All posts

Your daughter submits a book report. You glance at it and notice the writing sounds nothing like her. When you ask about the main character's motivation, she looks at you blankly. "ChatGPT wrote that part." She doesn't know whether it's right. She doesn't think it's her job to know. She used AI, and AI gave her something, and she handed it in.

This is the core problem with how most children are approaching AI right now: not that they use it, but that they've mentally reversed the authority relationship. The AI is running the project. They're just delivering its output.

The boss/employee mental model is the single most practical reframe a parent can offer. It shifts the relationship entirely, and it's simple enough to explain in five minutes.

The idea is this: when your child sits down to use AI, they are the employer and the AI is the new hire. A good employer doesn't just accept whatever their employee hands them. They give clear direction, they review the work, they catch errors, and they take responsibility for the final result. A good employer also knows when to assign a task to that particular employee and when to handle it themselves.

Most children (and most adults, honestly) are behaving like the employee's assistant. They receive the AI's output, tidy it up slightly, and pass it on. The mental labor has been outsourced. The accountability has gone with it.

Why this reframe works

The boss/employee model works because it gives children a clear, active role with real responsibility attached to it. It answers the question they're almost never asked: what exactly is my job here?

Their job is not to prompt. Their job is not to receive. Their job is to evaluate, direct, and decide.

Researcher Chahna Gonsalves at King's College London identified two distinct types of critical thinking that happen around AI. The first is critical thinking for the assignment -- using AI as a tool to help complete a task. The second is critical thinking toward the AI -- questioning the output itself, checking its accuracy, asking what assumptions the AI made, and deciding what to actually use.

Most students only do the first. A student practicing the boss/employee model is forced to do both. Before prompting, they have to think about what the task actually requires. While reviewing output, they're asking whether the AI did the job correctly. After, they're deciding whether to accept, reject, or revise what they got.

That three-part process (planning, monitoring, evaluating) is what turns a passive AI user into an active AI thinker. It also maps directly to the skills employers are already beginning to screen for.

What it looks like in practice

A seven-year-old who hears "you're the boss of AI" mostly needs to understand that AI can make things up, and their job is to check. That's the whole version for that age. They can do it with simple fact-checking: ask AI a question they already know the answer to and see if it gets it right.

A ten-year-old can start giving explicit direction before prompting. Instead of "write me a summary of the American Revolution," they practice: "what do I actually need here, what context matters, what would make a bad answer?" Then they prompt with that clarity. Afterward, they read the output with a simple question: does this match what I know, and how would I check what I don't know?

A thirteen or fourteen-year-old can learn to audit for specific failure modes -- AI's tendency to present plausible-sounding misinformation with the same confidence as verified fact, to fill in gaps with assumptions, to reflect the dominant perspective on contested topics rather than flagging complexity.

The cognitive demand scales with age. The mental model stays the same.

The accountability piece

There's something else the boss/employee frame accomplishes that's easy to overlook: it puts responsibility back where it belongs.

When a student submits work and it contains an error -- a hallucinated fact, a misattributed quote, a misrepresented study -- the question of accountability becomes complicated if they genuinely believed they were just the messenger. "AI said so" is the new "I copied it from somewhere." It's not a defense, but it does reflect a real confusion about who owned the work.

The boss mental model removes that confusion. The boss is responsible for what the company produces, even if an employee made the mistake. A student who approaches AI as an employer knows from the start that the final product is theirs. That means the verification is theirs. The judgment calls are theirs. The errors, if they slip through, are theirs to catch and correct.

That might sound like a burden. In practice, it tends to produce more careful work -- and children who can actually explain what their report says.

Starting the conversation

The easiest entry point is a question: "If you were hiring someone to help you with this project, what would you actually need them to do?" Walk through what good work looks like, what you'd check before approving it, and what would make you send it back.

Then make the connection explicit: that's exactly how this works with AI. You're hiring it for a specific job. Your job is to be a good boss.

One question worth adding to any AI session: "Would I accept this from a person, or would I ask them to revise it?" That single check shifts the posture from recipient to evaluator -- and it's a habit that builds quickly once it starts.

Sources cited in this post:

• Gonsalves, C. (2024). "Generative AI's Impact on Critical Thinking: Revisiting Bloom's Taxonomy." Journal of Marketing Education. Sage Publications. https://doi.org/10.1177/02734753241305980