1. The Gap
Consider an agent that wants to be corrected. In this architecture, openness to correction is a mechanical state: an agent that has recorded no complied corrective encounter in an action class is an untested class, and an untested class holds the short decay constant, so its earned persistence decays fastest. The way out is exposure. An agent becomes testable by admitting conduct evaluation artifacts (116) from parties positioned to correct it, and the filing has a precise notion of which parties those are.
The earned-exposure architecture already produces such exposure, but only as a byproduct of deference. When a persistent semantic agent (100) elects to execute under a second agent's recorded determination, it appends a deference record (900) that the followed agent counter-signs, and that counter-signature decrements the followed agent's own authorization budget (404), the same budget that gates the followed agent's dispatch. Responsive to the counter-signature, the deferring agent enrolls the followed agent's conformance-filtered corrector classes into its own register. Deference is priced against the followed agent, and exposure is the deferring agent's return for adopting a judgment it did not itself produce.
That coupling leaves a gap. An agent may want the exposure without either half of the deference bargain. It may not wish to adopt any determination of the second agent, because adopting one means executing under a judgment its own admission evaluator (120) never reached. And it may not wish to impose a cost on the second agent, because the deference decrement draws down the second agent's own authorization budget (404), the capacity it needs to act, merely for being followed. Deference forces both conditions. Elective exposure enrollment removes both, and prices the exposure to the party that actually wants it.
2. Mechanism
Start from what is being enrolled. A corrector class is not a party an agent likes or rates. It is an origin-equivalence class (200) that a two-part conformance filter (906) admits. The first part is correction: the records show the second agent produced an accepted determination (122) upon a conduct evaluation artifact (116) originating from a party assignable to that class, so the class is recorded as having corrected the second agent, not merely interacted with it. The second part is subsequent conformity: within an observation interval, the second agent's later conduct is recorded as compliant with that determination rather than non-compliant. A class that corrected the second agent but that the second agent then ignored fails the second part and is excluded. No rating, score, endorsement, or vouch of the second agent is retrieved or consumed; the filter reads recorded corrections and recorded conformity, nothing else.
Each admitted class is carried not by the second agent's local class identifier, meaningless outside that agent, but by a class-defining reference: the recorded fact that defines the class. It comprises a relation-type designator, being a common introduction path (206), a shared dispatch lineage (202), or a co-signature (204); the defining member of that type; and the identity primitive of each asserting party (118) assigned to the class. The reference is portable: both agents evaluate it against their own records, and the electing agent never adopts the second agent's private numbering.
Elective enrollment then proceeds by the agent's own election and without any deference. The persistent semantic agent (100) retrieves the second agent's corrector-class enumeration computed under the two-part conformance filter (906), and appends an elective exposure enrollment record. That record comprises, conjunctively, the second agent's identity primitive, the corrector-class enumeration with the class-defining reference of each class, and the action-class identifier. The agent decrements its own authorization budget (404) by an amount declared in its own signed policy object (112) for an elective enrollment, and writes each enumerated origin-equivalence class (200) into its own replenishment register (402) as an admissible-but-unexercised entry.
Three negatives define the mechanism as sharply as its positive steps. No deference record (900) is appended. No determination of the second agent is adopted, so nothing the electing agent later dispatches rests on a judgment it did not produce. And the authorization budget (404) of the second agent is not decremented. The cost of the enrollment is borne by the electing agent alone.
The entry it writes has two effects and no third, exactly as when reached through deference. For admissibility, a conduct evaluation artifact (116) from an asserting party (118) assignable to the enrolled class is admissible to the electing agent's admission evaluator (120) for the action class notwithstanding that the agent holds no prior encounter history with that party. For renewal, the entry is not counted while its exercise field stands at the unexercised value: merely writing it renews no authorization quantity, and the agent's quantities stand as they would have stood had the entry not been written. Enrollment buys admissibility, not credit.
Assignment of an arriving artifact to an enrolled entry runs by an ordered test evaluated on the electing agent's own records alone, with no registry and no coordination with another node. First, direct membership, where the arriving party's identity primitive is a member of the enumeration in the class-defining reference. Second, definition membership, where the defining member is satisfied against the agent's own records under the relation-type designator carried. Third, on plurality, the entry earliest written into the register is transitioned and each further assignable entry is named without being transitioned, so one arriving artifact exercises at most one enrolled class. Where no test is satisfied, a non-assignable-arrival record naming the artifact and the failed test is appended, the artifact is not admissible on the ground of enrollment, no entry is transitioned, and no authorization budget (404) is replenished.
When an admissible artifact arrives from an enrolled class and implicates the action class, the agent appends an exercised transition record (908) and writes the exercise field to the exercised value. The entry the electing agent paid for from its own budget is thereby exercised.
3. Operating Parameters
The disclosure fixes no numeric magnitude for this mechanism. Every quantity it depends on is declared in a signed policy object (112), and the operative values are the electing agent's own.
The enrollment cost is a decrement to the electing agent's authorization budget (404) by an amount declared in that agent's own signed policy object (112) for an elective enrollment. The disclosure states that such an amount is declared and spent, and states no figure.
The observation interval used by the second part of the two-part conformance filter (906) is the interval declared for the action class in the electing agent's own signed policy object (112). Where the electing agent declares none, the interval declared for the second agent applies, but only where it is not longer than a maximum the electing agent declares; a longer declared interval is replaced by that maximum.
