Mechanism
The architecture's cross-tier composite admissibility evaluator receives, for each candidate intent, a governance-credentialed intent observation carrying the intent payload, the issuing unit's authority credential, the unit's fidelity-tier assignment, and a set of supporting intent observations. The evaluator consults governance policy to obtain the tier-weighted evidential factors and the admissibility evaluation context that apply at that tier, and tests the candidate intent against them. Intents that clear the governance-policy-defined evidential warrant on the strength of the credential alone are admitted directly. For intents whose credential alone does not supply sufficient warrant, the evaluator accumulates tier-weighted supporting observations and asks whether enough corroboration exists to clear the admissibility threshold.
The accumulation is a composite over tier-weighted evidential factors. Each supporting observation carries its own credential and fidelity tier, and contributes a weighted increment to a composite admissibility evaluation. Tier 1 observations, full-fidelity cognitive-state observations broadcast by autonomous or highly integrated units, contribute the largest increments because the sharing unit reports its own cognitive state with full fidelity. Tier 2 observations, structured partial-fidelity intent signals such as turn-signal attestations, transponder codes, or integration-bus fields, are admitted at governance-policy-defined moderate evidential weight reflecting both their structured provenance and the operator-override susceptibility inherent in partial-fidelity sharing. Tier 3 observations, behavior-inferred intent produced by mesh observation of externally-visible behavioral cues, contribute the smallest increments and carry their inference confidence, because inference can be wrong. The composite is compared against the governance-policy-defined admissibility threshold, and the intent is admitted only when the composite clears it.
The output of the evaluator is itself a governance-credentialed observation recorded in the intent-lineage field: the candidate intent, the tier, the tier-weighted evidential factors in force at the time of evaluation, the supporting observations that contributed, the resulting composite admissibility evaluation, and the admission decision. Each intent emission, admission, fusion, verification, retraction, and downstream consumption is recorded in the governance-chain lineage, so an auditor can reconstruct the decision because the governing policy, the credential set, and the observation set at the moment of evaluation are all retrievable.
Operating Parameters
The tier-weighted evidential factors are governance-policy-configurable per deployment. A defense-context deployment operating against an adversarial population may apply different factors than a civilian-context deployment operating against a cooperative population: the former may discount Tier 1 broadcasts more, since broadcasts can be spoofed, while the latter may weight Tier 3 inferences less aggressively because most units in the environment are cooperating. The factors admit a plurality of representations, including scalar weights, probability distributions over intent confidence, bounded uncertainty intervals, and governance-policy-defined composite representations. Each configuration is recorded under governance policy so that admissibility decisions remain auditable across policy changes.
Tier boundaries and tier-weighted evidential factors are themselves governance-policy-configurable, and the framework admits additional tiers or finer-grained tier subdivision through governance-policy-defined tier specification without architectural modification. The admissibility evaluation composes with a threshold modulator that may elevate admissibility thresholds for proposed actuations classified as irreversible or as having irreversible sub-steps, so that a credential warranting a routine intent does not by itself warrant an irreversible one. Because admission turns on a composite over weighted factors rather than a single hard cutoff, intents that the credential alone does not clear pass into the corroboration regime rather than flipping abruptly between admitted and rejected.
Credential and corroboration are partially substitutable within the composite. An intent backed by a high-fidelity tier and a strong credential supplies more of the warrant from its own source and needs less corroboration; an intent inferred at a lower fidelity tier must draw more of its warrant from corroborating observation. Every admitted intent must clear the governance-policy-defined evidential warrant by some combination of the two.
Alternative Embodiments
In one embodiment the evidential factors are static and governance-policy-set. In another, the inference functions that produce Tier 3 observations are refined through the intent-verification feedback loop, which compares inferred intent against observed outcome and updates the inference function through the training governance primitive, producing inference-skill evolution. The verification variant maintains each inference function's track record so that consuming agents can weight it appropriately, and records every refinement in the governance chain so historical decisions remain auditable against the inference function in force at the time. A third embodiment supports governance-policy-defined factors that differ by deployment domain, so that a single deployment can apply civilian-context and defense-context evidential factors selected by the operating context rather than by global configuration.
A fourth embodiment routes conflicting cross-tier contributions through the integrity conflict resolution mechanism. When Tier 1, Tier 2, and Tier 3 contributions about the same unit agree, they reinforce the composite as concurring evidence. When they conflict, the conflict resolution mechanism produces a reconciled observation from the conflict set rather than admitting an unverified contribution, and that governance-credentialed observation is available to downstream consumers such as adversarial-aware operators and monitors. A unit whose Tier 1 broadcast conflicts with its Tier 2 structured signals or its Tier 3 behavior inference may be compromised, malfunctioning, or adversarial, and the reconciled observation preserves that conflict for downstream evaluation rather than collapsing it.
A fifth embodiment draws on the intent-retraction and correction mechanism, which supports governance-chain-preserving intent corrigibility. An operator whose intent does not clear the admissibility threshold may retract or correct the intent, or supply additional corroboration, and the correction is itself recorded in the intent lineage. This embodiment is appropriate where time permits the round-trip and where the operator can participate in supplying additional warrant.
