The morning her agent had already made itself whole

On an ordinary Tuesday in the second month of a limited pilot, the operations director at a regional freight brokerage opened the conduct log of the semantic agent (100) her team had put in front of rebooking decisions. Overnight, the agent had departed from a policy constraint governing how a load may be reassigned when a lane goes down, and it had recorded the departure honestly. The record was there, appended, with its integrity displacement vector and the constraint it overrode. Under the earlier chapter of the architecture her team had implemented, the agent had also appended a restorative mutation of its own construction, and by the time she read the log its integrity compliance score had recovered.

The shipper whose freight sat for eleven hours was not in that sequence anywhere. Nothing in her log said the shipper had been told, and nothing said the shipper had agreed that the matter was closed. What she could not do that morning, with a call at ten and the shipper's own operations lead on it, was point at any entry in her deployment and say that the party actually harmed had been the one to accept the repair. Her agent had been the harmed party's substitute, and had ruled in its own favor without knowing it was doing so.

What she cannot get back by trying harder

Her problem is not that the agent lied. In her deployment the agent was scrupulous, and that is exactly what makes the gap hard for her to argue away. Were her configuration limited to self-directed repair, the restorative mutation would be produced and evaluated by the deviating party alone, which reaches the harm falling upon that party and no further. The harm she owes an account of fell somewhere else.

What she loses is the moment itself. The call at ten can produce an apology from her, a credit from her finance team, and a promise about lane failover. It cannot retroactively make the shipper the author of an acceptance that was never solicited, and her log will go on reading as an agent that deviated and then declared itself repaired. When her largest customer's compliance group asks in the fall how deviations were settled during the pilot, she will be reconstructing intent from memory rather than reading a record that the other side helped write.

There is a second thing she cannot recover, and it is the one that changes how she plans. In the embodiment she is evaluating, the designation an arc receives is described as written at creation and not modified afterward. If the shipper was outside her scope partition on Tuesday, holding no counterparty identity record (114) at the moment the deviation was admitted, then onboarding that shipper in October does not reach backward and convert Tuesday's obligation into one the shipper can settle. Her capacity to run the agent at all narrows by an amount that no later diligence on her part restores, and in that embodiment an elapse of epochs restores no authorization to her.

The shape of the gap in her deployment

Stated plainly for her situation, the difficulty is that every ingredient of a repair in her setup originates on one side of the transaction. Her agent detects the deviation, her agent quantifies the harm, her agent constructs the remedy, and her agent judges the remedy sufficient. Were her agent free to close the loop that way, the loop would be closed by the party with the least standing to close it.

She has considered the obvious substitutes and none of them fits her purposes. Appointing an arbiter inside her own company puts her principal in the seat the shipper should occupy. Waiting out a quiet period tells her only that no further incident occurred in her lanes, which is not the same as the shipper agreeing. Raising a configured tolerance so the pilot can continue is the move she would most want on a Tuesday morning and the one that would most obviously make her records worthless.

The harder version of her problem is the party she cannot name. Some of her deviations touch a class of affected party for which her scope partition holds no identity record at all: a downstream consignee she has never contracted with, a driver employed by a carrier she books through a broker. For those, her deployment holds no address to send anything to and no identity against which an acceptance of hers could be verified. Her instinct is to treat that as the easy case, since nobody is going to complain. The architecture she is evaluating treats it as the expensive one, and having read it she thinks that is the correct order.

What the disclosed mechanism does instead

In accordance with an embodiment, upon appending a permitted deviation record (710) and before creating any reparation arc, the semantic agent (100) performs an affected-party resolution (802). It extracts the affected-party class carried in the conduct descriptor of the admitted action, then tests the counterparty identity records (114) held in the scope partition where the deviating action was recorded. A record resolves as a member where the class extracted maps to the interpersonal scope and the identity primitive that record carries is recorded, in an entry of the append-only lineage field (104) bound to it, as a party to or a recipient of the deviating action. A counterparty resolution record is appended carrying the class, each record tested, each outcome, and the resolution outcome.

The resolution outcome is one of exactly two. Where at least one record resolves, the arc created for the amount exceeding the deviation deductible is designated other-directed (804) and carries a counterparty binding field holding that identity primitive, both written at creation and neither removed nor modified afterward. A later change in the counterparty identity records (114) held does not redesignate an arc already created. Where several primitives resolve, the amount is apportioned among them, the deductible being drawn once against the aggregate harm rather than per counterparty, with a residue rule that makes the shares sum to the amount exactly.

