AI-supported reflection and human mentoring in internships: a randomised factorial field experiment on workplace judgement — open data and code (public tier)

Open data and code for the article "AI-supported reflection and human mentoring in internships: a randomised factorial field experiment on workplace judgement" (Higher Education, Skills and Work-Based Learning, manuscript HESWBL-08-2026-0707). A 2 x 2 factorial field experiment allocated 324 internship students across six Indian higher-education institutions and 87 host employers to standard structured reflection, AI-supported reflection, human mentoring, or both. This record is the OPEN TIER: person-level (324 rows) and weekly reflection (1,944 rows) files with randomly assigned release identifiers, banded age and placement duration, and without gender, discipline, delivery mode or baseline AI use; a codebook generated from the files; and the scripts that reproduce Table IV exactly (reproduce_primary.py), the sensitivity analyses S7 to S12 (analysis_R1_additional.py) and Figures 1 to 3 (figures.py). The complete de-identified files, with the same release identifiers, are held in a linked Zenodo record with Restricted access and are released to researchers through its request-access procedure; the restriction is a re-identification safeguard in a small six-site sample, not a consent limit. Software: Python 3.11, statsmodels 0.15.0, pandas 3.0.2, NumPy 2.4.4, SciPy 1.17.1. See README.md for layout and instructions.