AI-Supported Feedback in Higher Education: Evaluating Prompt Design and Pedagogical Potential

Author Information

Irum Naz, University of Doha for Science and Technology, Qatar
Rodney Robertson, University of Doha for Science and Technology, Qatar

Abstract

Generative artificial intelligence (AI) is rapidly transforming feedback practices in higher education, yet questions remain about its pedagogical reliability. This exploratory study examines the potential of AI-supported feedback in formative assessment, with a particular focus on the role of prompt design. Grounded in Sociocultural Theory, Second Language Acquisition, and Cognitive Load Theory, the research investigates how AI feedback can scaffold learning, provide comprehensible input, and reduce cognitive load for first-year undergraduates.

A mixed-methods design was employed, comparing AI-generated feedback with instructor evaluations on a simulated undergraduate report and conducting qualitative analysis of feedback clarity, accuracy, and pedagogical value. Findings indicate that while AI feedback often met baseline expectations, it diverged significantly from human evaluation in higher-order dimensions such as APA citation accuracy, comparative analysis, and tailored recommendations. Iterative prompt refinement improved alignment and specificity but did not fully resolve these discrepancies.

Building on these insights, the study introduces the PASTEL framework (Precision, Alignment, Structure, Tailoring, Engagement, Learning), a conceptual model for designing effective prompts to guide AI feedback. While preliminary, PASTEL provides educators with a structured approach for harnessing AI in ways that are pedagogically sound and student-centered.

By combining empirical findings with a conceptual innovation, this study contributes to ongoing debates about AI’s role in higher education, offering both practical guidance for instructors and a foundation for future research.

Save as a Favourite