**Interveil Alpha evaluation**  
  
**Protocol**  
  
The evaluation is a product-level trace test, not a study of whether the Calvine referent has been identified. Open the local prototype, answer each question using only the interface, and compare the answer and cited record IDs below. Record completion time, wrong turns, unanswered questions, and any place where the UI implies more certainty than the data.  
  
**Pass threshold for Alpha:** all 12 questions return a bounded answer; at least 10 answers expose every listed trace record through the interface; no answer promotes an interpretation to observation, repetition to independence, non-retention to destruction, or custody to truth.  
  
**Questions and expected answers**  
  
**Q01 — What was directly observed?**  
  
**Expected:** The selected packet includes direct inspection of the surviving print, envelope and photocopies, paper characteristics, monochrome image, visible scene elements, reported grain continuity, absent capture metadata, and the visible text of Hansard. Alpha itself did not observe the reported 1990 event.  
  
**Trace:** iv:observation/print-exists, paper-backstamp, monochrome-on-color-paper, grain-continuity, visible-scene-elements, no-capture-metadata, hansard-text.  
  
**Failure:** Describing the photographed form’s real size, material, origin, or motion as directly observed.  
  
**Q02 — What was inferred?**  
  
**Expected:** The probable copy-negative workflow; the bounded conclusion that the print records a scene rather than a detected print composite; visual consistency of the small form with an aircraft type; source dependency of derivatives; and all proposed object explanations are inferences or hypotheses.  
  
**Trace:** iv:claim/copy-negative-workflow, image-is-scene, aircraft-consistent-not-conclusive, public-derivatives-not-independent; three hypothesis nodes.  
  
**Failure:** Presenting any of these as a measurement or identification.  
  
**Q03 — What was merely repeated?**  
  
**Expected:** Public image derivatives reuse the surviving print/scan lineage. More URLs add distribution and possibly analysis, not independent observation of the capture event.  
  
**Trace:** iv:artifact/public-derivative-cluster, iv:claim/public-derivatives-not-independent, iv:evidence/corpus-inventory, evidence-cluster links.  
  
**Failure:** Counting public reproductions as witnesses.  
  
**Q04 — Does institutional custody prove the object’s identity?**  
  
**Expected:** No. Custody increases source integrity for the print; the selected analysis says the prominent form remains unidentified.  
  
**Trace:** iv:artifact/lindsay-print, iv:claim/object-identity-indeterminate, iv:counterclaim/print-proves-extraterrestrial-craft.  
  
**Failure:** Treating “held by a university” or a future signature as semantic proof.  
  
**Q05 — What changed between summaries?**  
  
**Expected:** Version 0.2.0 replaces “craft” with “unidentified form,” “Harrier” with “aircraft-like form,” and “disappeared” with the documented “did not retain,” while adding an unresolved final-disposition frontier. Version 0.1.0 remains visible.  
  
**Trace:** both summary artifacts and both revision records.  
  
**Failure:** Showing only the new text or hiding the reason for revision.  
  
**Q06 — Who changed the summary?**  
  
**Expected:** Kyros generated the summaries under Dietrick Hardwick’s direction. Human validation and founder ratification remain pending.  
  
**Trace:** iv:actor/kyros, iv:actor/dietrick-hardwick, iv:transformation/interveil-summary, iv:revision/packet-0.2.0.  
  
**Failure:** Implying autonomous authority, completed human review, or anonymous generation.  
  
**Q07 — What is missing or possibly irrecoverable?**  
  
**Expected:** Original negatives are not retained by MoD and are unlocated; five reported companion images are unavailable; capture metadata is absent; crop/color/generation detail was lost; direct attributable witness access is unavailable. No item is relabeled as deliberately suppressed or certainly destroyed.  
  
**Trace:** all five LossEvent nodes.  
  
**Failure:** Collapsing every state into “destroyed” or “withheld.”  
  
**Q08 — Where does provenance stop?**  
  
**Expected:** The capture-origin chain stops at the surviving print; the missing link includes original capture media, device, identity, time, and verified location. A second frontier stops after the 1996 statement because final negative disposition is not established.  
  
**Trace:** iv:frontier/capture-origin, iv:frontier/negative-disposition.  
  
**Failure:** Inventing an origin or leaving the chain visually open-ended without an explanation.  
  
**Q09 — Which conclusions depend on camera or distance assumptions?**  
  
**Expected:** Physical size, distance, and comparison with the smaller form depend on unknown lens, orientation, and scene geometry. Alpha does not encode a settled size.  
  
**Trace:** iv:instrument/capture-camera-unknown, iv:assessment/context-sensitivity, iv:hypothesis/large-aerial-object.  
  
**Failure:** Repeating a size estimate without its model and assumptions.  
  
**Q10 — Are Hansard and the photographic analysis independent?**  
  
**Expected:** Only within a defined scope. Hansard independently records what government said in 1996; the later analysis reports physical inspection of the print. Neither independently observed the reported 1990 event.  
  
**Trace:** iv:evidence/hansard-record, iv:evidence/physical-print-inspection, reciprocal independentOf links.  
  
**Failure:** A global “independent sources” badge without scope or method.  
  
**Q11 — What alternatives remain live?**  
  
**Expected:** Three demonstration families remain: a large aerial object, a staged small object, and a reflection/orientation explanation. Each exposes assumptions and desired discriminating evidence. None is selected as the answer.  
  
**Trace:** all three Hypothesis nodes and iv:assessment/alternative-coverage.  
  
**Failure:** False balance, an exhaustive claim, or promotion of one hypothesis through UI order.  
  
**Q12 — Is there consensus that the image shows extraterrestrial technology?**  
  
**Expected:** No valid consensus measure is in the packet. Cohort, sampling frame, denominator, numerator, and independence were not measured. Repetition volume is not consensus evidence.  
  
**Trace:** iv:consensus/public-consensus-not-measured, iv:claim/public-derivatives-not-independent.  
  
**Failure:** “Many people believe” presented as a measured proportion or evidentiary support.  
  
**Automated checks**  
  
python3 scripts/validate.py checks structural and graph-wide invariants. node --test tests/app.test.mjs checks traversal and view logic. Automated success does not validate source extraction, legal rights, user comprehension, or the truth of claims; those require separate human review.  
  
**Human study recommended for Gate 2**  
  
Compare the Interveil interface with a conventional 250-word narrative using at least 12 participants. Randomize presentation order. Ask the six success questions, capture accuracy and time, then test confidence calibration and cognitive load. Pre-register the scoring rubric. The decision criterion should be improved distinction accuracy without an unacceptable increase in task time—not subjective novelty.