Evidential Weighting in Governance Chain

Mechanism

Weighting operates as a governance-policy-defined function applied as an observation is admitted to the shared observation store. The disclosure enumerates the factors that determine an observation's evidential weight. First, the contributing device's authority level under the authority taxonomy: a high-authority contributing device carries high evidential weight and may be treated as a substrate condition, while a low-authority device carries low evidential weight and contributes primarily as advisory input to the composite admissibility evaluator. Second, the sensing modality's governance-policy-defined reliability characteristics. Third, the consistency of the observation with other observations from independent sources, the inter-source consistency that the disclosure also describes as corroboration.

The disclosure further incorporates staleness, dispositional context, and the contributing device's reputation track record, together with any other governance-policy-defined evidential-weighting factor. The factor set is therefore extensible by governance policy rather than fixed. Continuity-based identity, the trust-slope continuity through which device identity is established, conditions whether a contribution is admitted at all and feeds the authority determination; the disclosure does not state a fixed numeric continuity factor.

The factors combine through a composite weighting function declared by governance policy. The disclosure describes the resulting effective evidential weight as recorded in lineage alongside the observation, with the factors contributing to each weight, the outcome, and the governance-policy version applied. Downstream operations consume the weighted observation: an evidentially-weighted aggregation mechanism combines admitted observations through composite weighting rather than canonical selection, and derived observations carry an effective evidential weight computed from their input observations' weights through a governance-policy-defined derivation-weighting function. Auditors reconstruct both the weighting and the aggregation from the recorded lineage and governance-policy version.

Evidential weight is assigned by the chain rather than self-reported by the observer. An observation carrying a high-authority credential whose modality is reliable and whose reports are corroborated by independent sources receives high evidential weight; an observation from a low-authority source, a less reliable modality, or one inconsistent with independent sources receives lower weight. The weight, its contributing factors, and the governing policy version are recorded in lineage so that the standing of any observation can be reconstructed.

Operating Parameters

The function that combines the factors into an effective weight is governance-policy-defined rather than fixed. For derived observations, the disclosure states that the derivation-weighting function is selectable from a plurality of forms, including a minimum form producing weakest-input-limited derivation, a product form producing multiplicative derivation uncertainty, a weighted-average form, an uncertainty-aggregation form, a Dempster-Shafer combination form, a Bayesian combination form, a learned combination form, a composite form, or any equivalent governance-chain-preserving form. The disclosure does not designate any one of these as a default. The admissibility thresholds, the deferral-expiration parameters, and related gradations are configured independently by the deploying authority per governance policy.

Governance policy versions are themselves recorded in lineage, so a historical decision can be reconstructed by re-applying the admissibility rules and aggregation functions in force at the target time per the governance-policy-version lineage. The disclosure describes temporal reconstruction of the shared view state at any prior time for forensic, audit, regulatory, and legal purposes.

The disclosure provides graduated admissibility outcomes that interact with weight: an admit outcome stores the observation at the computed effective evidential weight, a gate outcome stores it at reduced effective evidential weight or subject to additional constraints, and a defer outcome holds it pending corroborating observations. The disclosure does not specify numeric weight floors, ceilings, or decay rates; where bounds or multipliers are applied, they are governance-policy-defined.

Alternative Embodiments

The combining function admits the plurality of forms the disclosure enumerates for derivation weighting: a minimum form, a product form, a weighted-average form, an uncertainty-aggregation form, a Dempster-Shafer combination form, a Bayesian combination form, a learned combination form, a composite form, or any equivalent governance-chain-preserving form. The learned combination form permits a machine-learning-derived weighting function while the function used remains identified in the governance-policy-version lineage.

Embodiments also vary in how the weight is represented. The disclosure states that evidential factors are governance-policy-configurable per deployment and that the factors admit a plurality of representations, including without limitation scalar weights, probability distributions over intent confidence, bounded uncertainty intervals, and governance-policy-defined composite representations. These tier factors compose with the evidential weighting of the shared observation store to produce unified composite admissibility.

The disclosure records, for each weight, the factors contributing to it and the governance-policy version applied, so that an aggregation can be reconstructed without re-running it. Derived observations carry their own effective evidential weight computed from their inputs, and the derivation-lineage record supports bidirectional traversal forward from an input observation to the derived observations depending on it and backward from a derived observation to its transitive inputs.

