Why K-12 AI Literacy Needs to Become AI Competency

By /

I’ve read a lot of AI literacy frameworks over the past two years, and most of them blur together. They list what kids should know about AI, they organize it neatly, and they stop there. Chiu et al. (2024) do something different, and it’s the reason this paper held my attention. They argue that knowing about AI was never the real goal. The goal is using it well, and the gap between the two is where most of our teaching falls short.

Their paper draws a clean line between AI literacy and AI competency. Literacy is the knowledge you can point to. Competency is how well you actually put that knowledge to work, which pulls in confidence and attitude, the parts a checklist can’t measure. Chiu and colleagues think we keep choosing the easier target.

The Literacy Trap, and the Case for AI Competency

Here’s the uncomfortable claim at the core of the paper. Chiu et al. (2024) write that “we frequently settle with literacy instead of striving for competency since it requires less effort” (p. 4). I think they’ve named something real. Literacy is easy to write into a curriculum and easy to test. Competency asks us to build confidence, judgment, and the habit of using AI thoughtfully, and none of that fits into a multiple-choice question. For a technology this disruptive, settling for the measurable half is a quiet failure.

AI competency framework

The Self-Reflective Mindset Is the Real Contribution

The five components of their framework are technology, impact, ethics, collaboration, and self-reflection. The first four show up in other frameworks. The fifth is the distinctive one. Chiu et al. argue that a child who keeps checking what they know about AI, and notices what they still need to learn, is the child built for a field that never stops moving. They put it directly: “to succeed in the AI age, students must be able to continuously evaluate their own understanding of AI and stay up-to-date on its advancements” (p. 4). They tie this to confidence too, since a student who feels capable is the one who keeps reading and watching on their own.

This is the same metacognitive muscle that Sidra and Mason (2026) place at the center of their work on collaborative AI literacy and metacognition. A student who reflects on their own thinking as they work with AI learns to use the tool without surrendering to it. Chiu et al. bring that idea down to the middle-school level, which is where the habit has to start.

What saves this framework from staying abstract is who built it. Thirty experienced AI teachers from fifteen Hong Kong middle schools co-designed it across four cycles, voting on every change with a 75 percent agreement threshold. You can feel the classroom in the result.

ai competency framework

A few of their fixes are small and smart. The word “perception” confused students, so the teachers replaced it with human-sensor language, see, hear, speak, read, and write, which kids could actually grasp. Public AI ethics codes often run to twenty principles written for lawyers, so the group cut them to five a twelve-year-old can use: fairness and bias, trust and transparency, accountability, social benefit, and privacy. And they linked prompt engineering to questioning skills, a sharp move, because teaching a child to prompt an AI well is partly teaching them to ask a good question, something schools already value.

What the Framework Still Owes Us

I’d be doing the paper a disservice if I sold it as finished. Chiu et al. are upfront about the limits. The framework hasn’t been field-tested in real classrooms yet, so its strongest claims are still promises. And it leaves out the question I care about most, which is teacher capacity.

Most teachers never received formal AI training themselves, a gap that the work on pre-service teacher preparation from Bilbao-Erana and Arroyo-Sagasta (2025) shows is far from solved. A framework that asks teachers to build student competency assumes a teacher competency we haven’t funded.

The authors also press the older frameworks they build on, warning that models stopping at content “may not properly provide them with a fundamental understanding and get them ready for a future with AI” (p. 7). That critique lands, and it echoes a worry I’ve raised about the framework reviews I’ve covered before, like the Royal Society rapid review from Hillman and colleagues (2025): the field has plenty of content lists and far fewer ideas about how children actually become capable.

Two years on, in a 2026 where students hand whole tasks to agentic tools, the literacy-versus-competency argument reads sharper than it did at publication. What AI actually is matters less every month. The real game is knowing when to trust it, when to question it, and how to keep learning as it changes. Chiu et al. didn’t solve that. They named it clearly, and they handed teachers a place to start. For a framework, that’s a lot.

References

Leave a Comment

Your email address will not be published. Required fields are marked *

Scroll to Top