Two Readings of One Recorded Latency

Paragraph [0358] of U.S. Provisional Application No. 64/117,812 opens on a measurement problem: the semantic agent (100) sets out to distinguish whether an observed confirmation latency is caused by a counterparty or by its domain, and does so by varying a property of its own dispatch rather than by passive observation. Paragraph [0359] converts that intervention into a single quantity. Over a matched class, the agent computes a first closure latency for interactions closed upon a receipt the counterparty controls and a second closure latency for interactions closed upon a receipt it does not control, each as the arithmetic mean of dispatch-to-closure intervals. It normalizes the first against a channel baseline by subtraction bounded at zero, and takes the latency differential as the normalized first latency less the second.

One number therefore reaches the decision stage, and what may be concluded from it is the subject of paragraph [0360].

Consider what a lone threshold test can express. Above the threshold, latency is attributed to the counterparty. Below it, whatever is left over is a remainder rather than a finding. That remainder is not one situation, because the differential is computed entirely from the two closure paths of the class in front of the agent. How other matched classes are closing on their second path is a different quantity, held in different records, and the filed text names it separately: the second-path baseline, the mean second latency across the agent's other matched classes with the counterparty under evaluation excluded. Paragraph [0360] makes exoneration turn on that separate quantity, and it says so in its own closing sentence, which places exoneration on the domain baseline and not on the differential alone.

Sequence matters as well as comparator. A matched class may hold few closures on one of its two paths, and the filed evaluation tests per-path volume before any threshold is consulted, appending a not-determinable outcome where either path falls short.

Ordered Branches, and a Comparator for Exoneration

Paragraph [0360] specifies an ordered branch sequence over the second-path baseline. Four conditions are evaluated in a fixed order, and each later branch is reached only where the earlier conditions do not hold.

The comparator. The second-path baseline is computed as the mean second latency across the agent's other matched classes, excluding the counterparty under evaluation. Second-path closures are those closing upon a receipt the counterparty does not control, per [0358]. Because the exclusion is written into the definition, the counterparty being measured contributes no second latency to the baseline it is measured against.

First branch, per-path volume. Where either path holds fewer than a declared minimum volume, a not-determinable outcome is appended and no attribution issues. The condition is stated of either path rather than of the class taken as a whole. Since the remaining branches are reached only otherwise, a class failing here is compared neither against the threshold nor against the baseline.

Second branch, attribution to the counterparty. Where the volume condition is met and the differential exceeds a declared threshold, latency is attributed to the counterparty by an attribute on the counterparty identity record (114). Within the four branches, this is the only outcome that records something against the counterparty.

Third branch, attribution to the domain. Where the volume condition is met, the differential does not exceed the threshold, and the second closure latency is itself elevated above the second-path baseline by a declared margin, latency is attributed to the domain and an exoneration record is appended, with nothing adverse written to the counterparty. The operative comparison here is of the second closure latency against a comparator drawn from other matched classes, not of the differential against the threshold.

Fourth branch, no adverse signal. Otherwise a no-adverse-signal outcome is appended.

Read together, the four conditions leave every evaluation ending in exactly one appended outcome. Attribution to the counterparty demands that one quantity exceed a declared threshold. Attribution to the domain demands that a different quantity exceed a different comparator by a declared margin. Neither result is the leftover of the other: the leftover is the no-adverse-signal outcome, which is itself appended and is a different record from the not-determinable outcome of the first branch.

Two adjacent paragraphs constrain what the sequence may conclude. Paragraph [0361] bounds the attribution threshold below by the standard deviation of the recorded durations of the qualifying closures from which the channel baseline is computed, being the positive square root of their variance and expressed in the units of the baseline and threshold. The filed reasoning is that a threshold below that bound attributes ordinary channel variation to the counterparty, and that the bound ties the decision threshold to the measured dispersion of the channel itself, so an attribution issues only where the differential exceeds the noise floor of the channel it is measured against.

Paragraph [0362] can withdraw a class from the sequence altogether. Over a matched class, the agent computes per closure path a measure of counterparty conduct other than latency, comprising the rate of non-closure, of returned exceptions, and of amendment requests. Responsive to that measure differing between paths by more than a declared manipulation margin, the agent appends a manipulation attribute to the counterparty identity record (114), distinct from the latency attribution, and neither attributes nor exonerates the latency of that class. The filed ground is that a difference across paths meant to be indistinguishable establishes that the assignment was in fact distinguishable to the counterparty, invalidating that class's latency evidence.

Where the third branch does issue, paragraph [0363] carries the result outward. The persistent semantic agent (100) appends an exoneration record identifying the matched class, both closure latencies, and the threshold applied, and transmits it to that counterparty as a governed observation. A further persistent semantic agent (100) presented the record by that counterparty admits it as an input to a first-encounter determination alone, verifying that it is signed by the agent that appended it, that the matched class is identified by the feature tuple recorded for it, and that its recorded time falls within a declared currency interval. So admitted, the record is written into the counterparty identity record (114) as a presented record, modifies no scoped integrity vector (106), and increments no counter. A record failing verification produces the not-determinable determination (126), with nothing adverse appended.

What the Text Fixes and What Policy Declares

The filed disclosure settles structure, ordering, and one relation between quantities. It states no numeric value, and none should be inferred from this article.

