Designing Flipped-Interaction Prompts: A Framework for Generative AI as an Intelligent Tutor in Higher Education

Students’ use of generative artificial intelligence (GAI) to avoid engaging in generative processing can undermine the validity of higher education. In contrast, Flipped-Interaction Intelligent Tutoring Systems (FIITSs) may promote active engagement by leading a personalised dialogue. The underutilisation of FIITS may stem from the lack of a framework to guide prompt creation and from a dearth of published FIITS prompt examples. This article presents the Flipped-Interaction Prompt (FIP) Framework, abstracted from two validated prompts. To achieve this validation, 26 preservice science education students at a South African university engaged with either prompt in a free GAI five times over ten weeks. The resulting 114 engagements, each involving at least 10 flipped-interaction dialogue exchanges, were analysed for implementation fidelity and for students’ engagement in generative processing. Findings were triangulated against questionnaire and group interview responses, as well as written reflections. The technical implementation was closely aligned with the prompt instruction, with minor deviations noted for not providing answers outright. Additionally, students demonstrated moderate to high levels of generative processing. Findings support the efficacy of the abstracted FIP Framework in guiding the creation of FIITS prompts. Investigating instantiations for additional subject domains would further strengthen confidence in this framework.