AI Has Entered the Exam Hall: West African Schools Need Assessment Reform Before Cheating Outruns Policy

For years, examination malpractice in West Africa has taken familiar forms.

A candidate might carry prepared notes into an examination hall. A leaked question could circulate before a paper began. A mobile phone could be used to photograph questions or communicate with someone outside the examination centre. In more organised cases, schools, invigilators or external syndicates might become part of the problem.

Generative artificial intelligence has now changed the equation.

A smartphone no longer needs to carry only a leaked answer. It can run a system that interprets a question, generates an essay, solves a mathematical problem, summarises a passage, produces computer code, or constructs a seemingly convincing explanation within seconds.

That does not mean artificial intelligence is responsible for every case of examination malpractice in West Africa. It is important not to exaggerate the evidence.

But it does mean that education systems now face a fundamental question:

What happens to assessment when the technology available to students can perform many of the tasks traditionally used to determine whether those students have learned?

That question is becoming urgent.
At an education forum in Lagos on October 27, 2026, focusing on examination integrity, ethics, policy, and future readiness, stakeholders called for stronger teacher preparation and clearer policies in response to the misuse of artificial intelligence. Participants warned that AI can be used to circumvent examination rules and argued that technical surveillance alone cannot protect academic integrity.

This should not become another argument about whether technology is good or bad.

Artificial intelligence is already becoming part of education, work and everyday life.

The more important challenge for West African schools is ensuring that assessment evolves quickly enough to distinguish between what technology can produce and what a student knows, understands and can do independently.

Examination malpractice was already a serious problem before generative AI

Artificial intelligence is entering an assessment environment that was already under significant pressure.

In Nigeria, WAEC withheld the results of 167,486 candidates in the 2026 WASSCE over reported cases of examination malpractice. That represented 8.59 percent of candidates who sat the examination. WAEC specifically identified the continued use of mobile phones and organised cheating in some schools as persistent concerns.

In Ghana, the 2026 WASSCE produced its own sobering figures.

Ghana WAEC cancelled the subject results of 7,434 candidates for bringing unauthorised materials into examination halls and cancelled the entire results of 861 candidates for possessing mobile phones. Other results were withheld over suspected irregularities, while cases involving alleged collusion at 169 schools remained subject to investigation.

Those figures should not be presented as evidence of AI use. They are not.

They demonstrate something equally important: West African examination systems are already fighting an integrity problem in which mobile technology, organised assistance and unauthorised access to information play significant roles.

Generative AI increases the capability of the same devices’ schools are already struggling to keep out of examination halls.

That should change how we think about the problem.

Keeping phones out of examination halls remains necessary - but it will not be enough

High-stakes examinations need controlled environments.

Candidates should not have unauthorised phones or connected devices during examinations. Strong invigilation, secure question handling, different question sequences, candidate identification systems and sanctions for organised malpractice all remain necessary.

Nigeria, for example, introduced additional measures for its 2026 examinations, including greater question randomisation, tighter continuous-assessment controls and a unique learner identification system intended to strengthen examination integrity.

Such measures are useful.
But the longer-term challenge goes beyond the examination hall.

Students increasingly encounter artificial intelligence while completing homework, projects, coursework, essays and research tasks.

A teacher may receive an assignment that is grammatically perfect, logically organised and factually plausible.

But who did the thinking?
Was the student using AI to generate ideas, receive feedback or check understanding?

Or did the machine effectively complete the assessment?

Those are very different situations.
A traditional examination room can prohibit phones.

A take-home assignment cannot realistically recreate that environment.

This is why the AI challenge cannot be solved entirely through prohibition.

Assessment itself must evolve.
We must ask what an assessment is trying to measure

Consider a mathematics assignment.
A student is asked to solve an equation and explain the method.

A generative AI system may now produce both the answer and a polished explanation almost instantly.

If the educational objective is simply to obtain the correct solution, technology can increasingly perform the task.

But suppose the teacher asks the student to:

identify an error in an incorrect solution.

compare two different methods.
defend why one method is more efficient.

explain the reasoning orally.
apply the concept to a new local problem.

or revise the solution after receiving feedback.

Now the teacher receives richer evidence of what the student understands.

The same principle applies beyond mathematics.

Instead of asking students only to produce an essay that AI can generate, an English or social studies teacher might ask students to defend an argument orally, critique conflicting sources, explain how their position changed during research or respond to follow-up questions about their reasoning.

A science student might be required to design an investigation, record observations, interpret unexpected results and explain methodological decisions.

A computing student might examine AI-generated code, identify its limitations and improve it.

In each case, the emphasis shifts from producing an answer to demonstrating thinking.

That is where assessment reform must go.

When answers become easier to generate, reasoning becomes more valuable

As a mathematics and STEM educator, I believe this is one of the most important lessons schools should take from the rise of generative AI.