Composition
Tier-weighted admissibility composes with the operator intent sharing primitive of the governed spatial mesh. The same fidelity tiers that classify intent emission, Tier 1 full-fidelity cognitive state, Tier 2 structured partial-fidelity intent, Tier 3 behavior-inferred intent, also supply the tier-weighted evidential factors applied during admission. The architecture maintains a single fidelity-tier vocabulary across emission, observation, and admissibility, so that a unit's tier means the same thing whether it appears as the source of an intent or as a contributor of corroborating evidence.
The mechanism also composes with the capability envelope of Chapter 7 and with the governed mesh's authority evaluation. Once an intent is admitted by the tier-weighted evaluator, the resulting action still passes through the capability envelope and the governing authority before execution: admission warrants only that the intent is evidentially admissible, not that the agent is physically capable of the action or authorized to perform it. Admissibility, capability, and authority are separate evaluations, and an action proceeds only when all concur. This separation prevents the failure mode in which a strong credential is allowed to override either physical infeasibility or governance prohibition.
Prior-Art Distinction
Prior vehicle-to-everything communication systems, DSRC/IEEE 802.11p and C-V2X/3GPP, define cross-vehicle message formats but lack governance-credentialed authority evaluation, fidelity-tier structure, and cross-source admissibility weighting. Prior advanced driver assistance systems infer other-unit behavior from onboard sensors without governance-chain integration, and prior driver-intent-prediction research produces probabilistic estimates without authority credentialing or attestation. None of these supports cross-source composite fusion of intent observations across fidelity tiers. The disclosed mechanism binds evidential weight to fidelity tier rather than to source count, so that a single Tier 1 cognitive-state observation can carry more warrant than many Tier 3 inferences when its informational value warrants it.
Prior coordination architectures also assume a uniform participation level, treating each participant as either able to share intent or not, without graduated-fidelity tiers or cross-source admissibility weighting. The disclosed mechanism explicitly admits fully-shared, structured-partial, and behavior-inferred participation in a unified framework, and couples the resulting tier-weighted composite admissibility to a governance-credentialed admissibility observation recorded in the lineage. That combination is not present in prior intent or coordination architectures.
Worked Example
Consider an urban traffic environment in which an autonomous vehicle is deciding whether to admit a yield intent attributed to an oncoming vehicle. A Tier 1 cognitive-state observation from the oncoming vehicle declares its intent to yield. A Tier 2 structured signal, a turn-signal attestation, corroborates it. A Tier 3 inference from mesh observation of behavioral cues observes brake-light illumination and deceleration consistent with yielding. All three concur, the composite admissibility evaluation clears the threshold by a substantial margin, and the yield intent is admitted. The autonomous vehicle proceeds.
Now consider the same scenario in which the Tier 1 observation declares yield but the Tier 2 turn-signal attestation is absent and the Tier 3 inference observes acceleration rather than deceleration. The integrity conflict resolution mechanism produces a reconciled observation from the conflict set, the composite admissibility evaluation remains below threshold despite the Tier 1 contribution, and the yield intent is not admitted. The governance-credentialed reconciled observation is available to downstream consumers, including any adversarial-aware monitor. The autonomous vehicle treats the oncoming vehicle as if no yield intent had been issued, and falls back to the trajectory the Tier 3 inference supports. The separation between admission and corroboration is what lets the architecture handle the conflict without conflating an unverified observation with a confirmed intent.
Deployment Considerations
Deploying tier-weighted admissibility into a heterogeneous fleet requires three coordinated choices. First, the fidelity-tier classifier must assign each unit a tier through a credentialed mechanism, whether self-declared, credential-based, observation-based, capability-based, manufacturer-attestation, dynamic, or a hybrid, so that tier assignments are auditable rather than ambient. Second, each contributing observation must carry its tier alongside its payload, so the evaluator can apply the correct evidential weight without inferring the tier from channel identity. Third, governance policy must support credentialed configuration of tier factors and thresholds without invalidating historical decisions; the architecture records each admission against the governing policy in force at the moment of evaluation, so that subsequent changes do not retroactively alter the meaning of a recorded decision.
Mixed-fidelity cooperation, a few highly integrated units sharing Tier 1 cognitive state, more units sharing structured Tier 2 signals, and legacy units inferable only at Tier 3, is the operating condition the primitive is built to address. The disclosed mechanism handles this regime by construction: tier-weighted evidential factors can be set to give Tier 1 observations greater warrant even when they are outnumbered by Tier 3 inferences, while still allowing concurring Tier 3 inferences to be reconciled against a single anomalous Tier 1 observation through integrity conflict resolution. This is the structural foundation on which adversarial-aware operation and cooperative coordination rest.
Disclosure Scope
This article describes subject matter disclosed in U.S. Provisional Application No. 64/049,409. The disclosure covers any embodiment in which the admissibility of an operator-issued intent is evaluated through a cross-tier composite over tier-weighted evidential factors, with intents whose credential alone does not suffice lifted by accumulation of tier-weighted supporting observations against a governance-policy-defined admissibility threshold, and with the resulting decision recorded as a governance-credentialed observation in the intent lineage. The specific evidential factors, threshold functions, conflict-handling rules, and verification variants are illustrative embodiments. The disclosure spans civilian and defense contexts and is governance-policy-configurable per deployment, and extends to any combination of the variants described above so long as the governance-chain lineage and the credential-and-corroboration warrant structure are preserved.
