Instrumented One-Dimensional Cutting-Stock Educational Exercise: Replication Package

This record provides Version 2.0.1 of the replication package for an instrumented browser-based educational exercise on the one-dimensional cutting-stock problem.

The package includes the current Brazilian Portuguese and English versions of the student instrument, byte-preserved historical versions used during the original study, a deterministic Python and JavaScript verifier, an English class dashboard, the session-log schema and examples, 22 consented session logs without direct personal identifiers, independent human coding, automated tests, reproducibility scripts, and generated results supporting the quantitative claims reported in the associated study.

The classroom task uses a fixed cutting-stock instance with 48 pieces and bar capacity 8.5. First Fit Decreasing (FFD) uses 14 bars for this instance, whereas the exact verifier proves an optimum of 13 bars. The replication workflow explicitly distinguishes result fidelity (matches_ffd_result) from procedural fidelity (followed_ffd_procedure and ffd_process_errors).

Version 2.0.1 additionally includes a reproducible same-problem transferability validation using a second cutting-stock instance. The second instance has 41 pieces, the same bar capacity of 8.5, and total length 110.5. Its volume lower bound is 13 bars, canonical FFD uses 15 bars, and the exact verifier proves an optimum of 14 bars. The package includes a verified 14-bar witness and an independent certificate explaining why 13 bars are impossible. This validation demonstrates reuse of the same FFD, state-verification, integer-scaling, and exact-search implementation on a structurally different instance of the same problem class; it does not claim demonstrated transferability to a different operations-research problem type.

The package also includes an explicit manuscript-claim audit. All entries in the current-manuscript comparison are required to match the reproduced values, including the Figure 2 reference for student 1 (ffd_process_errors = 11 and ffd_process_error_episodes = 4). A separate historical comparison preserves superseded values from the rejected manuscript version solely as archival provenance.

At release, all 57 automated tests passed, including JavaScript/Python equivalence, exact-solver, transferability, privacy, and current-manuscript consistency checks. Reproduced outputs match the bundled expected_results.json. The release audit reports 22 session files, 22 unique technical session identifiers, 22 human-coding rows, and no issues under the implemented direct-identifier checks.

Source Git commit:
5b2f0551ce8bffcbbdf49183f7f02374eb88f223

Archive SHA-256:
c2fd392e9f986461f5ccf4cf755513556b9d7e0a83f84378790724a0c2edcef1

The release archive was rebuilt twice from the same source commit and produced the same SHA-256 value.

Software components are licensed under the MIT License. Research data and non-software research materials are licensed under the Creative Commons Attribution 4.0 International License (CC BY 4.0).