{
  "sovereign_identifier": "EROSION-EMPIRICAL-AUDIT-01-CONTINGENCY-20260714",
  "governing_deposit": "EA-EROSION-EMPIRICAL-01 v0.1",
  "audit_date": "2026-07-14",
  "source_containers": [
    {
      "identifier": "ZENODO-DELETION-EXPORT-20260710",
      "role": "primary"
    },
    {
      "identifier": "ZENODO-DELETION-EXPORT-20260607",
      "role": "reference"
    }
  ],
  "population": {
    "description": "All rows of ZENODO-DELETION-EXPORT-20260710 with removal_date in 2026 and non-empty citation_text",
    "n": 100313,
    "filter_rationale": "Restricted to 2026 to establish the current classifier regime's behavior. Restricted to non-empty citation_text because spam-classified rows have citation_text stripped (\u00a77 finding) and cannot be classified on either dimension."
  },
  "classifiers": {
    "ai_signal": {
      "detection": "case-insensitive substring match on citation_text",
      "term_panel": [
        "chatgpt",
        " claude ",
        "gpt-4",
        "gpt4",
        "gpt ",
        "llm ",
        "large language model",
        "artificial intelligence",
        " ai ",
        "ai system",
        "ai review",
        "ai evolution",
        "ai coauthor",
        "gemini",
        "grok",
        "openai",
        "anthropic",
        "ai governance",
        "cognitive orchestration",
        "triune superintelligence",
        "agentic ai",
        "ai-assisted",
        "ai assisted"
      ]
    },
    "institutional_signal": {
      "detection": "case-insensitive substring match on citation_text",
      "term_panel": [
        "university",
        "universit",
        "universidad",
        "institut",
        " college ",
        "laboratory",
        " cern ",
        " nasa",
        " cnrs",
        "max planck",
        "lbnl",
        "ornl",
        "inria",
        "polytech",
        "school of",
        "department of"
      ]
    }
  },
  "contingency_matrix": {
    "cells": {
      "ai_and_institutional": {
        "count": 60,
        "row": "institutional",
        "col": "ai_signal"
      },
      "ai_and_not_institutional": {
        "count": 5965,
        "row": "not_institutional",
        "col": "ai_signal"
      },
      "not_ai_and_institutional": {
        "count": 1136,
        "row": "institutional",
        "col": "not_ai_signal"
      },
      "not_ai_and_not_institutional": {
        "count": 93152,
        "row": "not_institutional",
        "col": "not_ai_signal"
      }
    },
    "row_marginals": {
      "institutional": 1196,
      "not_institutional": 99117
    },
    "column_marginals": {
      "ai_signal": 6025,
      "not_ai_signal": 94288
    },
    "total": 100313
  },
  "derived_ratios": {
    "ai_signaled_institutional_fraction": 0.00996,
    "non_ai_institutional_fraction": 0.01205,
    "alive_side_institutional_fraction_estimate": 0.311,
    "compression_ratio": 31.2,
    "compression_interpretation": "Institutional AI-augmented composition is present in the alive-side sample at approximately 30x its representation in the AI-signaled deletion pool."
  },
  "underdetection_notes": {
    "field_scope": "Signal detection uses citation_text field only (essentially the record's citation string). Full-text detection of both AI signals and institutional signals would find higher absolute rates. Underdetection is expected on both dimensions.",
    "expected_symmetry": "Undercounting is expected to be roughly symmetric between the two dimensions (both use short-text substring matching on the same field). The compression ratio is therefore robust to symmetric undercounting.",
    "asymmetric_risks": "Independent-institution self-signaling in citation_text (e.g., 'EPINOVA LLC') is not caught by the institutional term panel. This means the deletion-side institutional presence estimate could be slightly higher under a more permissive institutional panel. The Wu case study shows an EPINOVA-affiliated uploader whose deletion event would not be caught by the current panel."
  },
  "reproducibility": {
    "script": "generate_audit.py",
    "input_files": [
      "deleted-head-20260710.csv.gz",
      "deleted-20260607.csv.gz"
    ],
    "runtime_seconds_approximate": 30
  }
}