AI-use levels, explained
This kit uses the AI Assessment Scale (AIAS) 2.1 (Perkins, Furze, Roe & MacVaugh) — a five-level scale that replaces vague “may I use AI?” rules with an explicit level for each part of a task. The level names below are the current AIAS 2.1 names, verified against the scale’s official site on 2026-10-25.
The five levels
Each level appears everywhere in this kit as a badge: icon + text (never an icon alone, so it works for screen readers and in plain text).
| Level | What students may do | What students must still do |
|---|---|---|
| 1 · No AI | Nothing generative: no AI drafting, translating, checking or brainstorming. (Spellcheck and a calculator are usually fine — say so.) | All the thinking, writing and problem-solving yourself. |
| 2 · AI Planning | Use AI to plan, outline, generate questions or suggest search terms — before the real work starts. | Do the work yourself; note in the disclosure log how AI shaped the plan. |
| 3 · AI Collaboration | Use AI during the work for feedback, critique, or language help — and critically evaluate what it gives you. | Make the decisions, verify claims, and disclose specifically what you used and what you changed. |
| 4 · Full AI | Use AI freely, including generating drafts and solutions. | Document the interaction, evaluate the output, and be ready to explain and defend the result. |
| 5 · AI Exploration | Co-create with AI as the point of the task: experiment, iterate, push the tools. | Reflect on the process and its limits; the exploration itself is what's assessed. |
Badge styling is provided by the kit’s stylesheet; in the raw markdown you will see the levels as plain text (“AI level 3 · AI Collaboration”) so nothing depends on colour or icons alone.
Setting a level per stage, not per course
The level belongs to the part of the task, because different stages serve different learning goals. A typical pattern in this kit:
| Stage | Typical level | Why |
|---|---|---|
| Choosing a topic, brainstorming | AI Planning (2) | Ideas are cheap; commitment to one is the student’s decision |
| Finding and verifying sources | No AI (1) | Verifying that sources exist is a core academic capability (and AI fabricates citations) |
| Producing the artefact | No AI (1) or AI Collaboration (3) | Depends on whether the course outcome includes AI-assisted production |
| Getting feedback on a draft | AI Collaboration (3) | Feedback-from-AI mirrors real professional practice, with critique required |
| Defending choices (oral/written check) | No AI (1) | This is the secure lane — the point is the student’s own reasoning |
| Reflecting on AI use | AI Planning (2)–AI Exploration (5) | Honest disclosure is assessed, not penalised |
What students must disclose, and how
Disclosure is required at levels 2–5 and takes one short paragraph at the end of the submission (template: assignment-ai-statement):
- Which level each part of the task used.
- Which tool(s), by name, for what.
- What you changed — where you rejected, rewrote or verified AI output.
- One thing you learned or would do differently.
Example disclosure statement (also shown in every student brief):
AI use statement. For this essay I used ChatGPT (free tier) at AI Planning level: I asked it for possible counter-arguments to my thesis and listed two in my plan. For the draft I worked without AI (No AI). After the draft I used the university’s AI feedback tool at AI Collaboration level to check paragraph clarity; I rewrote two paragraphs in my own words where it suggested changes I disagreed with. I verified every source by opening and reading it — AI suggested none of them.
Why the rules exist (the version we tell students)
The rules are about learning, not policing: the level for each stage is set so that you do the parts of the work that build your capability, and AI does the parts that don’t. AI detectors are not used in this course — they misfire, including against multilingual writers — so the honesty of the disclosure log is what matters. This mirrors the position in the kit principles (principles 2 and 6).
Choosing levels with the decision tool
If you are redesigning your own assessment, the decision tool recommends approaches and suggests starting levels per stage. The blank redesign canvas has a column for the level of every stage.