P66 — AIEP — Probability Metadata Envelope
Publication Date: 2026-02-26 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
Field of the Invention
[0001] The disclosure relates to structured metadata for evidence probability in machine-ingestible web artefacts.
[0002] More particularly, the disclosure concerns a probability metadata envelope specification defining a canonical structured format for declaring provenance type, confidence score, attestation source, and freshness indicator within AIEP machine mirror pages, without disclosing internal probability engine logic.
Framework Context
[0003] This invention 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.
[0004] The present invention extends deterministic canonicalisation, governance, and execution integrity mechanisms defined in the AIEP environment while remaining independently implementable as described herein.
Background
[0005] AI agents evaluating web content as evidence require structured probability metadata to assess confidence, provenance, attestation, and freshness. Without a standardised envelope, agents must infer these properties heuristically from unstructured content.
[0006] A structured probability metadata envelope separates the public declaration of probability metadata from the internal probability engine logic (which remains confidential per the AIEP trade secret architecture), enabling third-party agents to consume probability declarations without access to scoring internals.
[0007] Existing systems do not provide a standardised, implementation-agnostic probability metadata envelope suitable for inclusion in machine mirror pages.
Summary of the Disclosure
[0011] A probability metadata envelope specification is disclosed defining four required fields for inclusion in the probability object of AIEP machine mirror pages.
[0012] The four required fields are:
(a) provenance_type: a declared identifier for the class of provenance of the content, such as primary_source, secondary_source, derived, or synthetic;
(b) confidence_score: a normalised decimal value in the range [0.0, 1.0] representing declared confidence in the content, where 1.0 represents maximum confidence;
(c) attestation_source: a URI or structured identifier declaring the attesting entity or system; and
(d) freshness_indicator: a structured indicator of content freshness comprising a as_of ISO 8601 timestamp and an optional ttl_seconds integer.
[0013] A ProbabilityEnvelopeHash is computed as:
ProbabilityEnvelopeHash = H(provenance_type || confidence_score || attestation_source || freshness_indicator)
[0014] The probability envelope declares metadata only. It does not expose internal probability engine computation logic, scoring functions, weight distributions, or Bayesian parameters.
[0015] Agents must treat probability metadata as declared input to their own admissibility evaluation and must not treat confidence_score as a determinative admissibility threshold without applying their own version-bound evaluation rules.
[0016] The technical effect is provision of a standardised, implementation-agnostic probability metadata interface for AIEP machine mirror pages enabling structured agent consumption without disclosure of internal scoring logic.
Brief Description of the Drawings
[0017] Figure 1 illustrates the probability envelope field structure and valid value ranges.
[0018] Figure 2 illustrates the separation boundary between the public probability envelope and confidential internal engine logic.
[0019] Figure 3 illustrates the ProbabilityEnvelopeHash computation.
[0020] Figure 4 illustrates agent handling of probability metadata in admissibility evaluation.
ASCII Drawings
Figure 1 — Probability Envelope Field Structure
+------------------------------------------+
| probability object |
|------------------------------------------|
| provenance_type : "primary_source" |
| "secondary_source" |
| "derived" |
| "synthetic" |
|------------------------------------------|
| confidence_score : 0.0 ... 1.0 (decimal) |
|------------------------------------------|
| attestation_source: URI or structured ID |
| of attesting entity |
|------------------------------------------|
| freshness_indicator: { |
| "as_of": ISO8601, |
| "ttl_seconds": integer (optional) |
| } |
+------------------------------------------+
Figure 2 — Public Envelope / Confidential Engine Boundary
PUBLIC (declared in mirror page)
+------------------------------------------+
| probability envelope |
| - provenance_type |
| - confidence_score |
| - attestation_source |
| - freshness_indicator |
+------------------------------------------+
CONFIDENTIAL (not disclosed)
+------------------------------------------+
| Internal probability engine |
| - Scoring functions |
| - Weight distributions |
| - Bayesian parameters |
| - Computation logic |
+------------------------------------------+
No internal logic exposed via envelope.
