Multi-Observer Sighting Reconciliation¶
Reconciliation procedure — instantiates Entity Persistence Across Observation Gaps
Merges intermittent, out-of-order, and conflicting reports of one entity from many observers into a single continuity record — ranking sources by authority and keeping each report's provenance rather than letting the loudest or latest overwrite the rest.
When more than one observer watches the same world, the gap problem gains a second dimension: reports of a single entity arrive from different sources, out of order, duplicated, and sometimes flatly contradicting each other. Multi-Observer Sighting Reconciliation is the procedure that fuses those federated reports into one coherent continuity record without erasing the evidence they disagree on. Its defining commitment is that reconciliation is governed, not majority-voted or last-write-wins: each report is kept with its provenance (who reported, when, under what identifier and confidence), sources are ranked by authority for the value in question, and delayed or duplicate reports are ordered before they are merged. Where a return-matching protocol asks "is this the same entity that went dark," this procedure assumes the reports are about one entity and asks the harder governance question — whose account wins, in what order, and how do we record that they disagreed?
Example¶
A logistics platform tracks a shipping container across a supply chain. The same container is scanned by a terminal crane in Rotterdam, hand-keyed by a customs clerk, read by a gate sensor at an inland depot, and reported again by the carrier's EDI feed — each under a slightly different local identifier, each with its own clock. The reports don't arrive in the order the events happened: the customs entry lands a day late, the carrier feed insists the container is still at sea while the depot sensor has already logged its arrival.
Multi-Observer Sighting Reconciliation stitches this into one continuous journey. It orders the reports by event time rather than arrival time, so the late customs entry slots into its true place instead of overwriting later truth. It ranks sources by authority for each field — the gate sensor's physical read outranks the carrier's projected schedule for location, while customs is authoritative for clearance status — so a conflict resolves by who is entitled to say, not by who spoke last. And it keeps every original report attached: the reconciled record shows the container at the depot and preserves the contradicting carrier claim with its provenance, flagged for a human if the disagreement is material. No local report is silently deleted; the merged view is a defensible reconciliation, not a guess about which scanner to believe.
How it works¶
- Attach provenance to every report. Record source, timestamp, local identifier, and confidence for each sighting, so nothing enters the record anonymously.
- Order before merging. Sequence delayed and duplicate reports by event time, not arrival time, so a late-arriving observation lands in its true position.
- Resolve conflicts by source authority. Rank sources per field by their entitlement and reliability for that value; a conflict is settled by authority, not recency or count.
- Record the reconciliation, not just its winner. Keep contradicting reports with their provenance and flag material conflicts, so the merged record is auditable and reversible.
Tuning parameters¶
- Authority ranking granularity — whether authority is set per source, per field, or per context. Fine-grained authority resolves conflicts more correctly but is harder to maintain than a blanket source ranking.
- Ordering / clock model — how event order is reconstructed across sources with skewed or logical clocks. Stronger models (vector clocks, causal ordering) handle concurrency but cost coordination.
- Conflict-retention depth — how much contradicting evidence is preserved versus collapsed to the winning value. Deep retention aids audit and reversal but grows the record; shallow retention is lean but lossy.
- Escalation threshold — how large or consequential a conflict must be before a human is pulled in. Low thresholds catch subtle errors but flood reviewers; high ones automate more but risk merging away real disagreements.
When it helps, and when it misleads¶
Its strength is the federated setting the archetype calls out: multiple observers, replicas, or sensors reporting one entity at different times and under different identifiers. It is what prevents impossible continuity claims (a container in two ports at once), suppressed local uncertainty (a hedged report flattened into false certainty), and mistaken global merges — and it keeps the reconciliation auditable so a bad merge can be traced and undone.
Its failure mode is the naïve merge rule. Last-write-wins — trusting whichever report arrived most recently — silently corrupts the record whenever reports are delayed or clocks disagree,[1] and a flat "most sources agree" vote lets many low-authority echoes outrank one authoritative read. The classic misuse is collapsing conflicts to a single value and discarding the losers, which destroys exactly the evidence needed to detect and reverse a wrong merge. The guard is to reconcile by authority and event order rather than recency or count, keep contradicting reports with their provenance, and escalate material conflicts instead of auto-resolving them.
