By Louis Kruger and Suné Coetzee
Students from Stellenbosch University (SU) lived in a world without search engines, online databases or artificial intelligence (AI) tools when the first edition of Die Matie was published in 1941. 85 years later, a student can generate an essay or computer program in seconds. In a series of interviews with Die Matie, SU lecturers and AI experts explain how this change is forcing universities to ask difficult questions about assessment, the purpose of education, and what it means to be “smart”.
Assessment
The impact of AI on assessment often varies from module to module. According to first-year BCom (Financial Accounting) student Inge Lambrechts, in Digital and Leadership Acumen (DLA) 112, students were required to use ChatGPT to compare products and generate a report explaining their choices.
They also had to submit their conversations with ChatGPT. In many of her other modules, however, AI use was discouraged. The result, she said, is “essentially academic whiplash, forcing you to constantly switch between treating AI as a professional skill and a restricted shortcut”.
A professor at the School for Data Science and Computational Thinking and Pro-Vice-Chancellor for Artificial Intelligence and Quantum Technologies at SU, Dr Fransesco Petruccione, said that generative AI weakens the “inference” from submitted work to demonstrated understanding.
“A polished piece of writing or working piece of code no longer necessarily tells us what the student can do unaided, what they understand or how they reason,” he said. Petruccione emphasised the shift to “process-based” assessment, the diversification into mediums such as oral examinations, and problem-oriented learning – which rewards good questions, rather than answers – as hallmarks of the new mode of assessment. For him, assessment must increasingly ask: “Can the student frame the problem, evaluate the answer, explain the reasoning, identify the limitations and use powerful AI tools responsibly?”
The nature of intelligence
For Dr Hanelie Adendorff, Senior Advisor in the Centre for Teaching and Learning at SU, AI forces academics to re-evaluate “what kinds of thinking assessment should require”.
This may be because some modes of thinking prized in higher education have been shown to be replicable. Dr Debra Shepherd, Associate Professor of Economics at SU, said, “We tend to ascribe ‘intelligence’ to these models because they are very good at recognising patterns and generating coherent responses. And since we have spent the last 50 years treating the human mind like a computer, we see a machine calculating and think, ‘Hey, it’s just like us!’” But in reality, “that’s quite a narrow slice of what intelligence is”.
Dr Tanya de Villiers-Botha, head of the Unit for the Ethics of Technology at SU, said, “While I do think that there is a significant difference between the intelligence (or ‘intelligence’) possessed by AI and humans, I don’t think that this is always readily apparent in assessment situations.”
What kinds of thinking are intrinsically human? For Shepherd, “Intelligence is embodied and socially situated: it involves lived experience, judgement and having a stake in the world. AI doesn’t care about the outcome or understand the norms it’s reproducing.”
The purpose of education
Adendorff said, “Higher edu-cation will need to adapt at more than the level of individual assignments or AI rules.” While these considerations are important, “institutions may also need to ask deeper questions about what [they] reward”, she said.
If AI is being used irresponsibly, it is largely because education has “made the product more important than the learning”, according to Adendorff. “If the goal is only to submit the perfect essay, report, code or calculation, then a tool that can help produce that artifact quickly will be tempting. But if the goal is to become someone who can think scientifically, reason carefully, make sound judgements, work ethically and explain their decisions, then the learning process itself matters.”
For Dr Nuraan Davids, a Professor of Philosophy of Edu-cation and Chairperson of the Department of Education Policy Studies at SU, education is “a relational process that ought to initiate students into thinking about the world in which they find themselves; and to reimagine new ways of being and acting; to exercise human judgement.” In a presentation on the topic, she said, “If AI can ‘think’, the university’s purpose is not to compete with it, but to humanise, contextualise, and take responsibility for thinking.”
Moving forward
The SU Law Faculty’s response to Die Matie on 10 June highlighted “it is arguably still too early to tell” what the impact of AI will be. However, Adendorff said AI is an opportunity, here and now, to rethink education. “The advent of powerful generative AI has not created all the challenges we now see in teaching and assessment, but it has intensified and exposed them,” she said.
Adendorff said that the impact of AI on students is “mixed”. “On the positive side, AI can support access through explanations, feed-back, language assistance, idea generation and study guidance, amongst others. But students may outsource too much of the cognitive work that learning requires. They may become good at producing acceptable-looking answers without developing the underlying understanding or judgement,” she said.
Prof Liezl van Dyk, the Deputy Vice-Chancellor of Academics at SU, announced the launch of SU’s AI in Teaching, Learning and Assessment Resource website on 23 July. According to the website, SU is “approaching this moment through a learning-centred lens – one that upholds academic integrity, strengthens critical judgement, protects learning and prepares students for responsible participation in an AI-rich world”.
Adendorff said, “A universal institutional response is useful at the level of principles.” According to the website, SU aims to make “context-sensitive choices about GenAI in assessment, learning and teaching, with a focus on responsible use, digital literacy and student learning.”
Yet for her, “The more important question remains whether lecturers are making explicit, educationally defensible decisions about AI use in their own contexts, and whether students understand those decisions.”
“The students who will thrive are not necessarily those who use AI the most. They will be those who use it most thoughtfully,” said Petruccione. It may be that, as Davids claims, “the real risk is not AI itself, but complacency in how we respond to it”.
