Fair but Impersonal: Student Responses to AI Evaluation in Interpreter and Subtitler Training

Dataset: Student perception survey on AI-generated assessment of reflective writing in interpreter and subtitler training (Spring 2025) This deposit accompanies the article Fair but Impersonal: Student Responses to AI Evaluation in Interpreter and Subtitler Training (under review, The Interpreter and Translator Trainer). It contains the anonymised student-perception survey (N = 43) and the supporting methodological documentation required for independent inspection and reuse. Study context. Undergraduate students enrolled in two courses offered within the Minor in Translation programme at a public university in the Gulf during the Spring 2025 semester — Introduction to Subtitling (TRAN 314, two sections) and Introduction to Interpreting (TRAN 403) — submitted a reflective writing task, which was assessed by GPT-4.1 against a standardised rubric. Students then completed a voluntary perception survey on the AI-generated assessment and feedback they received. What this deposit contains. 01_Instruments — the survey instrument, the informed consent form, the reflective-task guidelines, the grading rubric, and the invitation email, as administered. 02_Survey_data — the anonymised survey responses (CSV, N = 43), a codebook documenting every variable, and a generalised redaction log covering the two categories of substitution applied to open-ended fields (one named third party and one physical condition). 03_AI_grading_protocol — a short protocol describing the model version, interface, grading schedule, inputs, and the prompt as administered for each course. 04_Documentation — a README / Data Statement, and an Anonymisation Plan describing the procedures used to protect participant confidentiality. What this deposit does not contain. The reflective report corpus and the AI-generated feedback files are withheld to avoid residual identifiability of a small student cohort. These materials may be requested from the author for scholarly use, subject to ethics review. Ethics. The study protocol was reviewed and granted exempt status by the institutional IRB on 1 May 2025. All participants provided written informed consent. Participation was voluntary and had no effect on course grades. The survey timestamp was truncated to date only, and survey participant codes (P1–P43) were randomly assigned at the point of anonymisation and do not link to any other file in this deposit or outside it. Reproducibility note. The grading procedure is documented for transparency, not as a guarantee of reproducibility: outputs from commercial LLMs cannot be expected to replicate exactly across model versions or sessions. Related work. https://doi.org/10.5281/zenodo.19634109