AI Literacy Beyond Tools Helping Learners Think With Not Forfeit Thinking To AI
These resources and notes are from my session at the Bowman Symposium on AI Literacy Day, March 2026.
Raise your hand if you’ve used AI this week.
Now keep it up if you made a decision the AI couldn’t have made for you.
That gap, between using AI and staying in the driver’s seat while you do, is what this talk is about.
Most AI literacy training teaches people to use AI better. That’s tool literacy. This is about something harder: keeping your judgment intact while using powerful tools.
AI as Atlas
Kate Crawford’s Atlas of AI (2021) offers a useful frame for how to see these systems. An atlas presents a particular viewpoint of the world with the imprimatur of science. Scales and ratios, latitudes and longitudes, a sense of form and consistency. Yet an atlas is as much an act of creativity (a subjective, political, and aesthetic intervention) as it is a scientific collection.
AI works the same way. The interface looks neutral. The confidence is high. But every model encodes choices about what to include, what to weight, and whose knowledge counts.
That’s the first thing students need to understand. Not how to prompt. How to see.
Two Practices Worth Naming
When people use AI, they tend toward one of two practices. These aren’t personality types. They’re habits, and the same person does both depending on the task. But naming them helps.
Orchestrating means staying in the driver’s seat. You have a frame before you prompt. You know what you want, roughly what it should sound like, and what matters to you about it. The AI is filling in or testing against your frame, not supplying the frame itself. You stay in dialogue with the output. You notice what’s off. You own the final judgment. You could explain every significant choice in what you kept, cut, or changed.
Outsourcing means handing the wheel over. The prompt is the task. “Write me a lesson on photosynthesis for 8th grade.” The frame, the voice, the judgment about what matters. All of it is delegated at the moment of prompting. The output is accepted, not interrogated. Light editing, maybe a sentence swapped. The AI’s confidence reads as competence.
It’s important to note that outsourcing isn’t cheating. It’s a rational response to low-stakes, high-volume work. Nobody needs to orchestrate a permission slip.
The problem is when it becomes the default regardless of stakes. Because outsourcing doesn’t just produce outputs. It shapes the person doing it. If you never do the meaning-making work, that capacity atrophies. Repeatedly.
The question for educators isn’t “did they use AI?” It’s: which practices are your students rehearsing every day?
What Full AI Literacy Actually Requires
Four Resources Model (Freebody & Luke) (1990) gives us a useful map. A fully literate person doesn’t just use texts. They bring four simultaneous practices:
- Code Breaker — How does this system actually work?
- Meaning Maker — What does this output mean to me, in my context?
- Text User — How do I use this tool purposefully?
- Text Analyst — What is this system doing to my thinking? Who built it and why?
Most classroom AI work develops only one of these. Guess which one.
Text User. The functional, pragmatic practice. How to prompt, how to evaluate outputs, which tool to use for which task.
The two that are almost entirely absent: Meaning Maker and Text Analyst. Those are the practices where your judgment lives.
Who Gets to Make Things?
Janet Emig described “writing as a mode of learning” in 1977. The act of composing is generative, not just expressive. Writing isn’t transcription of thoughts you already had. It’s how you find out what you think.
When students outsource the drafting, they don’t just skip a step. They skip the cognitive experience where the actual learning happens. The question isn’t whether the output is good. The question is: whose thinking produced it?
Prompt engineering is not drafting. Evaluating an AI’s essay is not writing your own. If we act like it is, we will produce very efficient managers of other people’s thinking.
What Are the Costs?
These concerns extend beyond individual students.
Hidden labor. Data labelers. Content moderators. The people who make AI safe and functional, often in the Global South, often exposed to the worst content the internet produces. Their work disappears into the product.
Environmental cost. A single image generation uses roughly as much energy as charging your phone. An LLM query uses significantly more water and electricity than a Google search. At scale, across millions of classrooms, this is not a rounding error.
The data feedback loop. Every prompt is a contribution. Student writing, teacher feedback, and curriculum materials fed into AI tools become training data for the next version of the model. Schools with resources adopt tools first. Student data trains better models. Better models are sold back to the same schools. The gap widens.
The transaction isn’t just student → prompt → output. There’s a third party who never shows up in the interface.
Four Moves You Can Make Tomorrow
None of these require new tools.
1. Redesign the task, not the policy. A well-designed task makes outsourcing useless. If the output requires the student’s specific students, specific classroom, specific moment. AI can’t complete it for them. The question isn’t “did they use AI?” It’s “could AI have done this without them?”
2. Make the dialogue visible. Don’t just collect the product. Collect the conversation. Require students to submit the full AI exchange and an explanation of every significant edit they made and why. What you’re now assessing isn’t the essay. It’s the judgment.
3. Teach the frame problem directly. Show them AI getting something confidently wrong that they know. Ask AI about your school, your town, a local event. Watch it hallucinate or go generic. Then ask: “What does this system not know that you know? Why does that matter?” The particular always escapes the aggregate.
4. Co-construct the policy with students. Students who set the rules show more metacognitive awareness than students handed them. Give them a draft policy that’s deliberately incomplete. Have them find what’s missing, what’s unfair, what won’t work. Negotiate the final version together. The negotiation IS the AI literacy activity.
Closing
Like all technologies before it, artificial intelligence will reflect the values of its creators. So inclusivity matters — from who designs it, to who sits on the company boards, and which ethical perspectives are included.
— Kate Crawford
If AI reflects the values of its creators, and teachers shape the values of the people who will build and live with these systems, then this work is not peripheral to education. It is the work.
Slides
[Embed or link slide deck here]
Stay Connected
Digitally Literate — I write a weekly newsletter on AI, digital literacy, and education. It’s free and it’s where I think out loud about exactly this stuff.
→ Subscribe at digitallyliterate.net
AI Literacy for Educators — I’m building a course grounded in this talk: the theory, the practical moves, and the classroom application. If you want early access, get on the waitlist.
Live at wiobyrne.com/ai-literacy-for-educators.
Share that link tonight in your keynote. Go.
Want to bring this to your school or conference? Reach out at wiobyrne.com.