Method · Transparency · Research
Research and methodology
This page documents how NO AI ACT is designed, which educational objectives it pursues, which simplifications it makes and — just as clearly — what has not yet been demonstrated. It is the starting point for researchers, lecturers and institutions who want to study or reuse the game.
01Educational objective
NO AI ACT trains three competences: (1) conceptual knowledge of the risk-based approach of Regulation (EU) 2024/1689 — risk categories, prohibited practices, transparency duties, provider and deployer roles; (2) the ability to apply it to concrete scenarios — classifying a system, selecting decisive evidence, assigning responsibility, choosing proportionate measures; (3) civic awareness of why AI regulation exists, made tangible through a fictional city where the AI Act never entered into force.
The design hypothesis — to be tested empirically, not proven — is that making decisions with visible consequences and receiving a reasoned debrief trains regulatory reasoning better than passive reading alone.
02Game mechanics
- Inspection case files: 13 cases on AI systems (social scoring, workplace monitoring, deepfakes, healthcare, biometrics, credit, public chatbots, procurement, EdTech, GPAI).
- Decision cycle: examine evidence → classify risk → assign responsibility → corrective measures → report → debrief.
- Visible consequences: city indicators respond to decisions, showing the cost of an unregulated society.
- Explanatory debrief: every decision receives an explanation of why it holds or fails, with a pointer to the simplified rule involved.
- Local teacher mode: discussion pauses and an anonymous classroom debrief, with no data collection.
The full flow is described in how it works; the pedagogical background in serious games and AI regulation.
03Target groups
- Upper secondary school (civic and digital education).
- University: law and technology, computer science, political science, media studies.
- Professional and adult training.
- Self-directed learners, citizens and curious professionals.
04Simplified legal model and source matrix
The game is a didactic simplification: it compresses procedures, timelines and interpretive nuance to make regulatory reasoning playable. It is not legal advice and does not replace the regulation. The themes it covers map to the following areas of Regulation (EU) 2024/1689 — the authoritative source is always EUR-Lex:
| Theme in the game | Area of the regulation | Read more on this site |
|---|---|---|
| Prohibited practices (e.g. social scoring) | Chapter II, Art. 5 | Prohibited practices |
| High-risk systems (healthcare, education, employment, essential services) | Chapter III and Annex III | High-risk AI systems |
| Transparency duties (chatbots, deepfakes, synthetic content) | Art. 50 | Transparency obligations |
| General-purpose AI models (GPAI) | Chapter V | Generative AI & GPAI |
| Roles and responsibility (provider, deployer) | Art. 3 and Chapter III | EU AI Act guide |
The classifications proposed in the cases are traceable case by case in the open-source game data; ambiguous scenarios are treated as contestable, not as a single right answer. A third-party legal review is recommended before formal institutional use.
05Privacy architecture
The public product is designed to be studiable without collecting personal data: no account, no backend, no gameplay telemetry; saves stay in the browser's localStorage. Public pages use only aggregate, cookie-free Cloudflare Web Analytics. Details and how to verify this in privacy by design.
The methodological consequence: any research instrument (questionnaires, pre/post tests) lives outside the game, under the research team's own ethics responsibility. The game does not embed surveys and will not.
06Validation status — what is and is not demonstrated
Verifiable software properties
Behaviours guaranteed by automated tests and open code: no network calls during gameplay, local-only saves, per-version pinned content and scoring, IT/EN bilingualism.
Educational effectiveness
Educational effectiveness has not yet been empirically validated: there are no controlled studies of learning or transfer. Informal classroom use has so far informed design iterations only.
Always distinguish three levels: usability (can people operate it), engagement (do they keep using it willingly) and learning (do the target competences improve). Positive results on the first two must never be presented as proof of the third.
07How to study or reuse the game
- Validation framework: theory of change, constructs, proposed pre/post measures, sampling and an anonymous participant-code protocol are in the research validation framework (repository).
- Versioned stimuli: cases, texts and scoring rules are reproducible by citing the exact version (GitHub tag +
release.config.json). - Reuse: GPL-3.0 code, CC BY-SA 4.0 content — adaptations and translations are permitted with attribution.
- External instruments: tests and questionnaires stay on paper or on platforms chosen by the research team, never inside the game.
08How to cite
Ready-made formats (APA, BibTeX, software citation with version) are on how to cite. The repository ships a CITATION.cff file readable by GitHub and reference managers. A Zenodo DOI is planned with the 2.0 release.
09Contact and collaboration
NO AI ACT is designed and developed by Matteo Angeloni, PhD candidate in Society in change: policies, rights and security at the University of Tuscia. The project is independent and is not an official output of the university.
For study proposals, classroom pilots, error reports or institutional collaboration: open an issue on the GitHub repository. We use no external forms and collect no email addresses on this site.