Methods: two-dimension and norms coding of the AIME-Con 2026 program

Methods: two-dimension and norms coding of the AIME-Con 2026 program Everything needed to inspect or reproduce the opening chart (see `../README.md` for the results). | File | What it is | |---|---| | `codebook-2d.md` | Codebook v2: `product` and `profession`, the coders' only instructions | | `codebook-norms.md` | Codebook v2.1: the `norms` question | | `items-text.jsonl` | The 335 program items as the coders saw them: id, session, title, abstract text | | `sessions-text.jsonl` | The 83 sessions as the coders saw them | | `coder1-items.csv`, `coder1-norms.csv`, `coder1-sessions.csv` | Primary AI coder, blind to the v1 coding | | `second-coder-sample.json` | The blind stratified sample of 98 items (design and seed) | | `coder2-sample.csv`, `coder2-norms.csv` | Second AI coder, blind to the primary coder | | `agreement.json` | Agreement, Cohen's κ and confusion tables; also how v1 codes map to v2 cells | | `audit-2d.md` | The chair's audit sheet: coder disagreements and items whose classification moved | | `merge.py` | Rebuilds `../program-items-coded.csv` and the norms agreement | Both coders are instances of the same model family (Claude). Agreement therefore shows the codebook can be applied consistently; it is not independent human judgment.