A wrong answer produced by a student can sometimes tell a teacher more than a perfect answer produced by a machine.

Suppose a student incorrectly expands:
3(x + 2)
as:
3x + 2
That mistake immediately reveals something about the student's understanding of the distributive property.

A teacher can work with it.
Where did the reasoning change?
What does three groups of (x + 2) mean?
How could the expression be represented another way?

The error becomes evidence of thinking.
Now imagine that the same student enters the problem into an AI system and copies a perfectly written solution.

The page looks better.
The evidence of learning may be worse.
This is the danger schools must understand.

A correct product is not always proof of a correct learning process.

In an AI-rich environment, assessment must increasingly capture that process.

Students need to show how they arrived at the answer

This does not mean every examination should disappear.

Nor does it mean every student must undergo a lengthy oral defence after every assignment.

It means schools need a richer mix of assessment methods.

Teachers can require students to maintain drafts.

Students can annotate the steps they followed.

Major projects can include brief oral explanations.

Teachers can conduct short conferences in which students explain their work.

Assignments can include reflection questions such as:

What was the most difficult part of this task?

What did you initially misunderstand?

What evidence changed your thinking?

What assistance did you use?
If you used AI, what did it contribute and what did you verify independently?

Classroom assessments can contain unfamiliar problems requiring transfer rather than reproduction.

Practical subjects can include demonstrations.

Group projects can require individual accountability.

Students can critique AI-generated answers instead of merely producing answers that AI can supply.

UNESCO's AI competency framework for students specifically recognises the need for diverse forms of assessment, including reflective writing, oral presentations, project work and other approaches capable of demonstrating students' competencies more directly.

That is an important direction for West African education systems.

We should teach responsible AI use before we punish irresponsible AI use

Another danger is responding to AI only as a cheating tool.

Students may conclude that artificial intelligence is something schools fear, but society expects them to use everywhere else.

That contradiction is unsustainable.
The workplace students are entering will increasingly require them to understand when AI is useful, when it is unreliable, how to verify its outputs, how to protect confidential information and when human judgment must override an automated response.

Schools therefore need clear distinctions between:

AI use that supports learning.
and
AI use that substitutes for learning.

A student asking an AI tool to generate ten practice questions on simultaneous equations may be using technology productively.

A student asking AI to explain why a particular mathematical step is incorrect could be learning.

A student comparing an AI-generated argument with credible sources and identifying inaccuracies may be developing valuable critical literacy.

But a student who submits AI-generated work as independent evidence of personal competence has crossed a different line.

Students need to understand that difference before assessment begins.

UNESCO's student AI framework emphasises responsible engagement, ethical awareness and critical judgement of AI outputs rather than simply teaching learners how to operate the technology.

West African schools should do the same.

Teachers cannot enforce rules they have not been prepared to understand

Any serious assessment reform must begin with teachers.

Teachers need more than a circular telling them that unauthorised AI use is prohibited.

They need professional learning that answers practical questions.

How do I design an assignment in an AI-rich environment?

When can students legitimately use AI?

How should students acknowledge AI assistance?

What should I do when I suspect AI-generated work?

How reliable are AI-detection tools?

How can I redesign an assessment rather than spending hours trying to catch students?

How can AI itself support formative assessment without replacing professional teacher judgment?

UNESCO's AI Competency Framework for Teachers identifies competencies across AI ethics, foundations, pedagogy and professional learning, and notes that many education systems still lack comprehensive national programmes for developing these capabilities among teachers.

A further UNESCO analysis published in September 2026 emphasised that teacher AI competency requires coordinated action across policy, curriculum, regulation, teacher preparation and classroom practice.

That should matter greatly to West Africa.

If students are experimenting with AI faster than teachers are being prepared to manage it, assessment policy will continually lag classroom reality.

Do not build the response around unreliable AI detection

Schools will understandably look for technology capable of identifying AI-generated work.

But detection should not become the foundation of academic-integrity policy.

Generative AI is changing rapidly and attempts to determine with certainty whether a particular passage was generated by a machine remain imperfect.

UNESCO's guidance on generative AI has cautioned that there is limited evidence that automated measures designed to identify AI-produced material are consistently effective. Its longer-term recommendation is more fundamental: redesign assignments so that they better assess human learning rather than relying primarily on tasks that generative systems can easily perform.

That is a far more sustainable strategy.

Instead of asking:
“How can we catch AI?”

schools should also ask:
“How can we design assessment that makes genuine student thinking visible?”

Assessment reform must not widen the digital divide

There is also an equity problem.
Not every West African student has equal access to artificial intelligence.

Some students have personal laptops, smartphones, reliable electricity, fast internet connections and paid AI subscriptions.

Others may depend on a parent's phone, a school computer laboratory or intermittent connectivity.

The current AI debate therefore contains two opposite risks.

One student may gain an unfair assessment advantage because of unrestricted access to powerful technology.

