Research object

Repository structure

site/                       Static prototype
research/cases/             6 decision cases (case_06 has NO AI in the scenario)
research/instruments/       Trace collection templates (confidence 0-100)
research/coding_scheme/     Construct definitions and coding manual
research/measurement_pipeline/ AI-coding flow, schema, and REAL example coded outputs
research/synthetic_data/    Reproducible generator + one pipeline-testing CSV
research/analysis_notebooks/ Runnable example analysis + validation scripts
validation/                 Human coding, agreement, drift, failure-mode protocols, thresholds
benchmark/                  Gold-standard benchmark plan and synthetic-data limits
docs/                       Proposal and submission notes

The repository includes real AI-coder output on six synthetic traces under research/measurement_pipeline/example_ai_coded_outputs/, plus a checker that validates the schema and shows the shape of the agreement computation. The near-perfect agreement it prints is a live demonstration of the synthetic-data circularity problem, not a validation result.

For Cloudflare Pages deployment:

Build command: none
Output directory: site