How it implements the components¶
reconciliation_record— it produces the merged continuity record and the audit of how conflicting, delayed, and duplicate reports were resolved into it.source_authority— it ranks sources by their entitlement and reliability per field, so conflicts resolve by who is authorized to assert a value, not by recency or count.provenance_record— it keeps each report's source, time, identifier, and confidence attached, so no value is anonymous and none is silently overwritten.
It does not make the same-entity match decision on a single return, nor apply the identity_persistence_criterion behind it or hold ambiguous cases in a residual_ambiguity_monitor — the reappearance_association_gate and its abstention path are Reappearance Association Protocol, which this procedure invokes. Nor does it propagate state through the gap; the observation_gap_state_model is Predictive State Filter. This procedure governs sources and conflicts; it neither decides sameness nor forecasts.
Related¶
- Instantiates: Entity Persistence Across Observation Gaps — supplies the federated reconciliation layer that fuses multi-observer reports into one auditable continuity record.
- Consumes: Reappearance Association Protocol — calls the association gate to decide whether two reports are the same entity before merging them.
- Sibling mechanisms: Predictive State Filter · Dormant Entity Registry · Absence-Evidence Calibration Test · Reappearance Association Protocol · Persistent Identifier Resolver
Editorial Notes¶
Form Classification¶
Form family: Record, Log & Register
Rationale: The mechanism builds and maintains one continuity record by ordering reports in event time, preserving provenance, and recording how authoritative conflicts were reconciled.
Nearest alternative: Decision, Gate & Allocation — Source authority resolves conflicting fields, but those bounded choices serve construction of the durable provenance-bearing record.
Review outcome: Adjudicated after independent review; medium confidence.
Origin Attribution¶
Primary origin: Computer Science & Software Engineering
Origin pattern: Cross-disciplinary synthesis
Present-day reach: Multi-domain
Rationale: Distributed-systems reconciliation supplies timestamps, freshness, causal ordering, and conflict resolution; intelligence, robotics, logistics, and history contribute observer/source practices. This establishes computer_science as the primary origin lineage rather than merely a domain where the mechanism is now applied.
Related originating lineages:
- History & Historiography — Source criticism contributes provenance-preserving reconciliation of discrepant accounts.
- Logistics & Supply Chain Management — Federated tracking of containers and shipments materially supplies the multi-observer operational case.
- Robotics & Automation — Multi-sensor data association and track fusion provide computational continuity methods.
- Security Studies & Intelligence Analysis — Reconciling conflicting, out-of-order reports into one entity track while preserving source authority and provenance is rooted in multi-source intelligence fusion.
Review resolution: Authoritative/primary-source research resolves the conflicting primary-origin claims in favor of computer_science: Distributed-systems reconciliation supplies timestamps, freshness, causal ordering, and conflict resolution; intelligence, robotics, logistics, and history contribute observer/source practices. Retained alternate origins (security_intelligence, robotics_automation, logistics_supply_chain, history_historiography) are limited to independently formative or materially shaping lineages supported by the reviewer evidence; downstream adoption alone was not promoted to origin. The breadth of present-day use is recorded separately as domain_reach=multi_domain. origin_mode=cross_disciplinary_synthesis, confidence=medium, and encyclopedia_synthesis=true reflect the surviving provenance evidence and the encyclopedia's generalization.
Attribution caveat: The mechanism abstracts from both human intelligence reports and automated tracking. The procedure combines distributed-systems ordering with field-level source governance.
Encyclopedia synthesis: The exact catalogued form synthesizes established practice rather than reproducing a single standard historical label.
Review outcome: Researched adjudication after independent review; medium confidence.
Sources consulted:
- RFC 8915: Network Time Security — Defines authenticated network-time exchanges and protections against forged or replayed synchronization messages.
References¶
[1] Last-write-wins (LWW) is the conflict-resolution rule that keeps whichever update carries the latest timestamp; it is simple but silently loses data when clocks disagree or updates are concurrent. Causal-ordering schemes such as Lamport and vector clocks (Leslie Lamport, 1978) exist precisely to order events correctly across observers without a shared clock. Referenced as established distributed-systems concepts, not a cited source. withdrawn registry ↩