On 25 February, Singapore’s Education Minister Desmond Lee stood up in Parliament and used a phrase that stopped me mid-read.
“Cognitive outsourcing.”
He was answering a question about whether AI use is causing a decline in students’ cognitive skills. He said MOE does not yet have Singapore data on this, but that international research shows inappropriate AI use can lead to over-reliance, harming basic recall and higher-order thinking. He said MOE is actively guarding against it.
Then he used that phrase. Cognitive outsourcing.
I have been writing about exactly this problem for organisations. I call it the Cognition Gap, the risk that people stop evaluating AI outputs and start simply approving them. The skills you stop using are the skills you lose.
Hearing a minister name it the same way, in Parliament, about students, made me pay closer attention to what MOE has been doing since.
This week, a student survey appeared. The Ministry of Education has been asking students, from Primary 4 all the way up to JC Year 3, about how they use AI. Which tools. How often. What for. Whether it has changed how they learn. What concerns them.
I have spent fourteen years working in Singapore government schools and eight years in private enrichment centres. That is long enough to have watched technology evolve through several generations.
I remember the overhead projector. The transparencies. Then came the camera version, where you could put an actual textbook page under it. Then came the screens, the interactive whiteboards, the learning management systems. Each replaced what came before. Each was adopted, embedded, and eventually superseded by something better.
That is how technology normally moves through schools. Steadily, with training days, rollout plans, and someone in charge of the transition.
AI feels different. Not because the technology is obviously better, though it is, but because students are already using it before anyone has decided whether they should.
That is genuinely new.
When the interactive whiteboard arrived, the teacher controlled it. When ChatGPT arrived, the student was already using it on their phone before the teacher knew it existed. The technology did not wait for a rollout plan.
So when MOE puts out a survey trying to understand what is actually happening, I pay attention.
What the Survey Is Really Asking
On the surface, it looks like a usage audit. Which tools? How often? For what?
But read it carefully and it is doing something more interesting. It is trying to map five different kinds of problem at once, even if it does not use those words.
I have a framework I call the Five Cs, five gaps that consistently stop AI from actually working in practice. Let me show you briefly what I see when I read this survey through that lens.
Capability: Do students understand what AI can actually do?
The survey asks which tools students know about, what tasks they attempt, and whether they feel they need to learn more. A student who has only ever used ChatGPT to finish homework and does not know NotebookLM exists is not AI-literate. They are AI-adjacent. That gap matters more than it sounds. Whether students of all ages should be using these tools at all is a fair question, but that’s for another day.
Communication: Are students communicating clearly with AI, and is AI understanding them?
AI is not just a technology interface. It is a language interface. When a student says AI “doesn’t work,” the more likely explanation is that their instruction was unclear, not that the tool is broken. Frustration and vague prompting are the same thing wearing different clothes.
Control: Is anyone actually in charge of how AI is being used?
The survey separates teacher-assigned tools from self-selected ones. That gap between those two columns is a direct measure of unsupervised, unaccountable AI use spreading faster than any school policy can follow. One of the survey options asks students whether they want clear rules about AI use from their school. Students having to ask for rules is itself the answer.
Culture: Are students actually changing how they learn, or performing the appearance of change?
The survey asks whether AI has changed learning, whether students prefer it, whether they feel more independent. These are behavioural questions disguised as attitude questions. Enthusiasm in a survey box does not mean changed behaviour on a Thursday afternoon when the assignment is due.
Cognition: Are students still thinking, or are they outsourcing the thinking?
This is the one I keep coming back to. The survey asks students whether they worry about becoming dependent on AI. The honest problem is that the students who need to worry most are probably the least concerned. A student who feels confident using AI to complete work but cannot evaluate whether the answer is correct has not gained a skill. They have traded one away. Minister Lee called this cognitive outsourcing. I call it the Cognition Gap. We are describing the same thing.
The Questions That Stayed With Me
Of all the concern options in the survey, two stood out.
The first: “I might become dependent on AI.”
The second: “AI might give wrong answers.”
They are connected. Dependency is how you stop forming your own judgements. Wrong answers are what you stop noticing when you do.
The students who tick both are the ones still thinking clearly. They have not surrendered their critical evaluation. They understand that the risk is not just what AI does, it is what it does to you over time.
The students who tick neither concern me more. Not because they are reckless, most are just acting naturally, but because AI produces fluent, confident responses. It does not flag its own errors. It does not look uncertain. If you are not bringing your own judgement to the output, you will not notice when it is wrong. And if you have stopped forming independent judgements altogether, you will not notice that either.
Why This Matters Beyond Schools
Singapore is asking the right questions now, while there is still time to shape the outcome. The survey is not just a data collection exercise. It is the beginning of a real conversation about what it means to learn well and think well in the age of generative AI.
And if a generation of students learns to use AI as an answer machine rather than a thinking tool, those habits do not disappear at graduation.
I have seen enough technology cycles to know that the tool is rarely the problem. It is always about what happens between the human and the tool.
If you want to go deeper on the Five Cs framework, Capability, Communication, Control, Culture, and Cognition, and how they apply to organisations navigating AI adoption, the full piece is over at Kerson.ai.