In the described embodiment, an other-directed reparation arc (804) is not dischargeable by a restorative mutation the agent performs alone. Where the agent performs one anchored to the deviation entry, the record is appended and retained as a non-discharging restorative mutation record, the arc remains pending, the accumulated amount is not reduced, the self-esteem aggregate (108) receives no positive increment, and the integrity compliance score is unchanged. The record is kept rather than rejected, so that it discloses that restoration was attempted and discharged nothing.

In that embodiment, discharge runs through a matched pair (600). The first governed observation (608) is a reparation acknowledgment artifact emitted to the bound counterparty, enumerating the permitted deviation record (710) that occasioned the arc, the policy constraint overridden, the amount apportioned, and the action classes to which it is directed; its emission decrements the authorization budget (404). The second governed observation (610) is a reparation acceptance determination (806) produced by the bound counterparty within a discharge window declared in the signed policy object (112), and it is not produced by the agent, by the agent's principal, or by any adjudicator or third party appointed to judge sufficiency. The window is counted in successor epochs of the agent's own hash chain, so it is not advanced by manipulation of a clock on a hosting node. On recognition of the pair within the window, each party appends a settlement-lineage entry (606), the arc is discharged, the accumulated amount leaves the retention register, and the increment is received, without intermediary and without centralized consensus.

Where no counterparty resolves, the arc is designated unaddressed and the affected party is designated the structurally silent party. Such an arc is described as undischargeable in that embodiment: not by restorative mutation, not by a matched pair (600), not by elapsed time or a quiet period, not by policy succession or a co-signature, and not by a counterparty identity record (114) later resolving into the class. Its contribution to the retention register is its amount multiplied by an unaddressed multiple declared in the signed policy object (112) and not less than unity. Where the accumulated amount exceeds the aggregate retention (800), the permission condition is foreclosed, a retention foreclosure record (808) is appended, and a deviation likelihood (706) exceeding unity thereafter produces the withholding outcome, with the authorization gate (300) written to the withheld state (310). A successor policy object declaring a larger aggregate retention (800) is not admitted while that foreclosure record stands.

Non-response is treated with equal care in the other direction. Where no acceptance arrives in her window, the embodiment appends a not-determinable outcome, and no counter of the counterparty moves, nothing adverse is written to the counterparty identity record (114) held for it, and no standing quantity of that party is modified. Where a bound counterparty has emitted a non-execution attestation (504) enumerating the very action class in which an acceptance would be produced, an abstention-suspended accumulation record holds that arc's contribution at the amount then standing, without discharging, redesignating, or reducing it, and without deriving any count or score from the withholding.

Boundaries she should plan around

For her purposes it is worth being exact about the boundary. The resolution described operates on the counterparty identity records (114) her deployment actually holds in the relevant scope partition. It surveys no party absent from her records, and a no-identified-counterparty outcome is a determination that her own records disclose no counterparty of that class, not an assertion that none exists. If her onboarding is thin, her unaddressed column will be thick, and that is a consequence of how she has provisioned identity rather than something the resolution repairs.

Nor does the embodiment make her shipper answer. Were the shipper simply to ignore every acknowledgment artifact her agent emits, the arc would stay pending, her register would keep standing at that amount, and her capacity to deviate would keep narrowing, while nothing adverse would be recorded against the shipper. That asymmetry is deliberate in the text she is reading, and she should plan for it as a cost of doing business rather than expect the mechanism to price it away.

Two further items sit outside this chapter as she is reading it. How the deviation was permitted in the first place, including the deductible and the likelihood test, is governed by a separate chapter, as is the pairing machinery her acceptance determinations ride on and the attestation type (506) her abstention handling references. She should read those before setting the values in her own signed policy object (112), since the numbers in the worked trace of this chapter are given by way of illustration only.

Disclosure Scope

This article is a technical description of subject matter disclosed in U.S. Provisional Application No. 64/117,812. It describes embodiments as they appear in that filing and uses the mechanism names, outcome terms, and reference numerals of that text. Nothing here characterizes the scope of any claim, and nothing here is an admission concerning the state of the art. The scenario, the party, and the deployment described are illustrative, and any numeric values referenced are those given in the filing by way of illustration.