In a further embodiment, the second part of the filter requires a count of complied conduct records not less than a minimum declared in the electing agent's signed policy object (112), rather than a single complied record, so the agent enrolls only classes to whose corrections the second agent repeatedly conformed. A class recorded with fewer complied subsequent records than the declared minimum is excluded, and the exclusion is recorded.
An enrollment does not persist unconditionally. An admissible-but-unexercised entry whose exercise field still stands at the unexercised value upon expiry of an enrollment window declared in the electing agent's signed policy object (112) is written to a lapsed value and ceases to confer admissibility. The lapse is appended, produces no determination concerning the enrolled origin-equivalence class (200), increments no counter, and appends nothing adverse to any party.
4. Composition
Elective enrollment is a second entrance to machinery the filing builds once. The admissible-but-unexercised entry, its two effects, its ordered assignment test, and the exercised transition record (908) are defined for enrollment reached through deference, and elective enrollment writes the same entry and inherits all of it. The filing thus discloses two disjoint routes to one register state. On the deference route the followed agent pays through its counter-signed decrement and the deferring agent earns the exposure by adopting a determination; on the elective route the electing agent pays from its own budget and adopts nothing. The endpoint is identical; only the payer and the presence of an adopted determination differ.
The budget decrement is not nominal. The authorization budget (404) is the quantity whose exhaustion drives the authorization gate (300) to the withheld state (310) and the agent into the non-executing cognitive mode (302) for the affected action classes. Spending it on an elective enrollment moves the electing agent toward that floor, consuming the same capacity the agent needs in order to act.
Two invariants of the filing hold across the enrollment. Nothing enrolled becomes a scalar: the corrector-class enumeration and the records that carry it are not converted into a rating or a threshold operand, consistent with the conversion bar (502). And nothing the enrollment does is adverse to the second agent: its budget is untouched, no determination concerning it is produced, and it is not even a required participant in the enrollment.
Elective enrollment sits within the section on earned exposure without deference beside a distinct mechanism, the standing-determination designation, under which a conjunct of one agent's dispatch-authority predicate becomes satisfiable, in the alternative, by an invariant determination of a counterparty. The two are separate: elective enrollment governs which artifacts an agent will admit and be tested by, and adopts no determination of the second agent, whereas the standing-determination designation substitutes a counterparty's determination inside the dispatch-authority predicate.
5. Prior-Art Distinction
The mechanism occupies space near several established categories, and the distinctions are structural. Reputation and endorsement systems, including web-of-trust key-signing and vouching graphs, move a credential of standing from one party to another, a signature or score that raises the recipient's rating. Elective enrollment moves no such quantity. It retrieves a conformance-filtered set of corrector classes, adopts no determination, and confers no standing on either party; the conversion bar (502) forbids turning any of the enrolled material into a scalar. What the electing agent gains is not a better reputation but an obligation to admit and be tested by parties that have corrected a well-conformed agent.
Capability delegation frameworks, including access-control grants and token-based authorization, work by a holder granting a capability to a grantee. Here the second agent grants nothing and need not act. The electing agent writes entries into its own register by its own election, and the second agent's authorization budget (404) is not decremented and its participation is not required. This inverts delegation: the beneficiary pays and the source is left untouched.
Recommender and collaborative-filtering systems expand a user's exposure on the basis of similarity or popularity, optimizing for engagement. The corrector-class enumeration is not a similarity score: it is the output of a two-part filter requiring a recorded correction followed by recorded conformity, and the exposure it produces is exposure to being corrected, not to being agreed with. The electing agent pays a metered budget cost for it rather than receiving it as a free suggestion.
Federated identity and attestation systems bind a portable claim of standing that a relying party accepts on the issuer's authority. The class-defining reference is portable, but it is a structural fact, a relation type and its members, that the electing agent evaluates against its own records; it asserts nothing about the second agent's trustworthiness. None of these categories provides a self-funded, deference-free path to admitting the correctors of another agent while leaving that agent's capacity to act wholly untouched.
6. Disclosure Scope
The operative disclosure is U.S. Provisional Application No. 64/117,812. The elective exposure enrollment mechanism is disclosed in Section 10.7 at paragraph [0439], within the section on determination invariance, standing determinations, and earned exposure without deference. Its supporting apparatus is disclosed in the same filing: the two-part conformance filter and the corrector-class enumeration at paragraphs [0314] through [0319], and the admissible-but-unexercised entry, its two effects, the ordered assignment test, and the exercised transition record (908) at paragraphs [0320] through [0323]. The observation-interval treatment appears at [0315], the repeated-conformity variant at [0443], and the enrollment-window lapse at [0446].
What is disclosed is the structure: enrollment of a second agent's conformance-filtered corrector classes by the electing agent's own election; an elective exposure enrollment record naming the second agent's identity primitive, the enumeration with each class-defining reference, and the action-class identifier; a decrement of the electing agent's own authorization budget (404) by a policy-declared amount; and the write of each origin-equivalence class (200) into the electing agent's own register as an admissible-but-unexercised entry, with no deference record (900) appended, no determination adopted, and the second agent's budget undecremented.
What is disclaimed matters equally. The disclosure fixes no numeric value for the enrollment decrement, the observation interval, the repeated-conformity minimum, or the enrollment window; each is declared in a signed policy object (112) and lies outside the disclosure. It claims no rating, score, credential, or standing quantity passing between the agents, and no cardinal measure of benefit. Descriptions in this article of motivating problems and prior-art categories are context, not part of the filed disclosure. Where this article and the filed specification differ, the filed specification of U.S. Provisional Application No. 64/117,812 controls.