Figure 3 — ProbabilityEnvelopeHash Computation
provenance_type ----+
confidence_score ----+
\
attestation_source -----+--> H(provenance_type ||
/ confidence_score ||
freshness_indicator ----+ attestation_source ||
freshness_indicator)
|
v
+-----------------------------+
| ProbabilityEnvelopeHash |
+-----------------------------+
Figure 4 — Agent Handling of Probability Metadata
Agent receives mirror page
|
v
Extract probability envelope
|
v
+------------------+
| Apply agent's |
| own version-bound|
| evaluation rules |
| to declared |
| metadata |
+--------+---------+
|
confidence_score is INPUT
NOT determinative threshold
|
v
Agent admissibility determination
(agent's own rules govern outcome)
Detailed Description
1. Required Fields
[0021] The probability envelope comprises four required fields: provenance_type, confidence_score, attestation_source, and freshness_indicator.
[0022] provenance_type declares the class of provenance. Permitted values include primary_source, secondary_source, derived, and synthetic. Extension values may be declared via the schema extension framework (P65).
[0023] confidence_score is a normalised decimal in [0.0, 1.0]. It is a declared value only and does not represent an internal engine output.
[0024] attestation_source is a URI or structured identifier of the attesting entity or system.
[0025] freshness_indicator comprises as_of ISO 8601 timestamp and optional ttl_seconds.
2. No Internal Logic Disclosure
[0026] The envelope declares metadata only. Internal scoring functions, weight distributions, Bayesian parameters, and computation logic are not declared in or derivable from the envelope.
3. Agent Handling
[0027] Agents must apply their own version-bound evaluation rules to probability metadata. The confidence_score field is declared input and must not be treated as a determinative admissibility threshold without agent-side evaluation.
4. ProbabilityEnvelopeHash
[0028] A ProbabilityEnvelopeHash is computed as H(provenance_type || confidence_score || attestation_source || freshness_indicator) for binding and cross-node equivalence verification.
Claims
-
A probability metadata envelope specification for AIEP machine mirror pages comprising: (a) four required fields: provenance_type declaring provenance class, confidence_score as a normalised decimal in [0.0, 1.0], attestation_source as a URI or structured identifier, and freshness_indicator comprising as_of ISO 8601 timestamp and optional ttl_seconds; (b) a ProbabilityEnvelopeHash computed as H(provenance_type || confidence_score || attestation_source || freshness_indicator); (c) a prohibition on disclosure of internal probability engine logic, scoring functions, weight distributions, or Bayesian parameters through the envelope; and (d) a requirement that agents apply their own version-bound evaluation rules to probability metadata rather than treating confidence_score as a determinative admissibility threshold.
-
The specification of claim 1 wherein extension provenance_type values may be declared via the schema extension framework.
-
A computing system configured to generate probability envelopes conforming to the specification of any of claims 1 to 2.
-
A non-transitory computer-readable medium storing instructions which, when executed, generate or validate probability envelopes conforming to the specification of any of claims 1 to 2.
Licence
Any person is granted a perpetual, irrevocable, worldwide, royalty-free licence to make, use, implement, modify, or distribute any system or method described in this disclosure for any purpose, without restriction, under the Apache License 2.0.
A copy of the Apache License 2.0 is available at https://www.apache.org/licenses/LICENSE-2.0
Abstract
A probability metadata envelope specification is disclosed for the probability field of AIEP machine mirror pages. Four required fields are defined: provenance_type declaring provenance class, confidence_score as a normalised decimal in [0.0, 1.0], attestation_source as a URI or structured identifier, and freshness_indicator comprising as_of ISO 8601 timestamp and optional ttl_seconds. A ProbabilityEnvelopeHash is computed as H(provenance_type || confidence_score || attestation_source || freshness_indicator). Internal probability engine logic is not disclosed through the envelope. Agents must apply their own version-bound evaluation rules to declared metadata and must not treat confidence_score as a determinative admissibility threshold.