The emergence of generative artificial intelligence (AI), particularly large language models such as ChatGPT, has rapidly transformed the academic landscape. For educators, the ability of these tools to generate human-like essays, code, and explanations presents a fundamental threat to the integrity of assessment. In computing disciplines, where students are often more technically adept, this threat is especially pronounced. The ease with which written assignments and programming tasks can now be produced raises pressing questions about the authenticity of student submissions, the credibility of academic qualifications, and the fairness of traditional continuous assessment.

This paper presents a case-based redesign of academic assessment within a postgraduate Machine Learning module at Atlantic Technological University (ATU) Donegal, Ireland. The original assessment relied on written literature reviews and code submissions, both now easily replicable using generative AI. Due to the limits of current detection tools and the ethical concerns of acting on unverified suspicion, a new approach was developed. It aims to reduce the success of inauthentic submissions by requiring students to explain their work live or in recorded format, rather than relying on post-submission detection. This method supports the module’s learning outcomes while helping students build oral communication and presentation skills.

The revised strategy draws from two well-established academic formats: the viva voce and the Three Minute Thesis (3MT®) competition. The viva voce, traditionally used in thesis defences and professional examinations, is built on live, spontaneous intellectual exchange. It values a student's ability to think, explain, and respond in real time, skills difficult to fake and nearly impossible for AI to replicate convincingly. Similarly, the 3MT format emphasises concise, engaging communication of complex ideas under strict time and visual constraints. Both models prioritise synthesis, clarity, and oral reasoning over polished but potentially unauthentic written work.

In the revised assessment, students present a topic of their choice (e.g., “Machine Learning and X”) through an 8-9 minute live presentation with a maximum of three slides. Reading from notes or slides is disallowed, and marks are deducted for scripted delivery. For the programming component, students submit five short (3-minute) recorded videos explaining the function and rationale behind their code, demonstrating how classic machine learning algorithms operate. In both elements, the focus is on verbal articulation, personal understanding, and evidence of original thought.

This format encourages meaningful engagement with the material and deters superficial, AI-assisted submissions. It supports students in developing professional communication skills and provides assessors with greater confidence in the authenticity of the work. While generative AI may still play a role in preparation, the requirement to explain and defend content in structured oral formats reinforces individual accountability.

This submission outlines the rationale for the redesign, its implementation, and initial outcomes. It argues that assessment strategies inspired by viva voce and 3MT practices can serve as effective, scalable tools to uphold academic integrity in computing education. It offers a constructive, student-focused response to the growing challenges posed by generative AI in higher education.

Keywords: Continuous Assessment, Generative AI, Integrity, Viva Voce, Presentations.