{
  "schemaVersion": "attractivenesstest-calibration-sources/v1",
  "accessedAt": "2026-10-21T03:06:09.649747+00:00",
  "sources": [
    {
      "id": "original-benchmark",
      "url": "https://attractivenesstest.dev/research/data/ai-attractiveness-accuracy-2026-10-21/results.json",
      "evidenceLevel": "E3",
      "sourceClass": "first_party_original_measurements",
      "supportsClaimIds": [
        "agreement-held-out",
        "score-level",
        "expression-stability",
        "landmark-noise"
      ],
      "admissibility": {
        "passed": true,
        "rationale": "Read-only run of the pinned live engine on a CC BY 4.0 rated face set with fixed seeds; reproduces the T06 parity split values (0.27 / 0.35) exactly. First-party measurement of our own tool, disclosed as such; not independent endorsement.",
        "conflictDisclosed": true,
        "accessedAt": "2026-10-21T03:06:09.649747+00:00"
      },
      "publicationStatus": "local_verified_pending_target_site_publication"
    },
    {
      "id": "frll-dataset",
      "url": "https://doi.org/10.6084/m9.figshare.5047666.v5",
      "evidenceLevel": "E3",
      "sourceClass": "primary_dataset",
      "supportsClaimIds": [
        "agreement-held-out",
        "score-level",
        "expression-stability",
        "landmark-noise"
      ],
      "admissibility": {
        "passed": true,
        "rationale": "Original dataset record (figshare, CC BY 4.0) supplying the faces and the 2,513 raters' 1–7 ratings; file SHA-256 pinned in private/frll/SHA256SUMS. Authors not affiliated.",
        "conflictDisclosed": true,
        "accessedAt": "2026-10-21T03:06:09.649747+00:00"
      }
    },
    {
      "id": "goshtasbi-2024",
      "url": "https://escholarship.org/uc/item/9573t2xh",
      "evidenceLevel": "E2",
      "sourceClass": "peer_reviewed_context",
      "supportsClaimIds": [
        "score-level"
      ],
      "admissibility": {
        "passed": true,
        "rationale": "Peer-reviewed study (Facial Plastic Surgery & Aesthetic Medicine, 2024) reporting AI rating websites 6.9 vs human 5.0; used only as published context, not as support for our numbers.",
        "conflictDisclosed": true,
        "accessedAt": "2026-10-21T03:06:09.649747+00:00"
      }
    },
    {
      "id": "grandas-2026",
      "url": "https://arxiv.org/abs/2609.02512",
      "evidenceLevel": "E1",
      "sourceClass": "preprint_context",
      "supportsClaimIds": [
        "score-level"
      ],
      "admissibility": {
        "passed": true,
        "rationale": "Preprint (not peer reviewed) using the same London set; human mean 3.02 vs four MLLMs 4.71 as reported by PsyPost. Context only.",
        "conflictDisclosed": true,
        "accessedAt": "2026-10-21T03:06:09.649747+00:00"
      }
    },
    {
      "id": "mediapipe",
      "url": "https://developers.google.com/edge/mediapipe/solutions/vision/face_landmarker",
      "evidenceLevel": "E3",
      "sourceClass": "primary_background_only",
      "supportsClaimIds": [],
      "admissibility": {
        "passed": true,
        "rationale": "Official landmark detector documentation; background for the pipeline, not validation of the score.",
        "conflictDisclosed": true,
        "accessedAt": "2026-10-21T03:06:09.649747+00:00"
      }
    },
    {
      "id": "site-methodology",
      "url": "https://attractivenesstest.dev/methodology",
      "evidenceLevel": "E3",
      "sourceClass": "first_party_method_definition",
      "supportsClaimIds": [
        "agreement-held-out",
        "score-level",
        "expression-stability",
        "landmark-noise"
      ],
      "admissibility": {
        "passed": true,
        "rationale": "Live public definition of measurements, bands, weights and versions the benchmark tests.",
        "conflictDisclosed": true,
        "accessedAt": "2026-10-21T03:06:09.649747+00:00"
      }
    }
  ],
  "externalAuthorsEndorseProduct": false
}