Teaching "Who's Responsible?" to Builder mode (ages 8–10)

Hook & Warm-Up

Builder mode students are ready for real-world stakes, not just a puzzle to giggle about — but keep the tone matter-of-fact rather than alarming. This is a "let's figure something out together" lesson, not a scary-news lesson.

Say: "A self-driving car hits a pedestrian. An AI denies someone a loan. Who is responsible? The AI? The company? The programmer?" Pause after each question and let it sit for a second before moving to the next. Then say: "These are real situations that have actually happened, and figuring out who's responsible turns out to be one of the hardest questions in all of technology right now. Today we're going to think like justice seekers and dig into why."

Warm-up question: "Raise your hand if you've ever used an app, a game, or a website that made a decision for you — like recommending a video, or deciding you leveled up." Let a few hands go up and a couple of examples get shared. Then ask: "If one of those apps made a BIG mistake — not just a boring video recommendation, but something that really affected someone's life — who do you think should have to answer for that mistake?" Take two or three guesses without confirming or correcting any of them yet — you want them arriving at the activity already turning the question over.

You can sharpen the warm-up with a quick comparison: "If YOU broke a classroom rule, your teacher knows exactly who to talk to — you. But if a computer program breaks a rule, who does anyone talk to? The computer can't sit in the principal's office and explain itself." This small joke usually gets a laugh, and it plants the exact idea the lesson builds on: normal rule-breaking has an obvious person to hold responsible, and AI mistakes often don't.

Main Activity

Walk the class through the lesson's three scenes together, treating each one as a mini case file to investigate rather than a list of facts to memorize. This lesson earns the "Justice Seeker" badge — you can frame the whole activity that way: "we're building a case, scene by scene."

Scene 1 — The Blame Problem. Present the four examples as short "cases": a self-driving car crash (who's at fault — the car maker, the AI developer, the car owner, or the AI itself?); an AI loan denial that can't explain itself (the lesson notes the EU now requires a "right to explanation" for decisions like this); and an AI giving a doctor wrong medical advice (is the doctor responsible for following it, or the company that built the AI?). For each one, ask the class to vote by show of hands on who they think is "most" responsible, then reveal: "Here's the tricky part — in real life, courts and lawmakers are still deciding these exact cases. There often isn't one clean answer — responsibility can be shared across several of these people at once." Show the picture of the justice scale and connect it explicitly: "Being fair means carefully weighing all these pieces before deciding — not picking the fastest, easiest answer."

Scene 2 — Transparency. Introduce "black box" directly: "Many AI systems are 'black boxes' — even the people who built them can't fully explain exactly why the AI made one specific decision instead of another." Contrast this with "Explainable AI" (XAI): "So researchers are working on ways to make AI show its work — pointing out which factors mattered most in a decision, sort of like showing your math instead of just writing down the answer." Bring in the science-fairness comparison from the "glass beaker" item: "In science class, you're taught to show your work so others can check it. AI decisions that affect real people deserve the same kind of checking — that's what an 'AI audit' is: testing a system regularly for fairness and accuracy, kind of like a health inspection for a restaurant."

Scene 3 — Rules & Laws. Explain simply: "Because AI can affect real people's lives, different countries are writing different rules. The European Union passed the first big, detailed law about AI, called the EU AI Act — it bans some risky uses of AI and requires companies to be transparent about others. The United States has taken a lighter approach so far, relying more on existing agencies and voluntary guidelines. And a group called UNESCO got almost every country in the world to agree on a shared set of AI values — though actually following those values is still up to each country." Ask: "Why do you think different countries might disagree about how strict AI rules should be?" and let a short discussion happen — there's no need to resolve it, just to notice that this is genuinely contested territory.

Anticipate a sharp student asking, "so is anyone actually in trouble when these things happen?" A solid answer for this age: "Sometimes companies get fined, sometimes laws change afterward, and sometimes courts are still figuring it out case by case — this is a newer problem than most laws were built for, which is exactly why it's still being worked out."

It's worth also naming, briefly, what an "AI audit" actually looks like in practice, since the base lesson names it but doesn't explain the mechanics: it usually means a team — sometimes inside the company, sometimes an outside group hired specifically to check — tests the AI system with many different kinds of inputs on purpose, looking for patterns like "does this system treat one group of people less fairly than another?" or "does it make more mistakes in certain situations?" It's less like a single inspection and more like repeatedly quizzing the system to find its weak spots before real people are affected by them. Comparing it to a school fire drill can help: you don't wait for a real fire to find out if the evacuation plan works.

Timing note: at a normal classroom pace with the vote-and-reveal structure and the AI-audit explanation, this runs about 15–18 minutes. If time is short, Scene 1 is the one to protect — it carries the lesson's core idea — and Scene 3's country comparison can be trimmed to a single sentence.

Discussion

• "If a self-driving car crashes, should the car company be treated the same way a human driver would be, or differently? Why?"
• "Is it fair for a company to deny someone a loan using an AI that can't explain its reasons? What would you want a company to have to tell you instead?"
• "Why might a company build an AI system that even they can't fully explain? Is that a good enough reason not to have to explain it?"
• "If you were writing one rule that every AI company in the world had to follow, what would it be?"
• "Do you think it's better to have one detailed law like the EU's, or lighter rules that let companies figure it out themselves? What's a downside of each?"
• "If an AI audit found that a system made more mistakes for one group of people than another, what should happen next — should the company have to fix it, explain it, or stop using the system until it's fixed?"
• "Do you think laws about AI should be the same in every country, or is it okay for different countries to have different rules? What could go wrong either way?"

Quiz Walkthrough

Have students answer individually first, then reveal and discuss each one — at this age, the "why" behind each answer matters more than getting it right on the first try.

After the last question, it's worth summarizing the pattern out loud for the class: notice that all four quiz questions point back to the same two ideas — that blame is usually shared rather than single, and that fixing this means making AI more explainable and better regulated. If a student can restate that connection in their own words, they've understood the lesson even if they missed a detail along the way.

Wrap-Up & Extension

Close with: "You investigated three real, tricky cases today and learned that when AI causes harm, blame is almost never simple — it's usually shared between the people who built it, sold it, and used it. That's exactly why the world is still writing new rules to catch up."

Extension activity (15–20 minutes): Split the class into small groups and give each group one short "case" (you can reuse the self-driving car, the loan denial, or the medical AI example, or invent a similar one — like a game's AI banning a player by mistake). Have each group list every person or organization that might share some responsibility, then rank them from "most" to "least" responsible and be ready to defend their ranking to the class. Expect disagreement between groups — that's the point, since real regulators and courts disagree about these exact kinds of cases too.

A shorter alternative: have the class design one rule for a "Rights to Explanation" poster — one sentence describing what a company should have to tell you if an AI made a decision that affected you. Post the best ones around the room as a visual reminder of the lesson's core idea.

If your class already talked about how an AI audit works like a fire drill, you can close with a quick connective question before dismissing: "So if AI audits are like fire drills for computer programs, what's one thing you'd 'test for' if you were in charge of auditing the AI systems at your own school?" A couple of quick answers — "does it treat everyone fairly," "does it explain itself" — is a great, low-pressure way to end, since it puts the lesson's vocabulary directly in their own hands.