Schools and universities are spending significant energy deciding when students may use generative AI.
The rules matter. But a rule about tool use cannot repair an assignment that lacks a clear learning intent.
If educators do not know which capability a task is meant to develop, they cannot decide which parts AI should support, which parts students must perform independently, or what evidence would demonstrate genuine learning.
The central question is therefore not, “Is AI allowed?”
It is, “What human capability is this learning experience designed to build?”
Technology policy cannot substitute for pedagogy
The OECD Digital Education Outlook 2026 argues that generative AI can support learning when its use is guided by clear teaching principles. This distinction is essential.
AI can explain, generate examples, simulate feedback, translate, summarize, and adapt material. The same capabilities can also remove the productive difficulty through which a learner develops understanding.
Whether AI helps or harms learning depends less on the tool itself than on the role it plays inside the learning design.
A blanket ban ignores useful support. Unrestricted use can hide weak understanding. Neither approach begins with the educational purpose of the task.
The missing layer is learning intent
Learning intent is more specific than a course topic or assignment instruction. It names the change that should occur in the learner.
For example, an essay may be intended to develop several different capabilities:
- retrieving relevant evidence;
- evaluating source credibility;
- constructing an argument;
- organizing ideas;
- revising language;
- defending a judgment.
AI support affects each capability differently. Using AI to improve grammar may preserve the work of argument construction. Using AI to generate the argument may bypass it.
Without a clear learning intent, both teachers and students are forced to debate tools at the surface level.
Design AI roles around the capability
A practical model is to assign AI one of four roles in a learning task.
1. AI as access support
AI may clarify terminology, translate instructions, adjust reading level, or help a learner enter the task. The objective is access, not substitution.
2. AI as practice partner
AI may generate questions, simulate cases, offer counterarguments, or provide repeated rehearsal. The learner remains responsible for the reasoning.
3. AI as feedback instrument
AI may identify gaps, test consistency, or suggest revisions. Students should still evaluate whether the feedback is valid and explain what they changed.
4. AI as object of critique
Students may compare models, detect errors, examine bias, verify sources, or analyze how a response was constructed. Here, evaluating AI is itself part of the learning.
Each role should be connected to a named capability and a visible form of evidence.
Evidence should match the learning intent
When final answers become easy to generate, assessment must look beyond the final artifact.
Depending on the objective, evidence may include:
- an oral defense;
- annotated source choices;
- a record of revisions;
- comparison between alternative approaches;
- explanation of rejected AI suggestions;
- application of the same principle in a new context;
- reflection on uncertainty and limitations.
This is not about surveillance. It is about making learning visible.
A design protocol for educators
Before assigning any AI-enabled task, educators should answer five questions.
- What capability should the student strengthen?
- What productive struggle is necessary for that capability?
- Where can AI increase access, practice, or feedback without removing that struggle?
- What evidence will show that the learner—not only the tool—can perform?
- How will students explain and evaluate their use of AI?
These questions turn AI policy into instructional design.
Conclusion
Education cannot govern AI well by focusing only on permission.
The deeper responsibility is to define what students should become capable of doing, then design the role of AI around that purpose. Clear learning intent allows institutions to move beyond the false choice between banning AI and accepting every use of it.
AI should not determine the learning design.
Learning intent should determine the role of AI.
Key Takeaways
- AI usage rules cannot compensate for unclear learning objectives.
- The same AI capability can support or bypass learning depending on its role.
- Educators should connect every AI use to a human capability and visible evidence.
- Assessment should reveal reasoning, revision, transfer, and judgment—not only final output.