Composition With the Five-Property Chain

Weighting is the second of the five governance-chain properties. The disclosure enumerates the chain as authority-credentialed observation, evidential weighting in a shared governed observation store, composite admissibility evaluation across cognitive domain fields, governed actuator execution, and lineage-recorded provenance. Authority-credentialed observation establishes the credentialed source identification that the weighting consumes; evidential weighting then assigns the graded contribution; composite admissibility evaluates every mutation against the dispositional, integrity, confidence, and capability fields before admission; governed actuator execution requires composite admissibility approval before any physical actuation; and lineage-recorded provenance links every observation, evaluation, and action through deterministic lineage. The chain is recursive: observations generated at any primitive's output feed back into the chain, and actuations emitted from the chain pass through every primitive's governance.

Composition with other primitives is structural. The matched-pair settlement primitive records each first observation, second observation, pairing determination, and downstream consumption in the governance-chain lineage field, and forecast observations enter the chain with evidential weight from forecasting-kernel track records. Adaptation artifacts produced through the training governance primitive carry a training-data provenance and attribution record in which each contributing training example's weight in the aggregate parameter update is recorded, so that a deployed adaptation artifact can be traced to the weighted evidence base from which it was derived.

Distinction From Prior Art

The disclosure distinguishes the composite admissibility evaluator from prior threshold-based admission mechanisms, prior voting-based admission mechanisms, and prior rule-based admission mechanisms: the evaluator computes a composite evidential weight integrating a plurality of factors, including authority, staleness, modality, dispositional context, reputation, integrity, and continuity, rather than applying a single-factor threshold, and it produces a plurality of outcomes rather than a binary admit or reject. The disclosure separately notes that prior centralized sensor-aggregation systems treat contributions homogeneously as undifferentiated sensor data, whereas each governed observation carries a verifiable authority credential that determines its evidential weight in each consuming agent's cognitive architecture.

The disclosure also distinguishes the shared observation store from prior spatial data substrates, prior cloud-hosted sensor databases, prior digital-twin architectures, prior federated data lakes, and prior blockchain append substrates: the store combines observations through evidentially-weighted aggregation incorporating authority, staleness, modality reliability, dispositional context, and reputation track record, whereas prior substrates apply canonical selection or last-writer-wins semantics, and it produces governance-chain-preserving lineage for every observation, admission, consumption, and correction event.

Worked Illustration

Consider a multi-source environmental observation fed into a downstream consuming agent. Three observers contribute reports of an event at approximately the same time and location. Observer A is a high-authority contributing device whose authority level places it among the highest-weighted sources; its reports may be treated as substrate conditions, and consistent reports from independent sources reinforce rather than duplicate it. Observer B is a derived inference produced by an automated process under a lower authority level; it carries lower evidential weight and contributes primarily as advisory input to the composite admissibility evaluator. Observer C is another contributing device whose modality has lower governance-policy-defined reliability characteristics, reducing its evidential weight relative to A.

The composite weighting function for this observation class is governance-policy-defined and recorded with its policy version. The evidentially-weighted aggregation mechanism combines the three weighted contributions through composite weighting rather than canonical selection, and the result carries lineage recording the factors that contributed to each weight, the outcome, and the governance-policy version applied. Where the evaluator cannot yet resolve admission, it may produce a defer outcome holding an observation pending corroborating observations, or a gate outcome admitting it at reduced evidential weight. A later review of how an observation was weighted can be evaluated against the recorded factors and policy version through temporal reconstruction, without re-running the entire aggregation.

Disclosure Scope

This article describes the evidential weighting property of the five-property governance chain disclosed in U.S. Provisional Application No. 64/049,409. The disclosure covers the weighting of contributed observations by authority level, sensing-modality reliability, inter-source consistency, staleness, dispositional context, reputation track record, and any other governance-policy-defined factor; the composite weighting function and the plurality of combination forms; the recording of each weight's contributing factors and governance-policy version in lineage; the evidentially-weighted aggregation mechanism and the derived-observation effective weight; and the composition of weighting with the remaining governance-chain properties. The numeric examples, factor weightings, and combining rules referenced in any deployment are governance-policy-defined and are not fixed by this disclosure.