The AUSPEX Confidence Rubric: How We Grade Incidents

Most UAP databases publish a single-axis credibility badge — CONFIRMED, HIGH, MEDIUM, LOW — and never define what those labels mean. The result is that Rendlesham Forest and the Hill abduction end up in the same tier, and neither the skeptic nor the believer knows what to do with it. Skeptics read the badges as credulity; believers read them as validation. Both are wrong, because the badge is not actually making a claim.

This page publishes the rubric AUSPEX uses to grade every incident in the corpus. It is deliberately two-axis because a single axis cannot express the distinction that matters most in UAP research: the difference between "this event happened" and "this event was anomalous." Rendlesham definitely happened — there is an official memo from the Deputy Base Commander. Whether what Colonel Halt saw was a genuinely anomalous craft is a separate question that rests on different evidence entirely. Any rubric that flattens those into one number is misleading by construction.

Documentation asks: what record of the event exists? This axis grades the strength of the claim that the event occurred at all, independent of what the event was.

• ◆OFFICIAL — a government, military, or state investigative body has an on-the-record account of the event. Examples: the Halt memo (Rendlesham), the Nimitz FLIR release (2017), the Fravor–Slaight radar tracks, congressional testimony under oath, GEIPAN case-file numbers, FBI Vault Part memoranda.
• ◆PRESS — an investigative organization with a documented field-investigation protocol (typically MUFON) has an on-file case report, or the event was covered by mainstream press with contemporaneous documentation. No primary official record.
• ◆TESTIMONY — the only surviving account of the event comes from witnesses themselves, with no corroborating official body or investigative organization. Includes most raw civilian self-reports (NUFORC-tier submissions).

Evidence class asks: what makes the event a UAP claim rather than a mundane misidentification? This axis grades the strength of the anomaly claim, independent of who documented it.

• ◆INSTRUMENTED — sensors recorded the event. Radar returns, FLIR/IR imaging, photographic or video capture, gun-camera footage, or other instrument data exists.
• ◆PHYSICAL TRACE — physical evidence at the site. Landing marks, recovered debris, radiation, or measurable environmental effects (dead vegetation, magnetized ground, etc.).
• ◆MULTI-WITNESS — multiple independent witnesses (typically four or more, or with independent-source corroboration) reported the same event, without instrumented or physical evidence to anchor it.
• ◆SINGLE-WITNESS — a small number of witnesses reported the event with no instrumented or physical evidence. The interpretive burden rests on the account itself.

The two axes are independent. An incident can be OFFICIAL × INSTRUMENTED (USS Nimitz Tic-Tac: radar tracks, FLIR video, Pentagon-released), OFFICIAL × MULTI-WITNESS (Washington 1952: PROJECT BLUEBOOK file, multiple pilots, air-traffic controllers), or OFFICIAL × SINGLE-WITNESS (many field-officer reports in the FBI Vault). It can also be TESTIMONY × MULTI-WITNESS (mass sightings with no institutional follow-up) or TESTIMONY × SINGLE-WITNESS (the weakest grade — one person, no record).

The badge next to every incident on AUSPEX shows both axes. Hovering the badge shows the definition. The pairing tells you exactly what kind of claim the entry is making before you spend attention on the details.

It is not a truth verdict. OFFICIAL × INSTRUMENTED does not mean "definitely a UAP." It means "the event is well-documented and instrument data exists," which is the strongest evidentiary posture the field currently produces. Whether the instrument data actually shows something anomalous — and not, say, a sensor artifact or a misidentified conventional object — is a separate analytical question that lives in the case study, not in the badge.

It is not a legal standard. The rubric describes the evidentiary posture of an entry in a research corpus, not the level of proof required in a courtroom or scientific journal.

It is not fixed forever. If new evidence classes emerge — biological samples that survive independent analysis, sustained multi-sensor tracks under scientific protocol — the rubric will grow.

Every incident in the AUSPEX corpus carries structured fields: source, witnesses, independentSources, radarConfirmed, physicalEvidence, photoVideoEvidence, officialInvestigation. The rubric grade is computed from these fields, not hand-tagged, which means:

1. ◆Consistency. Every incident is graded by the same function. There is no "some incidents got the CONFIRMED badge because the analyst was in a good mood that day."
2. ◆Auditability. If you disagree with an incident's grade, you can trace it directly to the underlying fields and challenge those. The rubric is inspectable, not opaque.
3. ◆Data-driven revision. When new evidence is discovered and the underlying fields are updated (e.g., a new sensor track surfaces, or a witness count is corrected), the badge updates automatically. No stale hand-annotations.

The mapping function lives in src/lib/rubric.ts — it is a single file, ~90 lines, and covers every incident in the corpus.

The existing single-axis credibility field (CONFIRMED / HIGH / MEDIUM / LOW / UNVERIFIED) is retained for backward compatibility and for the color-coding on the globe and maps, where a single-hue gradient is more useful than two pills side-by-side. But wherever a badge is displayed — incident lists, case studies, Atlas popups — the two-axis rubric is what you should read.

Over time, we expect credibility to become a computed summary of the rubric axes (roughly: OFFICIAL × INSTRUMENTED → CONFIRMED, OFFICIAL × MULTI-WITNESS or INSTRUMENTED alone → HIGH, and so on) rather than a hand-set field. That refactor is a separate follow-up.

• ◆USS Nimitz Tic-Tac (2004) → OFFICIAL × INSTRUMENTED. Radar tracks, FLIR video, Pentagon release, pilot testimony under oath, AARO case file.
• ◆Rendlesham Forest (1980) → OFFICIAL × PHYSICAL TRACE. Halt memo, USAF personnel, physical landing traces, no radar/FLIR.
• ◆Washington D.C. Carousel (1952) → OFFICIAL × INSTRUMENTED. Radar at Washington National + Andrews AFB, PROJECT BLUEBOOK file, multiple pilots.
• ◆Phoenix Lights (1997) → PRESS × MULTI-WITNESS. Extensive press coverage, thousands of witnesses, video evidence disputed as flare misidentification for part of the event; no official investigation of the anomaly claim itself.
• ◆Farmington mass sighting (1950) → PRESS × MULTI-WITNESS. Local paper coverage, hundreds of witnesses across three days, no official record.
• ◆Kenneth Arnold (1947) → OFFICIAL × SINGLE-WITNESS. Single pilot report, PROJECT BLUEBOOK file, no radar/FLIR/trace — the report that coined "flying saucer."

The rubric cannot distinguish between:

• ◆an INSTRUMENTED case where the sensor data is genuinely anomalous vs. one where the sensor data turned out to be an artifact under later scrutiny (Gimbal-style rotation-of-glare debates);
• ◆a MULTI-WITNESS case where the witnesses were genuinely independent vs. one where they shared expectations and one dominant account propagated;
• ◆an OFFICIAL record where the officiality reflects careful investigation vs. one where it reflects institutional CYA (an on-file memo that says "this happened" but did no follow-up).

These are exactly the questions the case-study pages exist to address, and they are exactly why the rubric badge on its own is not a truth verdict. The badge is the entry point to the analysis, not a substitute for it.

The rubric is under revision. If you have a proposed refinement — an evidence class we should split, a source that should map differently on the Documentation axis, a case where the derivation is clearly wrong — the code lives at src/lib/rubric.ts and the discussion belongs at /contact.