Towards a pedagogy for deliberate AI-assisted programming in computer science education
Generative AI is reshaping both higher education and professional software development, raising new challenges for how first-year computing students learn to program meaningfully. Recent research suggests novice programmers often engage with AI-assisted tools in ways that privilege speed and surface functionality over understanding, producing shallow prompting, limited code engagement, and reduced visibility of student reasoning. This paper proposes a five-stage pedagogical framework for deliberate AI-assisted programming in first-year computing, derived from the research team’s teaching practice, empirical studies of novice AI-assisted programming, and established software engineering practice. In doing so it reframes AI-assisted programming as a pedagogical design problem rather than a question of tool adoption or restriction, and offers a transferable model that preserves student responsibility for reasoning, validation, and understanding. As the first cycle of a design-based research study, the paper’s contribution is the framework and a mixed-methods evaluation plan through which it will be enacted and assessed in two first-year cohorts, rather than empirical findings.
Authors
- Édouard Amouroux
- Michael Cowling (ORCID: https://orcid.org/0000-0003-1444-1563)
- Thanh Pham
- Tom Hyunh
Institutions
- RMIT University (AU)
Publication Details
- Journal
- ASCILITE Publications
- Published
- 2026-11-17
- DOI
- https://doi.org/10.65106/apubs.2026.3353
- Primary Topic
- Teaching and Learning Programming
- Type
- article
- Field-Weighted Citation Impact
- 0.00