The gap

In mixed-fleet environments — human-driven vehicles alongside autonomous ones, military drones alongside piloted aircraft, surgical robots alongside human surgeons — intent is implicit. A human operator's intentions must be inferred from behavior, and there is no standard way for a machine to receive structured intent from a human across different fidelity tiers. Current systems either require full teleoperation (high fidelity but high bandwidth) or rely entirely on behavioral inference (low bandwidth but uncertain).

This creates a structural problem: a machine that cannot distinguish between competence-based risk and intent-based hostility cannot make correct escalation decisions, and a system that cannot receive retracted intent cannot honor withdrawal of consent.

The invention

A three-tier intent primitive: full cognitive-state broadcast (tier 1) where the operator's agent shares its complete cognitive state; structured partial-fidelity bus extraction (tier 2) producing governance-credentialed intent observations; and behavior-inferred intent from sensor cues (tier 3) producing inferred observations with explicit uncertainty.

Cross-tier composite admissibility fuses intent from all three tiers with tier-weighted evidential factors. Verification-feedback inference-function evolution compares inferred intent against observed outcomes. Risk-vs-hostility profile bifurcation separates competence-based risk from intent-based hostility as structurally distinct profiles. Governance-credentialed retraction enables withdrawal of previously expressed intent.

The inventive step

Prior art treats intent as a single signal — teleoperation commands, natural language instructions, or behavioral inference. This primitive treats intent as a multi-fidelity input stream where higher fidelity is preferred but not required, and where the system can operate on any combination of available tiers. The bifurcation of risk from hostility is structurally novel: existing systems conflate competence failure with malicious intent.

The departure is that intent is a governed observation at every fidelity tier, with tier-weighted admissibility and verification-feedback evolution — the system learns to infer intent better over time without compromising safety.

Alone, and in composition

On its own, operator intent applies to any mixed-fleet environment — autonomous vehicle handoff, defense engagement authorization, surgical procedure control, search-and-rescue coordination, and elder-care robotics. The risk-vs-hostility bifurcation and verification-feedback evolution are independently useful.

In composition, operator intent feeds the actuation governor and the confidence governor — the system knows what the operator intends, at what fidelity, with what uncertainty, and can factor that into every governed execution decision.