Qualitative single-case study reveals process-focused AI frameworks outperform unstructured chatbot access in university assessments, highlighting the necessity of structured institutional governance.
Generative AI makes student outputs ambiguous indicators of learning. This study examines how institutional governance can protect validity while enabling responsible AI use, aligned with the EU AI Act and the European Commission’s Ethics Guidelines for Trustworthy AI. A qualitative single-case study of a departmental Task Force (October 2024–April 2025) compares two pilots: unscaffolded retrieval-augmented chatbot access (n = 35 of 116 enrolled) and process-focused assessment using the DRIVE framework (n = 70 students; 1,450 prompts). Chatlogs, student surveys (n = 19), instructor evaluations, and governance documents were analysed using a unified validity framework. An efficiency paradox emerged: unscaffolded access produced high satisfaction but no exam-performance difference (d = 0.08), whereas DRIVEinteraction-quality scores correlated with essay scores (r = .54), separating collaborative partnership from passive delegation. Validity protection requires ethical review of pilots, process visibility, validated rubrics, faculty development, and equity-impact analysis. Findings are exploratory.
No takes yet. Share an insight, caveat, or question.
Sadowski et al. (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: