Mechanism
The reversibility-aware commitment-point evaluator sits between the policy layer that proposes an actuation and the actuator chain that effects it. It comprises a reversibility classifier that classifies each proposed actuation into one of a plurality of reversibility classes, a commitment-point detector that identifies the point within the actuation chain beyond which the actuation becomes irreversible, a threshold modulator that elevates admissibility thresholds for proposed actuations classified as irreversible or as having irreversible sub-steps, a path-preference engine that prefers reversible actuation paths over irreversible ones where both are admissible candidates, and a commitment-lineage recorder that records each commitment-point transit in the lineage field.
The reversibility classes are governance-policy-defined and comprise at minimum: reversible actuations, wholly reversible through an inverse actuation, such as a gate-opening actuation reversible by a subsequent gate-closing actuation or a lane-change actuation reversible by a subsequent return-to-prior-lane actuation; partially reversible actuations, where some effects are reversible and some are irreversible, such as a braking actuation reversible in kinematic effect but irreversible in elapsed time and consumed stopping distance; irreversible actuations, where no inverse actuation returns the physical system to its pre-actuation state, such as a weapon-system terminal actuation, a suppressant-deployment actuation, a medication-dispensing actuation, or a cutting-tool actuation; time-bounded reversible actuations, reversible within a governance-policy-defined temporal window but irreversible after the window elapses; condition-bounded reversible actuations, where reversibility depends on external conditions that may or may not remain satisfied; probabilistically reversible actuations, where reversibility depends on a stochastic outcome; composite actuations comprising sequences of sub-actuations with mixed reversibility classes; and any further governance-policy-defined reversibility class. The classifier does not invent these classes; the reversibility ontology is governance-policy-defined and the preference among classes is itself governance-policy-configurable per actuator class and per deployment domain.
The commitment-point detector is the structural element. Many actuations appear monolithic at the policy level but are in fact composite, comprising a sequence of sub-actuations with mixed reversibility. For each staged actuation the detector identifies the stage beyond which continuation becomes irreversible. This enables the graduated-actuation mode selector to interrupt a staged actuation prior to the commitment point upon receipt of a newly-consumed observation that reduces composite admissibility, without incurring the consequences of the irreversible final stages. Each commitment-point transit is a lineage-recorded event supporting post-hoc analysis of which commitment points were or were not transited.
Once the class and commitment-point are known, the evaluator modulates the actuation discipline. The threshold modulator elevates admissibility thresholds for actuations classified as irreversible or as having irreversible sub-steps, so that an irreversible actuation must clear a higher bar than a comparable reversible one. The path-preference engine prefers reversible actuations to irreversible actuations among admissible candidates of comparable projected outcome, prefers composite actuations with late commitment points to those with early commitment points, and prefers stage-gated execution of irreversible composite actuations over unstaged execution. The reversibility preference and the actuation determination are recorded together in the lineage field.
Operating Parameters
The evaluator's behavior is governed by the governance-policy-defined reversibility ontology and by the admissibility thresholds it modulates. The threshold modulator elevates admissibility thresholds for proposed actuations classified as irreversible or as having irreversible sub-steps. The reversibility preference is governance-policy-configurable per actuator class and per deployment domain, and both the preference applied and the resulting actuation determination are recorded in the lineage field. The specification states these as governance-policy-defined parameters rather than fixed numeric thresholds, frequencies, or latency budgets.
The commitment-point detector identifies, for each staged actuation, the stage beyond which continuation becomes irreversible. Because the detector locates that stage within the actuation chain rather than at the start of contemplation, the graduated-actuation mode selector can interrupt a staged actuation prior to the commitment point upon receipt of a newly-consumed observation that reduces composite admissibility. The path-preference engine prefers stage-gated execution of irreversible composite actuations over unstaged execution, which converts a single irreversible actuation into a staged sequence whose final stage is the irreversible one and whose earlier stages remain interruptible before commitment.
The evaluator composes with the graduated-actuation mode selector, whose modes range across disabled, simulated, advisory, consultative, constrained, stage-gated, and full. An irreversible or uncertain actuation can therefore be routed into a consultative mode, in which the actuator emits a governance-credentialed consultation request to a human operator, a higher-authority agent, or a governance-policy-defined consultation endpoint, or into a stage-gated mode, in which execution proceeds only stage by stage. Mode selection and any escalation are recorded in the lineage field alongside the actuation determination.
Alternative Embodiments
The evaluator admits several embodiments differing in how the reversibility ontology is expressed, how the commitment-point is detected, and how an irreversible or uncertain actuation is escalated. Because the reversibility ontology is governance-policy-defined, an embodiment may enumerate the reversibility classes explicitly per actuator class, or define them by governance-policy-defined predicates evaluated against the actuation and its context. New actuator classes acquire their reversibility classification through the same governance-policy flow that maintains the rest of the actuation policy.
