"The AI is broken" is the wrong diagnosis.

Most of the time the AI is doing exactly what it was taught to do — for some people and not others. This pack gives students the words to say that out loud, and the evidence to back it up.

Twenty-four scenario cards, six kinds of bias, one reference page. No login, no devices, no district approval: print, cut, hand out.

What you get (11 pages, US Letter + A4)

• 24 scenario task cards — an essay-grader that punishes one country's spelling, a voice assistant that loses accents, job ads that sort students by gender, a smart camera that under-counts one neighborhood, a feed that narrows a girl's world, a calculator answer nobody questions

• Student reference card naming dataset, stereotype, selection, proxy, automation and feedback-loop bias in one line each

• Detective organizer — seven boxes that end with "how would you test whether your change worked?"

• Cut-and-sort mat — 12 strips, bias or not bias, with a line for written evidence (strip 10 is the trap)

• 3 exit tickets, a teacher quick-start with 20/30/45-minute plans, and an answer key with discussion notes

How teachers run it

Bell ringer (20 min): one card, hands up, argue for two minutes. Investigation (45 min): card, organizer, then a short letter to the company that made the tool — letters on the wall make a strong display. Sub day: print the pack, cut the strips, leave the timer.

Grades 5–8. Fits computer science, character education, media literacy, digital citizenship, and any class that has started talking about AI.

Also from this shop: Phishing Lab, Privacy Lab and Password Lab — the digital-safety simulations, plus a free sampler if you want to try the style first.