spb/wp17_uqo

1# Codebook — FIRC-0002 N1 human-validation frames (seed 20261105)23Files: `validation_coding_sheet_blind.csv` (A stratified 400 + D top-by-year 100, shuffled, no scores shown — code this one),4`validation_stratified_key.csv` (same A rows with decile, scores and keyword counts — for the analyst, not the coders),5`validation_top_firms.csv` (B: 20 top chunks x 30 top firms of Table 4, IDA / HE flagged in `special`),6`validation_median_firms.csv` (C: 20 top chunks x 30 median firms), `validation_top_by_year.csv` (D with scores).78Columns to fill (two readers, independently; kappa computed on `label_ai_substantive`):9- `label_ai_substantive`: 1 if the passage describes the FIRM'S OWN adoption, deployment, investment, product or10  infrastructure in artificial intelligence / machine learning (including generative AI); 0 otherwise. Generic risk-factor11  boilerplate, third-party / customer AI, pure automation without AI, cybersecurity without AI, generic data centres or12  cloud capacity without AI purpose are 0.13- `label_ai_mention`: 1 if AI / machine learning is mentioned at all (any register), 0 otherwise.14- `label_boilerplate`: 1 if the passage is standardized language repeated across filings (risk factors, safe-harbor15  statements, exhibit lists, accounting policies), 0 if firm-specific narrative.16- `label_topic`: one of AI | automation | cybersecurity | data_center_or_cloud | IT_generic | analytics_or_data |17  utility_load_or_energy | financial_boilerplate | other.18- `reader`: initials. `notes`: free text.1920Precision by decile = mean of `label_ai_substantive` within each 2012 / 2024 decile (join on `sample_id` to the key file);21precision by sector from the `sector` column; the B/C frames give the top-firm precision (IDA / HE in particular).22Text is clipped at 2,000 characters (`text_truncated_2000` in the key files); the full chunk is in `filing_chunks` by `chunk_id`.23