Police a losing battle or teach students to think clearly? The choice is ours

Walk into almost any university classroom in 2026 and generative artificial intelligence is already there, whether or not the module handbook acknowledges it. Students use it to plan essays, summarise readings, debug code, translate sources and rehearse for presentations. The common institutional response has been to reach for detection software and prohibition, but banning these tools rarely changes how students learn. Detection tools are unreliable, and the students least able to contest a false accusation tend to bear the brunt. Both responses treat AI as a problem of academic integrity rather than one of teaching.

The more useful question is whether students can evaluate what these tools give them. In professional work, the skill is no longer in producing a first draft, but judging one: knowing where a model is confidently wrong and taking responsibility for the decision that follows. As a recent Campus piece on business students put it, the bottleneck in AI-supported work is not generation but evaluation. That is a teachable skill, and it belongs in every discipline.

• Spotlight guide: Boosting AI literacy across your institution
• Reaching for the pizza? How to help students make informed choices with GenAI
• Apply the principles of critical pedagogy to GenAI

We call that skill critical AI literacy: the judgement to know when a tool can be trusted and when it can’t, and how to tell the difference. Recently, we asked a room of colleagues which capability would matter most for students over the next three to five years – and this was their near-unanimous answer. Building it is a matter of curriculum design rather than policy. Here are five practical ways to start.

Five ways to teach critical AI literacy

1. Make student thinking visible

If an assessment can be completed in seconds by a chatbot, the problem is usually the assessment, not the chatbot. Coursework that asks students to show their reasoning, annotate their drafts, apply ideas to a local or live problem, or defend their work in a short oral discussion rewards understanding that AI cannot quietly supply. This is harder to design than a detection dashboard, and it is the right place to put the effort: GenAI has not so much broken assessment as exposed what it was already failing to measure.

2. Make students argue with the machine

AI models’ output is fluent, confident and often wrong, so the most direct way to build judgement is to have students take it on. Reihaneh Bidar describes asking business students to generate the strongest AI counterargument to a position they had already defended, then evaluate in writing where it overreaches and where it fails to engage the evidence. The task does not ask whether the AI was right; it asks whether the student can tell. The same move works in any field: hand students an AI-generated answer, artefact or piece of code and make their job to find where it breaks.

3. Model your own use openly

If staff use AI to draft feedback, build a marking rubric or generate worked examples, we should say so and show how. Students learn responsible use far more readily from seeing it practised openly, with its limits flagged, than from a policy that treats every use as misconduct. Transparency from the front of the room sets the norm we want from them.

4. Keep the rules simple

Students cannot follow guidance they cannot parse. Policies work best when they are clear, proportionate and workable: a short, plain-language statement on when AI use is expected, permitted or inappropriate in a given task, written into the assessment brief rather than buried in a regulations document. No one trusts vague rules and they’re only ever enforced selectively. 

5. Protect the relationship, not the perimeter

When supervision and tutorials tilt towards detection, the developmental relationship at the heart of teaching curdles into surveillance. Students stop bringing us the half-formed work where learning actually happens. 

Guard that relationship. A supervisory meeting should be the place a student can think aloud about how they used a tool and what they were unsure of, not a checkpoint where they expect to be caught.

Make responsible use part of the curriculum

Teaching students to use these tools well is also teaching them to use them responsibly. The principles that govern trustworthy AI elsewhere apply just as well in the classroom: transparency about how a tool works, fairness in who it serves, respect for privacy and data, accountability for the final output and meaningful human oversight at every step. Woven into ordinary teaching, these stop being abstract ethics and become questions any graduate will be asked. Can you explain how you reached this? Whose data did you use? Who is accountable if it is wrong?

Alongside critical AI literacy, students increasingly need data literacy, ethical reasoning, the confidence to work across disciplines and the ability to engage the people their work affects. None of these replace the deep subject expertise a degree exists to build. They complement it, and they are precisely the capabilities that automation cannot supply on its own.

Be upfront that this is asking more of staff. The same colleagues who worried about a loss of criticality in their students also recognised that the moment “creates a need to be more critical, not less”. They are right. AI amplifies whatever judgement a learner brings to it, so our job is to make sure students bring some.

Disruptive technology in learning and teaching is not a challenge waiting in the future. It is here, and our students are already navigating it, mostly without a map. We can spend the next few years policing a losing battle, or we can teach students to think clearly with these tools and to recognise their limits. The institutions that get this right will not be the ones with the strictest rules but the ones whose graduates can look at a confident machine and know when it is wrong.

Naeem Ramzan is professor of AI and computer engineering and director of the Artificial Intelligence, Viral Communication and Network Institute, and Shahid Awan is lecturer in computer science, both at the University of the West of Scotland. 

If you would like advice and insight from academics and university staff delivered direct to your inbox each week, sign up for the Campus newsletter.