{
  "schema":"ULTRACON_CON_VALIDATION_PILOT_RESULT_V1",
  "pilot_id":"CON-V42-PILOT-01",
  "date":"2026-09-14",
  "status":"PILOT_COMPLETED_PRE_ADJUDICATION_PRESERVED",
  "release_state_after_pilot":"REVISED",
  "coders":[
    {"id":"A","system":"GPT-5.6 Sol","role":"blind computational coder","file":"pilot-01-coder-a-gpt56sol.json"},
    {"id":"B","system":"Gemini 3.6 Flash","role":"blind computational coder","transport":"T^AI Bridge / Gemini Developer API","file":"pilot-01-coder-b-gemini36flash.json"}
  ],
  "independence_scope":"The two first-pass outputs were produced by separate model systems. Coder B received only the frozen packet/codebook; coder A output was not included. This is model-to-model coding independence, not human inter-rater validation.",
  "sample":{"units":14,"admitted_by_both":9,"out_by_both":5,"packet":"pilot-01-packet.json"},
  "agreement":{
    "admission":{"raw_agreement":1.0,"agreements":14,"n":14,"wilson_95":[0.7847,1.0],"krippendorff_alpha_nominal":1.0,"disagreements":[]},
    "primary_neighbor":{"raw_agreement":0.9286,"agreements":13,"n":14,"wilson_95":[0.6853,0.9873],"krippendorff_alpha_nominal":0.9213,"disagreements":[{"unit":"P09","coder_A":"N05_FOOLISHNESS","coder_B":"N11_NONCONFORMITY"}]},
    "audit_state_all_units":{"raw_agreement":0.9286,"agreements":13,"n":14,"wilson_95":[0.6853,0.9873],"krippendorff_alpha_nominal":0.9115,"disagreements":[{"unit":"P09","coder_A":"INDETERMINATE","coder_B":"SUPPORTED"}]},
    "audit_state_admitted_only":{"raw_agreement":0.8889,"agreements":8,"n":9,"wilson_95":[0.5650,0.9801],"krippendorff_alpha_nominal":0.8640,"disagreements":[{"unit":"P09","coder_A":"INDETERMINATE","coder_B":"SUPPORTED"}]}
  },
  "disagreement_matrix":{
    "admission":{},
    "primary_neighbor":{"N05_FOOLISHNESS__vs__N11_NONCONFORMITY":1},
    "audit_state":{"INDETERMINATE__vs__SUPPORTED":1}
  },
  "interpretation":[
    "The admission boundary survived this small contradictory set with complete agreement, including five OUT cases designed to test false-positive field inflation.",
    "The only disagreement occurred on the deliberately performative court-jester case P09.",
    "P09 reveals a codebook ambiguity: the current audit field does not clearly separate a documented social ROLE label ('the fool') from a TRAIT/COMPETENCE inference that the performer actually lacks judgment.",
    "The disagreement should not be erased by forced consensus. It triggers a codebook revision introducing attribution_type and a rule separating role-designation validity from trait-inference validity.",
    "The small sample makes all agreement estimates provisional; confidence intervals remain wide. No replication claim is warranted."
  ],
  "stress_test_coverage":{"VT01_ADMISSION_FALSE_POSITIVES":"PARTIAL_RUN","VT02_LABEL_SWAP":"NOT_RUN","VT03_SOURCE_BLIND":"NOT_RUN","VT04_EX_ANTE_EX_POST":"PARTIAL_RUN","VT05_COUNTERCASE_SENSITIVITY":"PARTIAL_RUN","VT06_RIVAL_HYPOTHESIS":"NOT_RUN","VT07_REFLEX_REVERSAL":"NOT_RUN","VT08_PROVENANCE_COMPLETENESS":"NOT_RUN"},
  "codebook_revision_required":{
    "field":"attribution_type",
    "allowed":["TRAIT","CONDUCT","ROLE","INSTITUTIONAL_CLASSIFICATION","SYSTEM_OUTPUT","OTHER"],
    "rule":"When attribution_type=ROLE, distinguish whether the social role/title is documented from whether a trait or competence deficit is inferred. A role label may be documented while the trait inference remains unsupported or indeterminate.",
    "trigger_unit":"P09"
  },
  "non_claims":["agreement does not establish truth","agreement does not establish cultural validity","two model coders do not substitute for independent human coders","this pilot is not a replication","P09 disagreement is preserved as information rather than treated as noise"]
}