A five-level guide helps teachers define where AI can assist, what students must do themselves, and what evidence to include with each assignment.
AI rules work best when students can apply them to the task in front of them. A broad instruction such as “use AI responsibly” leaves room for conflicting interpretations: one student may ask a chatbot to explain a term, while another asks it to write the response. Both might think they’ve followed the rule.
A clear permission level on each assignment can make those boundaries concrete. It tells students what AI may help with, which parts they must complete themselves, and what to disclose. The key decision is not how much AI a class should use in general. It’s what kind of evidence the assignment needs to show.
Match AI permission to the learning goal
An assignment’s AI rules should follow its assessment purpose. If students are being assessed on their ability to build an argument, asking AI to produce that argument could obscure the skill the teacher wants to evaluate. If the goal is to assess arguments, an AI-generated example may give students useful material to examine and critique.
The AI Assessment Scale, developed by Mike Perkins, Jasper Roe, Leon Furze, and Jason MacVaugh, connects AI permissions with assessment goals. The five classroom levels below offer a practical guide for assignment sheets, with categories and wording designed for that use.
| Level | Permission | What students may do | What remains student work |
|---|---|---|---|
| 1 | Complete without generative AI | Use the resources and supports specified by the teacher | The thinking and production being assessed |
| 2 | Use AI for preparation | Explore a topic, clarify concepts, or brainstorm within stated limits | The plan, response, and final submission |
| 3 | Use AI for feedback | Request comments on a student-created draft or solution | The initial work, revision decisions, and final response |
| 4 | Use AI for specified components | Generate or revise the particular elements named in the instructions | The components and decisions reserved for the student |
| 5 | Use AI throughout the task | Integrate AI across planning, production, and revision | Direction, verification, justification, and responsibility for the submission |
These levels describe permissions, not a ranking of learning quality. Level 5 isn’t automatically better than Level 1. The right choice is the one that lets the teacher assess the intended learning.
Related: AI Disclosure Statement Examples for Students
Set boundaries across five levels
Level 1: Complete the work without generative AI. This level fits tasks where the teacher needs evidence of what students can do without generative assistance, such as writing an initial response to a text, solving a math problem, or explaining an idea from class materials.
Complete this assignment without generative AI. Develop the ideas, reasoning, and response yourself using the resources listed below. Continue to use your approved accessibility supports. Ask me before beginning if you are unsure whether a tool or feature is permitted.
List the resources and supports students may use. That can include dictionaries, spellcheck, translation tools, or calculators. Approved accommodations should be addressed through the appropriate school process.
Level 2: Use AI to prepare. Students may get limited help before they start producing their own work. They might ask AI to explain a concept or suggest questions about a topic, then use assigned readings and verified sources to develop their response.
You may use AI to clarify unfamiliar concepts and brainstorm possible questions before you begin. Create your own outline and write the submission yourself. Do not use AI to generate or rewrite passages for submission. Briefly describe any AI assistance you used.
Check that preparation isn’t the skill the assignment is meant to measure. If students need to show they can independently develop research questions, having AI supply those questions changes the evidence they produce.
Level 3: Use AI for feedback. Students produce an initial draft or solution, then ask AI to point out possible weaknesses. For instance, they could ask whether an argument includes an unsupported claim or whether an explanation skips a step.
Create your first draft independently. You may then ask an approved AI tool for feedback on clarity, organization, or reasoning. Do not ask it to rewrite the assignment. Submit your original draft, your revision, and a short explanation of one suggestion you accepted or rejected.
A prompt can keep the tool focused on feedback rather than authorship:
Read my draft and identify two places where my reasoning needs more explanation. Ask questions that help me revise. Do not rewrite the text.
Student work should only be uploaded through tools and processes approved by the school.
Level 4: Use AI for named components. Some tasks allow AI to produce one defined element while students handle the rest. A student might use AI to generate a counterargument, then assess its assumptions and evidence and write a response.
You may use AI to generate one counterargument to your position. You must independently evaluate that counterargument, check its claims, and write your response. Include the AI-generated counterargument and identify the tool you used.
Name the permitted component precisely. “AI assistance allowed” doesn’t tell students whether they can generate an introduction, translate a passage, write code, or produce the whole submission.
Level 5: Use AI throughout the task. This fits assignments where directing and evaluating AI-supported work are central to the learning. Students might develop a public information resource with AI, verify its claims, compare versions, and explain their editorial choices.
You may use approved AI tools throughout this assignment. You are responsible for checking claims, selecting appropriate sources, and explaining your decisions. Submit the final product with a short account of how you used AI, what you changed, and why.
The assessment should make those decisions visible. A polished final product by itself may reveal little about what a student understands.
Make the instructions usable
A permission level needs concrete instructions to guide student choices. Each assignment should explain:
- Purpose: What learning should students demonstrate?
- Permitted uses: What tasks may AI support?
- Boundaries: Which parts must students complete themselves?
- Disclosure: What should students report about their AI use?
- Evidence: Which drafts, annotations, explanations, or demonstrations should they submit?
Keep the evidence request in proportion to the assignment. A short feedback exercise may only need a brief account of revision choices. A larger AI-supported project may call for selected examples and an explanation of significant decisions.
Copyable assignment language
AI permission for this assignment:
[Insert the permission level and description.]You may use AI to:
[List permitted uses.]You must complete these parts yourself:
[Identify the thinking and production being assessed.]Use only these approved tools:
[List tools or explain the approval process.]When submitting, include:
[Specify disclosure and process evidence.]If access is a problem or you are unsure about a use:
Speak with me before beginning so we can clarify expectations or arrange an appropriate alternative.
Discuss examples with students before they begin. Is requesting feedback allowed? Can they accept a rewritten paragraph? May they ask AI to explain a reading they’re expected to interpret independently? Working through cases can expose unclear wording and give teachers a chance to explain the purpose behind a boundary.
Clear permissions make AI’s role in the work visible to both students and teachers. The practical test is whether students can tell what help is allowed, what work remains theirs, and what they need to show when they submit.
The post Assignment AI-Permission Scale: A Practical Guide for Teachers appeared first on Educators Technology.
Discussion
Sign in to join the discussion.