Reliable but supervised: evaluating a generative AI-rubric model for consistent and fair assessment in postgraduate education

This study presents and evaluates a generative AI-supported, rubric-aligned assessment model implemented in a postgraduate business economics course, utilizing a design-based research (DBR) framework. Drawing on ChatGPT, faculty-generated prompts embedded macroeconomic scenarios and behavioural biases—such as overconfidence and anchoring—to simulate real-world decision-making. Students submitted narrative analyses, which were scored using a structured four-part rubric and received AI-assisted feedback refined by instructors. Statistical analyses (ordinary least squares and logistic regression) confirmed the internal consistency, fairness, and pedagogical alignment of the model, with no evidence of bias due to gender or cohort. Faculty reflections indicated growing instructional fluency and trust in the model, while the quasi-complete separation in logistic models paradoxically underscored the precision of the rubric. These findings demonstrate how AI, when embedded in a faculty-led system, can enhance transparency, maintain academic judgment, and support scalable, ethically grounded assessment in higher education.