Abstract

The emergence of Generative Artificial Intelligence (GenAI) has introduced new challenges and opportunities for postgraduate supervision, prompting a reconsideration of how doctoral learning and intellectual ownership are evaluated. Increasingly sophisticated AI tools can now assist with literature reviews, coding, statistical analysis, academic writing, and even viva preparation. While these tools offer substantial opportunities to enhance learning and productivity, they simultaneously raise important questions surrounding cognitive engagement, ethical AI use, and the extent to which students genuinely understand and can defend their research. Central to these concerns is the question: Who is doing the thinking?

The authors previously developed a guiding rubric for PhD supervision in Statistics, designed to support both the development of doctoral candidates towards the graduate attributes expected of PhD graduates and the ongoing development of supervisors in their supervisory practice. Although developed within the context of Statistics in South Africa, the rubric's underlying principles are readily transferable to other disciplines. The emergence of GenAI has brought doctoral education into a new and largely uncharted era, in which even experienced supervisors must navigate unfamiliar challenges and reconsider established approaches to supervision. This raises important questions about how supervisory practices must evolve to ensure that doctoral candidates develop the critical thinking, independent judgement, research competence, and intellectual ownership required of doctoral graduates in an increasingly AI-mediated academic environment.

This presentation explores how the original rubric has evolved to address these challenges and support the responsible and transparent integration of GenAI into PhD supervision. The revised rubric moves beyond AI detection or prohibition, instead placing AI literacy, critical reflection, verification, and scholarly judgement at the centre of postgraduate research training. It introduces developmental maturity levels to guide candidates from uncritical AI use towards reflexive and ethically grounded AI stewardship, while recognising the corresponding need for supervisors to develop their own AI capability and establish shared supervisory norms. The expanded framework offers a practical model for navigating doctoral supervision and academic integrity in this changing landscape, while keeping the central purpose of doctoral education firmly in view.