How does GenAI-integrated learning tasks affect students’ knowledge retention?

Abstract

Australian universities are increasingly permitting the use of GenAI in assessments, and yet, empirical evidence on its effects on student knowledge retention is lacking. Methods: Guided by Process-Oriented Pedagogy, we designed a student-centred assignment for first-year marketing (MM110) students (n = 183). Students self-reported their Gen-AI use and submitted reflections on the process, documenting iterations, actions, rationales, and supporting evidence. More importantly, informed by Open Systems Theory and Triple-Loop Learning, the marking rubric was developed incorporating “Input–Process–Output” elements with six sub-criteria scaled across three learning levels. Assignment marks were then compared with the marks of the invigilated exam to evaluate knowledge retention. (HREC HE-2025-2429-4050). Findings: GenAI users achieved significantly higher assignment scores than non-users (M = 11.30 vs. 9.86, p= .001). Effects varied by usage pattern: data analysis (M = 11.48 vs. 10.02, p= .001) and minor editing (M = 11.23 vs. 10.63, p = .036) were associated with significant gains, whereas content writing was not. Process marks correlated strongly with other assignment components (r = .50–.80) but only weakly with exam performance (r = .18, p = .016). Overall, the use of different forms of GenAI in assignments did not significantly influence exam outcomes, suggesting that GenAI neither enhanced nor diminished knowledge retention. Implications: GenAI tools were most effective when used to scaffold analytical and editorial processes rather than for generating the content. Students who preserved an authentic voice while employing GenAI for analysis scored better in assignments, fostering critical thinking without undermining knowledge acquisition.