DMV Numbers Pack v1.3: Reproducible Record-Content Adequacy Metrics for California Autonomous Vehicle Reporting, 2021–2025

This dataset provides the reproducible numerical evidence pack for a record-content adequacy analysis of California Department of Motor Vehicles (DMV) autonomous vehicle reporting records from 2021–2025. The pack supports the paper “Compliance Without Content: Pathway Variation and Schema Underspecification in Mandatory AI Event Reporting” (DfP26 #18 working manuscript). It analyses three California DMV reporting pathways: standard disengagement reports, driverless reports, and first-time filer records. The dataset does not evaluate autonomous vehicle safety, manufacturer performance, legal compliance, or operational quality. It evaluates public reporting records as governance data, asking whether formally filed records preserve description content relevant to post-event reconstruction. The pack includes:- a clean reproduction document summarising all locked claims;- a reproduction script;- results JSON;- summary CSV;- source-file integrity hashes;- claim definitions for zero-word rate, description uniqueness, top-5 repeated-description concentration, and pathway-level comparison. The core reproducible findings are:- standard pathway: N=45,477 records; zero-word rate 37.3%; 2024 zero-word rate 85.9%; description uniqueness 4.1%; top-5 description coverage 28.5%;- driverless pathway: N=5,266 records; zero-word rate 98.3%; description uniqueness 3.3%; top-5 description coverage 100.0%;- first-time filer pathway: N=30,273 records; zero-word rate 0.6%; description uniqueness 1.2%; top-5 description coverage 80.7%. These metrics are used as directly observable record-content indicators. They should not be interpreted as direct measures of autonomous vehicle safety, system performance, manufacturer quality, or substantive legal compliance. The underlying source records are public California DMV autonomous vehicle reporting records. Source-file SHA256 hashes are included to support reproducibility and file integrity checking.