Fixed by the disclosure:

  • Branch order. Per-path volume, then the differential against the threshold, then the second closure latency against the second-path baseline, then the residual outcome.
  • Second-path baseline construction. The mean second latency across the agent's other matched classes, with the counterparty under evaluation excluded.
  • Third-branch operand. The second closure latency, compared against the second-path baseline, rather than the differential compared against the threshold.
  • Central-tendency measure. The arithmetic mean of dispatch-to-closure intervals, applied to both closure paths under [0359].
  • Normalization. Subtraction bounded at zero against a channel baseline pooled from first-path closures admitted by a qualifying predicate evaluated on the agent's own records alone, consulting no counterparty-supplied score.
  • Where outcomes land. Attribution and manipulation attributes are written to the counterparty identity record (114); a presented exoneration record is recorded there as a presented record, modifies no scoped integrity vector (106), and increments no counter.

Declared, with no value stated in the filing:

  • Minimum per-path volume. Tested on each closure path separately in the first branch.
  • Attribution threshold. Carrying one stated relation, the standard-deviation lower bound of [0361], expressed in the units of the baseline and threshold.
  • Margin above the second-path baseline. The amount by which the second closure latency must be elevated before latency is attributed to the domain.
  • Manipulation margin. The cross-path difference in non-latency conduct above which the class is neither attributed nor exonerated.
  • Currency interval. Governing whether a presented exoneration record is admitted at a receiving agent.
  • Qualifying-predicate quantities. From [0359], a declared minimum of first-path closures, each within a declared responsiveness bound.

Position Within Section 10.2

Section 10.2 of the filing carries authorization decay, scope-partitioned containment, and interventional attribution together, and within the attribution group this branch is the decision stage. Upstream, [0358] supplies the intervention: matched interactions identical in action class, stakes, urgency, and scope partition, routed to close upon a receipt the counterparty controls or upon one it does not, with paths assigned by deterministic alternation or by a pseudorandom selection seeded from a value committed to the append-only lineage field (104) before the first assignment, and with the counterparty not notified. Paragraph [0359] supplies the measured inputs. Paragraph [0360] turns those inputs into an outcome, [0361] floors the threshold, [0362] can invalidate the class, and [0363] makes one outcome portable.

The outcomes land in records the filing already defines. Attributions are attributes on the counterparty identity record (114). A presented exoneration record is recorded there too, but [0363] confines it to a first-encounter determination and bars it from modifying the scoped integrity vector (106) or incrementing any counter.

Recording and capability are likewise kept apart across the section. Paragraph [0355] appends a containment determination while stating that the authorization gate (300) is not written by reason of suspension, containment withdrawing only reach into other partitions. The outcomes [0360] specifies are, in the same manner, appended records and an attribute on an identity record.

Downstream, the record produced by the third branch enters the first-encounter admission that Section 10.4 of the filing develops, admitted for that purpose alone and only where the three verification conditions of [0363] are satisfied.

Adjacent Categories and How They Differ

Several established categories sit near this mechanism. The differences below are structural rather than matters of tuning.

Service-level and threshold monitoring compares a recorded duration against a limit and raises a state when the limit is crossed. Its in-bound condition is a remainder: whatever failed to trip. The branch of [0360] instead appends four distinct outcomes, and its exoneration outcome carries its own condition against its own comparator.

Peer-baseline and cohort anomaly detection compares a subject against a population statistic, commonly one that includes the subject, which suffices when the question is ranking within the population. The second-path baseline is defined with the counterparty under evaluation excluded, and the comparison asks whether that counterparty's second closure latency stands above the level recorded in the agent's other matched classes.

Statistical testing convention treats a failure to reach a threshold as a failure to establish an effect rather than as evidence of its absence. Here the distinction is carried by the record structure: a differential at or below the threshold leads to an exoneration outcome or a no-adverse-signal outcome according to a second and independent condition, so the two are never merged.

Reputation and trust systems commonly express exoneration as the absence of a negative entry, and an absence is not itself transmissible. The exoneration record of [0363] is a signed artifact identifying the matched class and carrying both closure latencies and the threshold applied, transmitted to the counterparty as a governed observation and admitted at a receiving agent only on signature, class identity, and currency checks.

Dispute and chargeback processes resolve attribution by submitting evidence to an adjudicating authority both parties recognize. No adjudicator appears in this mechanism. The agent varies a property of its own dispatch, evaluates its own records, and signs the record it appends.

None of these comparisons turns on a named product, party, or system, and nothing here should be read as asserting that any such system infringes. The comparisons are of mechanism.

Disclosure Scope

This article describes subject matter disclosed in U.S. Provisional Application No. 64/117,812, Section 10.2, principally paragraph [0360], "Attribution-versus-exoneration branch with domain baseline," together with the paragraphs on which that branch depends: the receipt-path control of [0358], the latency differential and channel-baseline normalization of [0359], the standard-deviation lower bound on the attribution threshold of [0361], the distinguishability check of [0362], and the portable exoneration record of [0363].

What is disclosed is the structure: a second-path baseline computed as the mean second latency across the agent's other matched classes with the counterparty under evaluation excluded, and an ordered branch sequence running a per-path minimum volume gate yielding a not-determinable outcome, then a differential-against-threshold attribution to the counterparty, then an elevation-against-baseline attribution to the domain accompanied by an exoneration record, and otherwise a no-adverse-signal outcome.

No numeric parameter value is disclosed here or fixed by this publication. Each quantity the branch consumes is declared, and the filing states a relation among two of them rather than any magnitude. Worked numeric examples, alternative comparator constructions, deployment illustrations, and any further sub-mechanism or variant not set out in the filed provisional fall outside the scope of this article, and nothing in it should be read to place such material in the public domain.