AI didn’t break university assessments — it exposed a dangerous lack of graduate capability

South Africa cannot build problem-solving universities on assessments that reward polished language more than judgement, application and capability.

Universities are panicking about artificial intelligence because it can write essays. They should be more troubled by what this reveals: many essays were never reliable evidence of understanding in the first place.

This matters profoundly in South Africa. A country confronting unemployment, infrastructure decay, water insecurity, energy instability, weak public services and low industrial competitiveness cannot afford universities that certify fluent writing without testing real capability. The issue is not only academic integrity. It is a national capability.

AI did not break assessment. It exposed it.

For decades, higher education has relied heavily on assessment formats that reward the appearance of learning. Students submit essays, reports, literature reviews and examination answers that look correct, sound scholarly and follow the familiar rituals of academic presentation. If the language is polished, the structure coherent, and the referencing acceptable, we often treat the work as evidence of understanding.

That assumption has become dangerously weak.

AI has exposed what I call “the fluency fallacy”: the belief that coherent academic language is the same as understanding. A student may submit a beautifully written essay and still not understand the subject. A report may be grammatically elegant and intellectually empty. A policy analysis may sound sophisticated while evading the difficult work of judgement.

This problem did not begin with ChatGPT or any other AI tool. It has been present for years. AI has simply made it impossible to ignore.

Generative AI can now perform academic fluency at speed and scale. It can produce the tone, structure and vocabulary of university work. It can imitate scholarly confidence. It can generate the surface of learning without proving that a particular student has gone through the discipline of learning.

That is why the AI debate in higher education cannot be reduced to cheating. Academic dishonesty matters. Universities must protect integrity. But if universities respond only with bans, panic and detection software, they will miss the deeper crisis.

The real question is not: How do we stop students from using AI?

The deeper question is: What kinds of assessments still prove that a student can think?

Assessment theatre

Too many assessments have become a form of “assessment theatre”. They create the appearance of rigour through long assignments, formal rubrics, referencing rules and polished submissions, while failing to test whether students can reason, apply, defend, create or solve. AI is dangerous to assessment theatre because it can perform the theatre perfectly.

It can write the essay. It can imitate the report. It can satisfy the rubric. But it cannot, by itself, prove that a student has developed judgement, ethical responsibility, disciplinary understanding or the capacity to act under real-world constraints.

That is the difference universities must now defend.

The goal should not be to make an assessment “AI-proof”. That phrase is too defensive and already outdated. AI is inside classrooms, workplaces, research, professional practice and everyday life. The goal is not to pretend that AI can be excluded. The goal is to make learning visible.

A serious university should not ask only whether a student submitted an answer. It should ask whether the student can explain the answer, defend the method, justify the evidence, critique the assumptions, acknowledge the limitations and apply the knowledge to a new situation.

In other words, universities must move from submission-based assessment to judgement-rich assessment.

Judgement-rich assessment does not abandon writing. Writing remains essential. But writing can no longer stand alone as unquestioned proof of learning. A written submission should become one part of a wider assessment ecology that includes process, defence, application and reflection.

The future assessment question should be simple: what can this task reveal about the student’s own understanding that AI alone cannot convincingly fake?

That question changes everything.

A student who submits an engineering design should be able to explain the design choices, load assumptions, material constraints, safety factors and possible failure scenarios. A student who submits a policy paper should be able to discuss trade-offs, consequences and ethical risks. A student who submits code should be able to explain the logic, debug errors and adapt it to a new problem. A student who submits a business plan should be able to test assumptions against real users, markets and constraints.

This is not anti-AI. It is pro-learning.

AI should be brought into assessment honestly, not smuggled in secretly. Students should be required to declare how they used AI, what outputs they accepted, what they rejected, what they verified and how human judgement shaped the final product. That would move universities away from a culture of concealment towards a culture of accountable use.

The future graduate will not be someone who never uses AI. That is unrealistic. The future graduate must be someone who can use AI critically, ethically and intelligently. Such a graduate must know when AI is useful, when it is misleading, when it fabricates, when it reproduces bias, when it oversimplifies and when human judgement must take over.

The South African context

This is where the South African context becomes unavoidable.

