Dataset for "Clinical-phase medical students' decisions when artificial intelligence conflicts with clinical reasoning or the medical hierarchy: a vignette-based cross-sectional study" (N = 286)

De-identified dataset from a single-centre, cross-sectional, vignette-based questionnaire study of Year 4 to Year 6 medical students at one Turkish medical faculty (May 2026). Three standardized clinical vignettes each paired a categorical decision (A autonomy, B AI-compliant or defensive, C deferral) with a 5-point Likert item, plus a closing preparedness item, measuring automation bias, deferral, and defensive medicine when an AI recommendation conflicts with clinical reasoning or the medical hierarchy. 288 respondents returned the questionnaire; 286 are analytic after excluding 2 with a missing vignette decision (respondents 284 and 288). The workbook contains the coded dataset, a bilingual (Turkish/English) codebook, and verbatim Turkish responses; no direct identifiers, timestamps, or free-text are included. See the README for full provenance, ethics approval, and reproducibility notes.