATE v1.0.1 — https://aiep.dev/schemas/aiep-os-spec-template/v1.0.1

Framework Context

[0001] This disclosure operates within an Architected Instruction and Evidence Protocol (AIEP) environment as defined in United Kingdom patent application number GB2519711.2, filed 20 November 2025, the entire contents of which are incorporated herein by reference.

[0002] The present disclosure defines a protocol for aligning evidence artefacts referencing heterogeneous temporal contexts — resolving discrepancies between evidence publication date, evidence observation date, evidence effective date, and reasoning query date — into a unified TemporalAlignmentRecord that enables a reasoning chain to correctly interpret the temporal relevance of each artefact relative to the facts in question.

Field of the Disclosure

[0003] This disclosure relates to temporal metadata alignment and reasoning protocols for evidence artefacts in artificial intelligence evidence systems.

[0004] More particularly, the disclosure concerns: a TemporalAlignmentRecord schema; a taxonomy of temporal date types; temporal alignment computation; temporal relevance scoring as a complement to freshness scoring (P147); temporal gap detection for a reasoning query’s target time window; and integration with the evidence taxonomy (P160) and reasoning-ready packages (P174).

Background

[0005] Evidence artefacts carry multiple temporal coordinates that are distinct and must not be conflated. A study published in 2024 may report observations collected in 2019: the publication date governs freshness decay (P147) but the observation date governs whether the evidence is relevant to a query about the state of the world in 2020. A regulation published in 2018 may have an effective date of 2019 and may have been superseded in 2023: a query about the regulatory position in 2021 requires the 2018 regulation but not the 2023 successor.

[0006] Standard freshness scoring (P147) uses ingestion date and publication date to compute temporal currency relative to the current query time. Temporal alignment is orthogonal: it maps evidence observation windows to query target windows, enabling reasoning chains to surface the evidence that was observationally relevant at the time the query question was true — not merely the most recently published evidence.

Summary of the Disclosure

[0007] TemporalAlignmentRecord Schema: A temporal metadata record attached to each evidence artefact:

• temporal_record_id — SHA-256 of canonical serialisation
• subject_deid — DEID (P162)
• publication_date — the date on which the artefact was publicly published (ISO 8601 date)
• observation_window_start — the start of the period during which the observations or data in the artefact were collected (ISO 8601 date; null if the artefact does not claim a bounded observation window)
• observation_window_end — the end of the observation window (ISO 8601 date)
• effective_date — for regulatory or policy artefacts: the date from which the artefact became operative (ISO 8601 date; null for non-regulatory artefacts)
• superseded_date — the date on which a successor artefact made this artefact no longer operative (ISO 8601 date; null if still operative)
• temporal_scope_type — one of: POINT_IN_TIME (artefact reports state at a single date), OBSERVATION_WINDOW (artefact reports observations collected over a window), PROSPECTIVE (artefact makes forward-looking claims or projections), REGULATORY_EFFECTIVE_PERIOD (artefact was operative from effective_date to superseded_date)
• temporal_confidence — confidence in the accuracy of the temporal metadata fields: HIGH (explicitly stated in the artefact), MODERATE (inferred from context), LOW (estimated by the tagging process)
• computed_by — node fingerprint (P46)

[0008] Temporal Relevance Score: For a reasoning query with a specified query_target_date, the Temporal Alignment module computes a temporal_relevance_score in [0.0, 1.0] for each artefact:

• POINT_IN_TIME artefacts: score decays with |query_target_date - observation_window_start| using a configurable half-life (default: 365 days — an artefact observing conditions 365 days before or after the query date scores ~0.5)
• OBSERVATION_WINDOW artefacts: score = 1.0 if query_target_date ∈ [observation_window_start, observation_window_end]; decays as the query date moves outside the window
• REGULATORY_EFFECTIVE_PERIOD artefacts: score = 1.0 if query_target_date ∈ [effective_date, superseded_date]; 0.0 otherwise — regulatory artefacts are binary (either operative or not at the query date)
• PROSPECTIVE artefacts: score based on whether query_target_date > publication_date and within the prospective horizon stated in the artefact

[0009] Temporal Gap Detection: For a reasoning query specifying a query_target_date and a taxonomy scope, the Temporal Alignment module identifies temporal coverage gaps — date ranges within the query window for which no artefact provides relevant temporal coverage. Gap detection proceeds by: (a) computing temporal_relevance_score for all artefacts in scope; (b) identifying the sub-periods of the query window not covered by any artefact with temporal_relevance_score ≥ 0.5; and (c) generating TemporalGapRecord entries identifying those uncovered sub-periods.

[0010] Integration with Evidence Packages: EvidencePackage AssemblyRequests (P174) may include a query_target_date parameter, causing the PackageAssembler to include each artefact’s temporal_relevance_score (for that target date) in the EvidencePackageItem alongside the standard freshness and trust scores.

[0011] Temporal Alignment Inference: Where an artefact’s observation_window_start or effective_date is not explicitly stated in the artefact metadata, the Temporal Alignment tagging process infers these dates from: (a) explicit date references in the normalised content (P10); (b) the CitationRecord (P157) publication year; (c) domain-specific conventions (e.g. for regulatory artefacts, gazette publication date is typically the effective date unless stated otherwise). Inferred dates receive temporal_confidence: MODERATE or LOW.

ASCII Architecture

Evidence Artefact
        │ temporal metadata extraction
        ▼
TemporalAlignmentRecord
(publication_date, observation_window, effective_date,
 superseded_date, temporal_scope_type, temporal_confidence)
        │
        ▼ query time: given query_target_date
Temporal Relevance Scoring
  POINT_IN_TIME: decay from observation date
  OBSERVATION_WINDOW: 1.0 if in window
  REGULATORY: 1.0 if operative at date
  PROSPECTIVE: decay beyond publication
        │
        ├──▶ Per-artefact temporal_relevance_score
        │    → EvidencePackageItem (P174, with query_target_date)
        │
        └──▶ Temporal Gap Detection
             identify sub-periods with no artefact coverage
             → TemporalGapRecord
             → Reasoning chain: "temporal coverage gap at {period}"

Operational Detail

[0012] Temporal Alignment vs. Freshness: Freshness scoring (P147) addresses the question: “How stale is this artefact relative to today?” Temporal alignment addresses: “Is this artefact relevant to the specific time period the reasoning query is about?” Both are needed. A 10-year-old artefact may be stale (low freshness score) but highly temporally relevant (high temporal relevance score) if the query target date is 10 years ago.

[0013] Temporal Uncertainty Propagation: Where temporal_confidence: LOW, the temporal_relevance_score is attenuated: adjusted_score = temporal_relevance_score × temporal_confidence_weight where temporal_confidence_weight is 1.0 for HIGH, 0.8 for MODERATE, and 0.6 for LOW. This propagates temporal metadata uncertainty into the relevance score received by reasoning chains.

[0014] Regulatory Supersession Cascade: When a regulatory artefact’s superseded_date is set (because its successor has been admitted to the corpus), the Watch Protocol (P180) notifies any subscriber watching that DEID with event type ARTEFACT_UPDATED. Reasoning chains subscribed to regulatory artefacts in their domain are thereby immediately notified of the operative-period closure.

Claims-Exclusion Notice

This specification is published as open-source prior art. No patent claims are asserted by the author in respect of the mechanisms described. Any third party seeking to patent mechanisms substantially equivalent to those described herein is placed on notice of this prior art disclosure.