Activities · Ready to use
Classroom activities
Five self-contained activities for teaching the EU AI Act and AI literacy — with target learners, timing, materials, instructions, discussion questions and a concrete reflection output for each. Use them stand-alone or around the game.
01The five activities
Risk-ladder sorting
Learners: secondary and university · Time: 20–30 minutes · Materials: 12 scenario cards you write on the board or slips (mix mundane and serious uses)
Teacher instructions: In pairs, students sort scenarios into the four risk levels and mark the two they found hardest. Regroup and compare sortings; only then reveal the intended level.
Discussion: Which facts moved a scenario up or down the ladder? Which scenario would change level if you changed one detail?
Reflection output: Each pair writes one sentence: "the detail that changes everything in our hardest scenario was …".
The boundary debate: score vs score
Learners: ages 15+ · Time: 30 minutes · Materials: two half-page scenarios: a bank's credit score; a city's "civic reliability" score
Teacher instructions: Split the class: each side must argue its score is lawful and the other's is not. Swap sides after 10 minutes — the swap is the point.
Discussion: What made the two scores different: the data, the context, the consequences, the exit options?
Reflection output: A shared two-column list: "what makes a score acceptable / what poisons it".
Oversight autopsy
Learners: university and professional training · Time: 25 minutes · Materials: one paragraph describing a fictional office where staff "validate" 400 automated decisions a day
Teacher instructions: Groups list every reason this oversight is fake, then redesign the workflow so a human check would be real (time, information, authority, incentives).
Discussion: What does effective human oversight minimally require? What does it cost?
Reflection output: A 5-line "real oversight" checklist the group would defend to an inspector.
Transparency hunt
Learners: secondary · Time: 20 minutes · Materials: students' own devices or a printed set of screenshots
Teacher instructions: Students find three interactions in their daily tools where AI is plausibly involved and check whether it is disclosed. Map findings against transparency obligations.
Discussion: Where were you told? Where should you have been? Did the disclosure change anything for you?
Reflection output: One paragraph: "the disclosure I would add, and where".
EdTech privacy audit
Learners: secondary and university · Time: 30–40 minutes · Materials: the question list from privacy-conscious learning; one real tool the class uses
Teacher instructions: Apply the audit questions to the chosen tool using only public information; mark each answer as verified, claimed, or unknown. Optionally contrast with this project's privacy page, where answers are verifiable in code.
Discussion: How many answers were "unknown"? Is "unknown" acceptable for a tool used on students?
Reflection output: A short letter to the (fictional) vendor asking the three most important unanswered questions.
02Using the activities with the game
Every activity has a natural companion case in NO AI ACT: the sorting activity previews the whole game; the boundary debate mirrors the "Civic credit" case; the oversight autopsy matches the recruitment and triage cases; the transparency hunt pairs with the chatbot case. Play first for energy, or play after as verification — both orders work.
For full sessions with objectives and assessment, see the lesson plans; for facilitation style, the teacher guide.
03Printable worksheet: classifying the risk of an AI system
An individual or pair worksheet to fill in on paper (this page is print-optimised: use your browser's Print function). It works with any game case or with a real AI system the class chooses.
| Guiding question | Your analysis |
|---|---|
| 1. What does the system do? Describe the function in one sentence, without jargon. | |
| 2. In what context is it used? Who uses it, who is affected, what is at stake. | |
| 3. What data does it process? Personal, biometric, behavioural, children's data? | |
| 4. Proposed risk category (prohibited / high-risk / transparency risk / minimal) and why. | |
| 5. Who is responsible? Provider, deployer or both — justify. | |
| 6. Which proportionate measure would you propose? | |
| 7. What would make your classification contestable? The strongest argument against your own choice. |
Review before filling in: risk categories and the glossary.
04Printable case-discussion rubric
A formative rubric (not a grade) for observing the quality of discussion after a case: useful for the debrief and for peer self-assessment.
| Criterion | Beginning | Developing | Advanced |
|---|---|---|---|
| Use of evidence | Opinions without reference to the case file. | Cites at least one relevant piece of evidence. | Connects several pieces of evidence and separates decisive from marginal ones. |
| Regulatory language | Does not use AI Act categories. | Uses the categories with some imprecision. | Uses categories, roles (provider/deployer) and obligations correctly. |
| Consideration of affected people | Ignores who is affected by the system. | Names the groups involved. | Analyses differentiated effects and the rights at stake. |
| Openness to contestability | Defends their answer as the only possible one. | Acknowledges that counter-arguments exist. | States the strongest counter-argument and answers it on the merits. |
The contestability logic mirrors the game's: see serious games for AI regulation and the methodological framework on research and methodology.