Another student may be educationally disadvantaged because the school never teaches them how to use the technology at all.

The answer cannot be to ignore AI.
Nor should schools assume that every learner has the same access.

Digital literacy programmes, school-based access, teacher-guided AI activities and clear assessment rules must therefore be designed with equity in mind.

Otherwise, a technology with enormous educational potential could deepen existing inequalities.

High-stakes examinations still matter - but they should not carry the entire burden

WASSCE and other national examinations perform important functions.

Universities need credible information.
Employers need qualifications they can trust.

Students deserve certificates that genuinely represent achievement.

Examination integrity is therefore not a minor administrative issue.

As one participant at this week's Lagos forum put it, examinations represent a promise that a certificate means what it claims to mean.

But no single high-stakes examination can capture everything an educated young person should know and be able to do.

Schools should gradually strengthen the quality and credibility of continuous assessment alongside external examinations.

Projects, investigations, practical tasks, presentations, classroom tests and teacher observations can provide additional evidence of learning—provided that those systems themselves are well moderated and protected from manipulation.

The objective should not be to replace one unreliable form of assessment with another.

It should be to create multiple credible forms of evidence.

West Africa needs clear AI assessment rules before every school invents its own

One of the biggest risks now is inconsistency.

One teacher may prohibit all AI use.
Another may encourage it without restrictions.

One school may treat AI-assisted brainstorming as misconduct.

Another may permit complete AI-generated drafts.

Students can hardly be expected to behave responsibly when expectations differ dramatically from classroom to classroom.

Education authorities should therefore establish clear principles.

Students should know:
when AI is permitted.
when it is prohibited.
what use must be disclosed.
how AI-generated information should be verified.

when work must be entirely independent.

and what consequences follow deliberate misrepresentation.

Teachers should also have practical guidance for different subjects and assessment types.

Policies should be understandable enough to use in real classrooms, not written only for technology experts.

Because WAEC examinations cross national boundaries, regional dialogue would also be valuable.

Artificial intelligence does not recognise national borders.

Assessment integrity cannot afford to develop in isolated policy silos.

AI should force us to improve assessment, not abandon technology

The easiest response to a new technological threat is prohibition.

But West African education systems should aim higher.

AI can support teachers.
It can generate practice material.

It can assist with feedback.
It can help students explore ideas.

It can support language learners.

It can provide alternative explanations.

It can help teachers create differentiated instructional resources.

It can also be misused.
Both realities can be true at the same time.

The goal should therefore be neither AI everywhere nor AI nowhere.

The goal should be AI where it strengthens learning, human thinking where human competence must be demonstrated, and clear boundaries between the two.

The certificate must still mean something

Examination malpractice has always threatened education's credibility.

Generative AI does not create that problem.

But it changes its scale and sophistication.

A phone in an examination hall was already capable of carrying unauthorised information.

Now it can also generate information.
That difference matters.
West Africa therefore cannot rely only on stronger invigilation, confiscated phones or punishment after misconduct occurs.

Those measures remain necessary, but they address only part of the problem.

Schools must also redesign assessments.
Teachers must be prepared.
Students must learn AI ethics and digital responsibility.

Education systems must protect equity.
And assessment must increasingly capture reasoning, explanation, application, revision, and independent judgment.

The fundamental question is no longer whether students can obtain an answer.

Artificial intelligence has made answers easier to obtain than ever.

The question education must learn to ask is:

Can the student explain it, defend it, apply it, question it, and demonstrate that the understanding is genuinely their own?

That is the kind of assessment that can survive the age of artificial intelligence.

If West African education systems begin that reform now, AI won't have to destroy examination integrity.

It could become the pressure that finally makes our assessments better.

Shadrack Agyapong
Mathematics & STEM Educator | Education Researcher | PhD Student
U.S.-Based Ghanaian Educator

Sources
The strongest current news hook is the October 27, 2026, Lagos forum on “Exam Integrity, Ethics, Policy and Future Readiness,” where education stakeholders called for teacher training and policy responses to AI misuse and argued that surveillance alone is insufficient. Punch Newspapers

For the regional examination-integrity context, Nigeria's WAEC reported 167,486 withheld 2026 WASSCE results over alleged examination malpractice, while specifically identifying mobile-phone use and organised cheating as continuing problems. News Agency of Nigeria: Ghana's 2026 WASSCE results included 7,434 subject-result cancellations for unauthorized materials and 861 entire-result cancellations for possession of mobile phones; these figures should be described as examination-malpractice figures, not AI-specific cases. WAEC Ghana

UNESCO provides the strongest policy foundation for the solutions proposed in the article: its frameworks call for student AI literacy, critical judgment and ethical use, while its teacher framework addresses AI ethics, pedagogy, assessment and professional learning. UNESCO's generative-AI guidance also argues that institutions should rethink assessment design rather than rely primarily on imperfect AI-detection approaches. UNESCO Docs