The commitment-point detection admits more than one embodiment. In one, the commitment point is a named stage within the staged actuation chain, and the graduated-actuation mode selector holds the actuation at the stage immediately before it. In another, the commitment point is a function of actuator state, and continuation past it is conditioned on the pre-commitment conditions remaining satisfied. The choice is governed by how the underlying actuator is staged and controlled.
Escalation for irreversible or uncertain actuations also admits embodiments. Where a human operator or higher-authority agent is reachable within the decision window, the actuation can be routed into a consultative mode that emits a governance-credentialed consultation request to that operator, that higher-authority agent, or a governance-policy-defined consultation endpoint. Where no such party is reachable, the elevated admissibility threshold and the stage-gated discipline still apply. The evaluator does not itself authorize the irreversible actuation; it elevates the admissibility bar, prefers reversible and late-commitment paths, and records each commitment-point transit in the lineage field.
Composition
The evaluator composes upward with the policy layer and downward with the actuation chain. Upward, the path-preference engine prefers reversible actuations to irreversible actuations among admissible candidates of comparable projected outcome, and prefers composite actuations with late commitment points to those with early commitment points. This lets the system favor a reversible or late-commitment alternative over an irreversible or early-commitment one where both are admissible, rather than committing the irreversible candidate by default.
Downward, the evaluator composes with the graduated-actuation mode selector and the actuation chain. Where an actuation is staged, the commitment-point detector identifies the stage beyond which continuation becomes irreversible, and the mode selector can interrupt the staged actuation prior to that stage upon receipt of a newly-consumed observation that reduces composite admissibility. Where an actuation cannot be staged, the elevated admissibility threshold for irreversible actuations still applies before execution proceeds.
Laterally, the evaluator composes with the lineage field. Each commitment-point transit is a lineage-recorded event, and the reversibility preference applied is recorded alongside the actuation determination. The lineage record supports post-hoc analysis of which commitment points were or were not transited, and is the basis on which governance policy reviews and updates the reversibility ontology and preferences over time.
Cross-Domain Application
The evaluator operates across domains in which proposed actuations carry differential reversibility, with the reversibility ontology and the admissibility thresholds governance-policy-defined per actuator class and deployment domain. The specification's irreversible-class examples span domains: a weapon-system terminal actuation, a suppressant-deployment actuation, a medication-dispensing actuation, and a cutting-tool actuation are each cited as actuations for which no inverse actuation returns the physical system to its pre-actuation state. In autonomous surgery, a low-confidence determination can produce consultative-mode outcomes for proposed surgical actuations pending operator input, consistent with elevating the admissibility bar for irreversible steps. The braking example illustrates partial reversibility, reversible in kinematic effect but irreversible in elapsed time and consumed stopping distance.
In each domain the structural mechanism is invariant. The evaluator classifies the proposed actuation into a governance-policy-defined reversibility class, the commitment-point detector identifies the stage beyond which the actuation becomes irreversible, the threshold modulator elevates the admissibility threshold for irreversible actuations, the path-preference engine prefers reversible and late-commitment paths among admissible candidates, and the commitment-lineage recorder records each transit. The domain-specific component is the reversibility ontology itself, which is governance-policy-defined and encodes the domain's understanding of where reversibility ends. A deployment in a new domain imports its reversibility classification through the same governance-policy flow that maintains the rest of the actuation policy, without re-engineering the classification, detection, modulation, and recording discipline.
Prior-Art Distinction
As the specification states, prior architectures treat actuations uniformly regardless of reversibility. A single admissibility check applied to every actuation is appropriate for reversible actuations but does not distinguish the case in which an irreversible actuation is committed under the same assumptions that govern a reversible one. The specification distinguishes the present commitment-point evaluation on the ground that it prefers reversible paths where feasible, where prior architectures do not.
The evaluator described here generalizes by classifying each proposed actuation into a governance-policy-defined reversibility class, detecting the commitment point within the actuation chain, elevating admissibility thresholds for irreversible actuations, preferring reversible and late-commitment paths among admissible candidates, and recording each commitment-point transit in the lineage field. This combination of reversibility classification, commitment-point detection, threshold modulation, path preference, and commitment-lineage recording is the distinction the specification draws against architectures that treat actuations uniformly.
Disclosure Scope
This article describes subject matter disclosed in U.S. Provisional Application No. 64/049,409. The disclosure covers the reversibility-aware commitment-point evaluator: the reversibility classifier operating on a governance-policy-defined reversibility ontology, the commitment-point detector that locates the stage beyond which an actuation becomes irreversible, the threshold modulator that elevates admissibility thresholds for irreversible actuations, the path-preference engine that prefers reversible and late-commitment paths among admissible candidates, and the commitment-lineage recorder. It covers the embodiments above and any embodiment that preserves these elements together. The reversibility ontology and the reversibility preference are governance-policy-configurable per actuator class and per deployment domain.
