Many of us have attended artificial intelligence workshops for lecturers featuring well-designed slides and the best intentions. People leave inspired, nodding along, taking photos of the slides. Two weeks later, their courses look exactly the same.
I have seen this so often that I stopped blaming lecturers and started questioning the model: we keep treating AI literacy as something that people attend rather than something they design.
The pattern repeats across many institutions. A talk here, a paragraph in the policy there, perhaps an AI module is added to week three of the syllabus. All this may be well intentioned, but it sends students a silent message: AI is a separate topic, detached from the real content of the course. This teaches them precisely the opposite of what we intended: that AI is either forbidden or should be treated as a shortcut for submitting work quickly. It’s not a tool for thinking.
I propose a shift: AI literacy is not content to be added; it is a quality of design. It should be woven into tasks, assessment and rubrics such that students cannot do well in the course without exercising judgement about AI. When it is designed in this way, AI literacy stops being a topic and becomes a habit.
Of course, that’s simple to say – but it’s trickier to implement.
Put AI into the task
Instead of teaching prompt writing in the abstract, redesign an assignment you already use so that AI becomes part of the process and part of the critique. For example, ask students to generate a first draft or a solution with AI and then compare it with the criteria of the discipline, document what was inaccurate, incomplete or misaligned with disciplinary criteria, and then improve it. The learning is not in the machine’s answer; it is in the comparison.
In my data-related courses, when students use AI to draft a summary of sources and then verify the references one by one, they discover for themselves something that no classroom warning would have taught as effectively: many of those citations simply do not exist. That productive disappointment is more valuable than repeated warnings from the teacher.
Assess judgement, not the product
Here is the uncomfortable question that we often avoid: if AI can produce the deliverable in 30 seconds, then the deliverable itself cannot be the main thing we assess. Our focus must change. What matters is no longer only the final product, but the reasoning behind it: the decisions, the verification, the revision and the ability to defend the work.
This can be translated into three concrete adjustments to the rubric:
- Reward critical use: Add an explicit criterion that values whether the student questioned, compared and corrected what the tool produced, instead of copying it without review.
- Require honest verification: Assess whether the student checked sources, data and claims. Detecting an AI error should count as evidence of learning, not as an invisible detail.
- Ask for transparency: Make the declaration of how AI was used part of the assignment, not a punishment. What is named openly can be learned ethically.
Turn reflection into a habit
For every task that will be completed with AI support, add a closing section in which students write two or three sentences explaining how the tool helped them and where it failed them. It may seem minor, but this is where metacognition begins. A student who becomes used to asking “Is what it gave me true, and is it mine?” does not need an honour code pasted on the wall; they already carry it with them.
We should also remember the equity dimension. Not everyone can pay for advanced versions of AI tools, not all tools handle data in the same way and bias exists. AI literacy also means teaching students to ask whose power sits behind an easy answer.
What we are really trying to achieve
In the end, the goal was never merely to train students to use AI. They will do that on their own, whether we want them to or not. The goal is to educate students who can think clearly with the machine and, above all, despite it. If they leave the course knowing how to doubt an impeccable-looking answer, we have done our job. If they leave believing everything a screen tells them with confidence, we have left them more defenceless than before.
Three ideas are worth keeping: bring AI into a concrete task instead of treating it as a separate workshop; assess judgement and verification, not a product that anyone can generate; and make transparency a habit, not a threat.
AI literacy is not a module to attach at the end of the syllabus. It is a way of designing. When we design it well, it stops being something students learned about AI and becomes a way of thinking and acting in relation to it.
Lucy S. Méndez Santiago is director of e-learning at Universidad del Caribe (UNICARIBE).
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.
Leave a Reply