13-05-2025
AI prompts us to rethink the very foundations of higher education systems
Rebecca Davis raises some important questions in her Daily Maverick article, ' Cheat GPT crisis – SA universities faced with a burgeoning degree of AI-written academic assignments '.
She points out, as others have, that 'students are turning to artificial intelligence (AI) to do their assignments and as detection tools fail, universities are scrambling to rethink their assessment methods'.
She suggests further, quite correctly, that some institutions are in denial about the scale of the problem.
In a follow-up piece, ' Chat GPT poses an existential threat to universities and the integrity of degrees ', Associate Professor Andy Carolin from UJ suggests that the availability of Chat GPT has created the risk of an 'ever-increasing number of students who hold certificates that fraudulently certify their mastery of skills and content knowledge that some of them may have only barely attempted'.
He concludes by suggesting that while 'it may go against the enthusiastic embrace of technological innovation', there may be no other choice than to insist on a return to in-person discussions, tests and examinations.
There can be no doubt that AI raises serious questions about how universities should conduct assessments, but this is really only the tip of the iceberg.
AI 'offers' or 'threatens' (depending on your outlook) to fundamentally transform knowledge, teaching, learning and assessment as we currently know it, as well as the people and the institutions that offer it. Consequently, the discussion and debate should prompt us to rethink the very foundations of our current higher education systems.
It is important to highlight that it is not only large language models (LLMs) which are being applied in the education sphere, but these are being combined with symbolic AI, for example, which is more effective in structured, rules-based areas such as learning pathways, competency alignment, student progression, adaptive testing and so on. This is sometimes referred to as neuro-symbolic AI or LLM and planner architectures.
Some of the areas which are of relevance to this debate – and which of course all require much more depth than is possible here – include the following:
Rethinking what constitutes knowledge and competence;
The development of teaching and learning materials and accompanying assessments;
The delivery of teaching and learning and the conducting of assessments;
The future for teaching and administrative staff;
Qualification and changes in hiring decisions by employers; and
The future of tertiary institutions.
Rethinking knowledge and competence
Historically, universities have placed significant value on knowledge retention, ie testing whether students can recall facts, formulae, or processes. However, with hybrid AI tools now capable of delivering detailed, real-time answers, the educational benefit of this type of memorisation must be questioned.
We have historically managed the transition away from times tables and recalled theorems to calculators which do this work for students and working people, and it is this mind-shift – in this case with AI – that we are being asked to make today.
This is not to suggest that some retained knowledge is not important – it clearly is.
But most tertiary institutions worth their salt will be far more focused on building the competence of their students to think critically, to be creative, to collaborate and communicate effectively – the well-versed core 21st Century skills. These are already widely acknowledged as more important metrics than recalling specific data points.
It surely cannot be correct to suggest that the use of AI stands in opposition to these core competencies? AI applications certainly raise the bar in terms of what can be achieved, but those who make extensive use of AI will know that the refining and synthesising of AI outputs through intelligent and creative prompting can be a very challenging process requiring high levels of creativity and critical thinking.
This suggests, therefore, that if private and public universities wish to have students capable of using AI effectively when they enter the new world of work, they will need to assess students' 21st Century skills in both unaided as well as aided formats.
'Unaided' assessments which also seek to test some measure of knowledge retention may well, as is suggested by Carlin, require in-person formats, though arguably this could include online face-to-face formats. 'Aided' assessments would involve required AI use, disclosure by the student and a close examination of the discovery and prompting processes to reach the final learning outcome.
The development of teaching materials and accompanying assessments
A major part of the life of a lecturer has traditionally been the development of lectures and the assessments that go alongside these materials.
Hybrid AI has, however, already demonstrated its ability to generate teaching materials and accompanying assessments at various levels of complexity. This form of AI can create course outlines, lesson plans and reading materials by drawing on datasets that cover an extensive range of academic subjects. These models can synthesise concepts into coherent, digestible materials suitable for different learning outcomes.
Hybrid AI can also automate the creation of formative and summative assessments. Through natural language processing, it can craft questions ranging from multiple-choice quizzes to more open-ended prompts designed to test higher-order thinking skills.
It would be naïve to assume that AI is not already being used by some lecturers to do precisely this – to create and update lecture notes and to develop new assessments which reflect changing global events.
Are we to regard this as some form of cheating? It would seem churlish to deny lecturers the opportunity to be more efficient and effective in their work by using AI applications, judiciously, of course, when the rest of the world is doing precisely this.
Teaching, learning and conducting assessments
Taking this a step further in the life of a lecturer, it should be recognised that hybrid AI can increasingly assume the role of a digital lecturer and learning facilitator by delivering material through interactive, multimedia presentations and real-time discussion prompts.
Intelligent tutoring systems, for instance, can parse student queries and tailor responses with immediate and context-relevant explanations, mirroring the support offered by a human instructor.