Our country does not need graduates who can merely produce polished assignments. It needs graduates who can solve problems. We need graduates who can address unemployment, water insecurity, energy instability, public health challenges, infrastructure decay, township economic exclusion, weak local manufacturing and failing public services. In such a context, assessment cannot remain trapped in academic ritual. It must test capability.

South Africa cannot build a developmental state, an innovative economy or problem-solving universities on assessment systems that reward fluency without competence.

Assessment is not only a university matter. It is a national capability system. Every weak assessment eventually appears somewhere in society: in a weak bridge, a weak classroom, a weak clinic, a weak municipality, a weak policy, a weak enterprise or a weak public institution. When universities certify students without testing their ability to reason and apply knowledge, society carries the cost.

The country does not only need more graduates. It needs more capable graduates.

This is where universities of technology should lead. Their applied mandate places them in a powerful position to pioneer the post-AI assessment model. They should not imitate traditional universities by overvaluing essays, abstract academic performance and paper-based demonstrations of competence. They should show the future by assessing prototypes, portfolios, workplace learning, simulations, design reviews, community projects, entrepreneurial experiments and applied problem-solving.

The AI era may prove that the university of technology model was not behind the times, but ahead of its time.

For too long, applied education has been treated as less prestigious than abstract academic work. AI should end that illusion. When machines can produce fluent academic prose in seconds, the value of higher education must increasingly be found in judgement, practice, creativity, ethical reasoning and the capacity to solve real problems.

Traditional universities should also change. Philosophy students should defend arguments orally. Economics students should analyse real budget trade-offs. Law students should reason through live social dilemmas. Education students should design learning interventions. Science students should explain experimental uncertainty. Management students should diagnose failing organisations. Public administration students should solve service-delivery problems.

Every discipline must ask: Where is the evidence of independent judgement?

This does not mean every assessment must become complicated or expensive. Large classes, limited staffing and unequal student access are real constraints. But poor assessment is also expensive. It produces graduates who may have certificates but lack confidence, competence and adaptability. It weakens public trust. It damages the credibility of universities. It creates the illusion of quality while employers and society absorb the consequences.

The reform can be practical.

Assessment audits

Every South African university should conduct an assessment audit within the next academic year, asking one simple question: Does this assessment test capability, or merely fluency?

That audit should lead to practical changes. Major written assignments should include a short process note explaining how the work was developed. Final-year modules should include oral defence, presentation or viva-style verification for selected high-stakes tasks. Capstone projects should use real-world problems, local data or applied outputs. Students should be allowed to use AI in clearly defined ways and required to acknowledge that use. AI detectors should never be the sole basis for a misconduct finding. Rubrics should reward judgement, evidence, originality, method and application, not only structure and language.

This is not about making assessment harder for students. It is about making assessment more honest.

The university must also stop confusing difficulty with rigour. A long essay is not necessarily rigorous. A three-hour examination is not necessarily rigorous. A complex rubric is not necessarily rigorous. True rigour lies in the quality of thinking demanded by the task. A rigorous assessment requires students to interpret, decide, justify, adapt and take responsibility for their reasoning.

This demands a different academic culture. Lecturers must be supported to redesign assessment. Departments must be given time and resources to rethink programme outcomes. Quality assurance systems must stop rewarding compliance alone and start asking whether assessment genuinely tests capability. Professional bodies must encourage authentic evidence of competence. Students must be trained in ethical AI use instead of being left to experiment in fear.

A permanent change

Universities should not treat AI as a temporary disruption. It is a permanent change in the knowledge environment. The workplace into which students are graduating will use AI. Research will use AI. Industry will use AI. Public administration will use AI. Entrepreneurship will use AI.

The moral duty of the university is not to pretend otherwise, but to prepare graduates who can work with AI without surrendering judgement to it.

AI should push higher education back to the oldest and strongest forms of learning: dialogue, demonstration, critique, apprenticeship, defence and practice. In this sense, AI may make assessment more human, not less human.

The university of the future will not be the one with the harshest AI policy or the most aggressive detection software. It will be the one with the most intelligent assessment design. It will ask students not only to submit, but to explain; not only to answer, but to defend; not only to write, but to demonstrate; not only to know, but to judge; not only to complete tasks, but to solve problems.

AI did not break assessment. It exposed the fluency fallacy.

Now universities have a choice. They can protect assessment theatre with detection software, or they can rebuild assessment around the one thing machines cannot replace: responsible human judgement. DM