ARTIFICIAL INTELLIGENCE IN EDUCATIONAL ASSESSMENT: A HIGH-SCHOOL CASE STUDY

This dataset was created to transparently document the empirical basis of a qualitative study on educational assessment practices observed during a supervised high-school Biology practicum, and to illustrate how contemporary AI methods could address the identified challenges (feedback, personalization, collaborative work, equity) from a formative and ethical perspective. Nature. This is a documentary and derivative package, with no student microdata and no personal identifiers. The units of analysis are textual excerpts from the practicum report (a single, lengthy document), coded via thematic content analysis. The package includes: README.md (overview), protocol.md (methodological protocol), codebook.csv (variable dictionary and allowed values), categories.csv (axis labels), evidence_matrix.csv (matrix of indicators and evidence type), data_aggregated_likert.csv (axis-level descriptive summaries on a 1–5 scale for visualization), and exported figures (PDF/PNG). License: CC-BY-4.0. Scope. It covers six analytical axes: (1) interest in lecture-based lessons; (2) engagement in practical activities; (3) understanding of content; (4) teamwork; (5) oral communication; (6) post-intervention interest. The Likert measures are researcher-generated summaries used solely for descriptive visualization; they do not correspond to student grades nor to AI-based automated scoring. The dataset supports traceability of the article’s interpretations (linking text → codes → tables → figures) and is reusable for teaching qualitative methods, transparency auditing, and reproducing the figure from the .csv files. Limitations: single-case study and no population-level inference.