The AI Fluency Loop

An experimental framework

A student asks a chatbot for sources. A convincing list appears, but one citation leads nowhere. What happens next? That is where the learning becomes interesting.

I want to give that learning a shape: notice, question, make, test, teach. The AI Fluency Loop is an experimental framework for this book, a proposal to try and revise. Its usefulness still needs to be tested with learners.

1. Notice

Begin with something that matters: a question, a frustration, a possibility. What is happening? Who is trying to do what? Start with a student struggling to find a source or a song you have always wanted to write.

2. Question

Examine the situation before choosing a tool. What are you assuming? Where could AI help? Decide what a useful result would look like, what remains your responsibility, and what evidence would change your mind. You may choose to work without AI.

3. Make

Give the idea a form you can examine: an explanation, a brief with sources, a lesson, a song, or a small website. Describe the purpose and constraints to the AI, then choose what to keep and change. Explore with sample data.

4. Test

Compare the result with your purpose. Open the sources. Try the task. Ask an intended user to use the prototype. An AI’s reassurance is not independent evidence. Decide whether to keep, revise, or abandon the attempt. Name what remains unchecked.

5. Teach

Explain what you did and why: where AI contributed, what you checked, and what remains uncertain. Show a colleague, talk with a friend, or rehearse the explanation yourself. A question you cannot answer gives you somewhere to begin again.

Move back and forth as the work requires. Testing may reveal a misunderstood problem; teaching may send you back to the sources. AI literacy belongs throughout the loop. Fluency means putting that understanding and judgment to work and adapting it.

One small library experiment

A librarian notices students arriving with citations they cannot locate. She questions whether they need another source-finding tool or help checking citations. She uses AI to make a one-page guide with labeled practice examples. She tests it: can a willing student locate a real source through the catalog or publisher and explain whether it supports a claim? She revises confusing instructions, then teaches a colleague how the check works and where the guide falls short. Their questions begin another turn.

Use it with this book

The “Try this” prompts invite you to explore, investigate what has changed, or build something. Follow one, then decide what comes next. An explanation is something you can make; checking current evidence can be a test. You need not complete every step in one sitting.

In teaching, compare similar tasks before and after introducing the loop. Look for a clear purpose, a reasoned AI choice, a useful attempt, an independent check, and an explanation of choices and limits. Later, try a different task. These observations can guide revision; they do not establish that the framework caused an improvement.

This loop organizes practice, rather than providing a validated assessment or complete curriculum. For a broader competency model, see Rick Dakan, Joseph Feller, and Anthropic’s 4D framework: delegation, description, discernment, and diligence.¹