Academic Nudge: behaviorally calibrated choice architecture for sequential course recommendation — replication data and code

Contents
- `code/` — complete pipeline (dataset construction, BERT4Rec encoder, conditional-logit calibration, counterfactual with common random numbers, comparators, bootstrap, aggregation, figures, manuscript generation). Execution order in `code/README.md`; environment in `code/requirements.txt`. License: MIT.
- `data_public/` — course- and program-level data and the aggregated result tables (R1–R14, main as-of construction and retrospective sensitivity). No individual-level records. License: CC BY 4.0. Codebook in `DATA_README.md`.
- `data_synthetic/` — synthetic trajectories with the schema of the controlled-access tier, for executing the code end to end only. License: CC BY 4.0.
- `DATA_README.md` — codebook of all data files, pseudonymization procedure and description of the controlled-access tier.

Controlled-access tier
Pseudonymized student–term records are not part of this deposit. They are available under a data-use agreement with Universidad del Magdalena [AUTHORS: office, procedure, or restricted Zenodo record].

Reproducing the main results from the controlled-access tier: see `code/README.md`, steps 4–8. Reproducing only the tables and figures from the aggregated outputs: `code/fill_rev_v5.py` reads `data_public/tablas_revision_asof/` and `tablas_revision_retro/`.