Additionally, hybrid AI can track each student's engagement and comprehension through analytics and offer timely, data-driven guidance to help them grasp difficult concepts. This kind of on-demand, personalised assistance ensures that students receive support at their own pace.
More sophisticated AI-based tools can even adapt materials in real time, based on learner feedback, to support the level at which the learner is most comfortable.
When it comes to assessments and marking, hybrid AI can automate much of the process, from generating relevant test questions to scoring responses. Beyond merely checking correct or incorrect answers, modern AI-driven systems can interpret open-ended responses and evaluate them for conceptual clarity and depth of understanding.
Implications for teaching and administrative staff
The capacity of hybrid AI in terms of teaching, learning and assessment clearly suggests that the roles of many lecturers will change over time.
Initially, they are likely to be freed from many menial and repetitive administrative tasks associated with student administration. Then, increasingly with information being so readily available in such great quantities instead of being focused on 'the lecture', it is likely that lecturers will gravitate more towards student support and in-person assessment.
However, as support and self-learning become increasingly personalised through hybrid AI applications, the requirement for lecturer support in even these areas is likely to diminish.
Likewise, with administrative staff, hybrid AI is well suited to take on many tasks associated with admissions, quality assessments, accreditations and qualifications' administration. This is already a reality.
So over time, maybe quite a short time, significantly fewer lecturer and administrative posts are likely to be required.
For current and future students – we will all be lifelong learners – the upside will probably be significantly cheaper education as tertiary institutions that offer online options, having recovered their initial investments, will be able to offer qualifications at significantly lower fees than is now the case.
Again, this trend is already very discernible.
Qualifications and employer hiring decisions
It has already been suggested that it is likely that the assessment of knowledge retention will decline in importance, so qualifications will increasingly need to reflect less about what learners know and more about how learners can apply their own knowledge and the knowledge that they are able to generate through AI applications.
Furthermore, as the pace of change speeds up (largely through technology) and competencies have to be rapidly updated, qualifications are likely to become more modular and flexible.
Instead of granting a single, comprehensive degree, institutions may increasingly move towards stackable micro-credentials or digital badges tied to distinct competencies, allowing learners to continuously update and demonstrate their abilities.
Additionally, non-traditional providers of education — such as industry organisations or even AI-based platforms — could gain more legitimacy in awarding recognised qualifications.
Students of the future may therefore earn credits across multiple providers and compile them into a unified record or 'portfolio', verified and authenticated by blockchain or other secure systems. This decentralised approach would diversify educational pathways and enable continuous upskilling, moving beyond the one-time awarding of a conventional degree.
As qualifications begin to change, employer hiring decisions are likely to change as well, and these two processes will each drive the other synergistically.
Employers are already increasingly focusing on demonstrable skills and practical outcomes. Their recruiters often weigh assessments of real-time problem-solving or proof of project-based achievements more heavily than a traditional diploma or degree.
This will be supported by hybrid AI systems and applications which enable recruiters to verify skill levels through performance data, such as coding challenges, simulations or portfolio reviews.
Recruiters themselves are likely to dwindle in numbers as the formal interview is replaced by hybrid AI-driven platforms which take on tasks like candidate matching, pre-screening, and initial interviews.
The end of universities?
Does this mean an end to bricks and mortar and online universities and colleges as we know them today?
Well, not tomorrow.
It is likely that some tertiary institutions will be able to continue offering their teaching and learning in a more 'traditional' format, which at least eschews some of the pressures from the AI revolution. But these are likely to be already well-branded, boutique institutions which cater in the main part for wealthier students who can afford premium fees.
But for the vast majority of learners in South Africa and the world more broadly, the cost of education is paramount and hybrid AI is already making education increasingly affordable.
The day is probably not long off when we see private and public teaching and learning institutions involving automation at virtually every stage of the student lifecycle.
Admission decisions will be determined algorithmically, factoring in everything from academic records to psychometric assessments derived from online interactions. Class content — lectures, readings, tests — will be generated and updated in real time by hybrid AI, with intelligent tutoring systems offering one-on-one support tailored to individual student needs.
Assessments will be administered and graded by hybrid AI tools capable of interpreting both structured and unstructured responses, providing instant feedback and analytics on a student's strengths and weaknesses.
Administrative tasks — from financial aid calculations to scheduling — will also run autonomously, with minimal human oversight required for exception handling or strategic decisions.
In this scenario, students will interact with a near-seamless digital ecosystem that delivers personalised learning and credentials requiring only minor human intervention to handle aspects of policy compliance, complex student welfare matters or moral and ethical queries.
Good or bad development?
From a pedagogical or andragogical perspective, this is probably a question to which we don't really know the answer, and one which we should probably approach with as much data as possible, rather than through dogma and belief.
