The Best AI Policy is Better Teaching

A student submits an eloquent essay. The argument flows, the language is polished, the conclusion convinces. Yet one question remains: what has the student learnt? That question existed long before generative AI. AI simply makes it harder to avoid, and it is why I believe the best AI policy is better teaching.

Universities everywhere are reviewing plagiarism rules and drafting guidelines on responsible AI use. This is necessary to protect academic integrity. But rules on how students may use AI cannot, on their own, answer a more fundamental question: what should students learn, and how should we teach them?

My background is in learning design, so I begin with the intended learning, then consider the authentic experiences, support and assessment students need to achieve it. Technology is a means to that end. A responsible AI policy should help academics make sound teaching decisions and help students understand their responsibilities.

An AI Policy Should Make Thinking Visible

It is tempting to respond with tighter restrictions. But an AI policy centred on catching misuse offers little guidance on designing worthwhile learning. The first question is whether an assessment measures the learning it claims to assess. If an assignment can be completed with AI and little contribution from the student, we need to re-examine it. A convincing submission is not necessarily evidence of understanding.

Consider an assignment asking students to describe a public health issue. An AI system can generate a plausible answer in seconds. A redesigned task could ask students to interpret a local scenario, weigh competing interventions and justify a recommendation using verified evidence, then explain how their decision would change if resources were cut.

The principle is simple: an assessment should reveal how students arrive at, and evaluate, an answer. Yet a more complex question does not automatically make an assignment secure, because AI can produce analysis with the right prompt. Evidence may need to come from several sources: selected drafts, feedback discussions, live demonstrations and brief explanations of decisions.

Written assignments and examinations remain valuable. Foundational knowledge matters, because students cannot critique an answer if they lack the knowledge to recognise its weaknesses. Some learning must still be demonstrated independently, without AI.

Design the Learning Before Assessing It

Assessment reform will have limited effect if everyday teaching rewards passive reproduction. Students need to practise the capabilities we assess: discussing competing interpretations, working through cases, receiving feedback and revising their thinking before a final submission.

AI can support this. A lecturer might use it to generate a deliberately flawed response for students to dissect. The educational value lies in what students do with the output. Do they verify it? Do they notice missing context? Can they explain why one recommendation is stronger than another?

Inclusion matters too. An oral defence must not become an unnecessary barrier for students with disabilities or communication difficulties. And where AI use is expected, universities should provide access, so that paid subscriptions do not become an advantage.

Support Educators and Set Clear Expectations

Academics select worthwhile problems, recognise misconceptions and give feedback suited to the learner. If AI helps prepare materials or feedback, the academic remains responsible for its accuracy. Redesigning learning and assessment requires real support: practical development, examples from their own disciplines, and time to work with learning designers and colleagues.

Better teaching strengthens integrity, but boundaries are still needed. Before starting an assessment, students should know whether AI is prohibited, permitted within limits, or built into the task, in plain language tied to the learning purpose. Disclosure should be proportionate: a short note on the tool used, its purpose and how the output was verified is more useful than a log no assessor can review. Where misconduct is suspected, processes must be fair. An automated AI-detection score should never substitute for an investigation.

Integrity also means accountability. A graduate who accepts an inaccurate recommendation without checking it may make a poor professional decision, even after honestly declaring AI use.

Keep the Purpose of AI Policy in View

AI is prompting universities to revisit what knowledge matters and what evidence demonstrates learning. The best AI policy is better teaching because its principles must become visible in learning experiences, assessment and academic support. Policy provides direction; teaching gives that direction meaning.

Our goal should be graduates who can use AI responsibly and know when independent thinking is essential. Universities cannot regulate their way through this transformation. They must invest in the people who design and deliver learning, as national efforts to build digital skills do. While AI may help produce an answer, it is the university’s responsibility to help students develop the understanding, judgement and accountability to stand behind it.

By Associate Professor Ts Dr Enna Ayub, Director, Centre of Learning and Teaching, University of Cyberjaya .

The post The Best AI Policy is Better Teaching appeared first on RISEMALAYSIA.my .