Classifying and Indicating Permitted AI Use in Coursework

Regarding students’ use of AI in classes, each instructor needs to state clearly what AI use they permit, for each class and each assignment (see About the Use of AI Tools in Classes (ver. 1.0)). This page offers a set of typical choices for that purpose: a classification of AI use scenarios (S1–S5), and permitted AI use labels (T0–T5) indicating how far along that scale AI use is permitted. For each class and assignment, please choose one of these or “Other” and show it to your students.

From the 2026 A Semester, you will be able to select such a label and show it to students when you post an assignment on UTOL (scheduled for introduction at the end of September).

* Please note that communicating to students what AI use is permitted is necessary for every assignment, whether or not it is posted on UTOL.

Why Stating Permitted AI Use is Important

• When it is left unclear or ambiguous whether AI use is permitted, unfairness can arise between students who are restrained about using AI (some out of their own judgment in light of their learning goals, others out of fear that even minimal use will be treated as misconduct) and students who are not. This situation itself is a source of anxiety for students
• Our survey, conducted since 2026, shows that whether AI use is permitted often goes unstated, and that this leaves students with the kind of anxiety described above
• Whether AI use is treated restrictively or permissively, it is important to state it clearly. This helps create an environment in which students can learn and work on their assignments without needless anxiety
• Stating it will not by itself eliminate misuse, nor will it make misuse easier to detect. But without a rule, there is no basis on which to decide whether a particular use is misconduct at all. Setting a rule is a minimum prerequisite for addressing misuse

The University’s Basic Policy

• Whether it is appropriate for students to use AI should be decided in light of the educational goals of each individual class and assignment. This varies with the learning objectives of the assignment, the form of the work submitted, and the students taking the course (their year of study, etc.)
• Therefore, the University’s basic policy is that instructors should decide and give clear instructions for each individual situation and assignment. We value flexibility and the ability to reflect each instructor’s own philosophy, and so we do not draw a uniform line
• This page and this feature do not address the question of what policy is right --- in which situations AI use should or should not be permitted
• The basic principles are set out for instructors in About the Use of AI Tools in Classes (ver. 1.0) and for students in Notification to Students on the Use of AI Tools in Classes (ver. 1.0), which encourage awareness of responsible use in line with educational goals
• Some thoughts on the underlying principles are also given at the end of this page
• On that basis, whether AI use is permitted still needs to be stated clearly for each individual assignment

How to Indicate Permitted AI Use

• Even for a single assignment, there are many possible positions between a blanket ban and blanket permission. For example: “you may ask AI questions when you get stuck, but you may not copy AI output directly into your submission.”
• To reduce the burden on instructors of indicating what they permit, and to reduce ambiguity and the misunderstandings that follow from it, we offer a set of permitted uses to choose from. Below we first present typical use scenarios S1–S5, and then a set of choices (T0–T5) expressing how far along that scale AI use is permitted.

Classification of AI Use Scenarios

Examples of uses falling under each scenario:

• S1
  • ask about a point where you got stuck on a basic practice problem
  • ask about a point in the class explanation you did not understand
  • ask questions in order to acquire the background knowledge needed to follow a class
  • research a topic you are unfamiliar with
  • translate a non-native-language text you need to read for the assignment into your own language and read it
• S2
  • consult about the choice of topic or the line of argument for an essay
  • consult about goal-setting in an open-ended creative assignment
• S3
  • have AI review your work, particularly its writing and form, as in an academic essay
  • ask where your own answer went wrong
  • obtain feedback on your own answer in an exercise admitting multiple solutions
• S4
  • have AI produce a draft in an assignment whose learning goal lies in the work that follows the draft (evaluation, verification, improvement, experiment, analysis, etc.)
  • have AI generate baseline code for a programming assignment (simple code serving as the object of evaluation or the starting point for improvement)
  • have AI produce charts, tables, or data visualizations in an assignment whose main focus is the analysis and discussion that follows
• S5
  • have AI generate the claims to be refuted, in an academic essay or other essay in which students take a position and argue for it
  • critically examine AI output to learn the limitations of AI --- errors, bias, (lack of) reproducibility, differences in answers depending on the prompt (lack of consistency), and so on
  • evaluate AI output by checking it against peer-reviewed literature and other sources
  • learn how to work with AI itself: writing programs that incorporate AI (APIs), working with RAG, working with agents, and so on

Permitted AI Use Labels to Display to Students

• Ti means that S1 through Si are permitted
  • For example, T3 means S1 through S3, that is, up to and including questions/research, discussion/idea-bouncing, and feedback
  • Choose T5 when, in addition to permitting everything up to S4, you also permit other uses (e.g., working with or examining AI is itself part of the goal of the assignment)
• Note:
  • To repeat what is said in “The University’s Basic Policy” above, which of these labels is appropriate should be decided in light of the learning objectives of the assignment, the form of the work submitted, and so on. A higher number does not mean it is more “advanced,” let alone “better,” in any way

• If none of these fits, choose “Other” and describe what AI use you permit
  • Examples: permitted uses that are not cumulative (permitting S1 and S3 but not S2, for instance)
• When indicating a label, instructors will find it easier to gain students’ understanding and acceptance by stating the learning objectives of the assignment (not what students are supposed to hand in, but what they themselves are meant to gain or learn) and adding a brief word on why that label was chosen in light of those objectives

Beyond Labeling…

Messages to Students

Students are of course asked to follow the rules (not to commit misconduct; AI use that has not been permitted is treated the same as any other form of misconduct). In addition:

• It is important to be aware that most of the purpose of learning (not only in your university studies, but in every setting) is to inscribe something in your own brain and body, not the “work product” itself
• There are of course cases in which the work product has unique value in itself, where writing, creation, invention, or discovery emerges that is worth being widely read, appreciated, or used in its own right --- and aiming at such work is a good thing
• Much of the activity of learning, however, is devoted to practice in preparation for making such contributions in the future (e.g., writing an academic essay including its writing style and conventions), to learning established theories and writings, or to internalizing them (e.g., solving problems, writing programs)
• In that kind of learning, it is not at all the case that the effort is meaningless without an outcome of the sort described above: the significance and the value lie in your own growth --- that is, in the change that has taken place in your brain, your spinal cord, your nerves, and so on
• Put plainly, this is the same thing that has always been said --- “don’t do what does you no good” --- but being conscious of it has become extremely important now that generating the “work product” itself has become so easy
• Please do not miss the opportunity to grow (= change)

With that in mind, here are some general remarks on what we would like you to keep in mind for each label:

• S1: broaden your understanding of the assignment by clearly putting your questions into words (what it is that you do not understand), and by exploring and organizing the information and background knowledge you need
• S2: after putting your own thinking into words, encounter different perspectives and counterarguments, and thereby deepen, broaden, and refine your points, hypotheses, and ideas
• S3: look back at the answers and work you have produced, judge for yourself whether the AI’s comments are valid, and improve your work while choosing what to accept and what not to
• S4: rather than taking an AI-generated draft at face value, examine its accuracy, validity, structure, and so on for yourself, and select, revise, and restructure it in line with your own thinking and argument, so as to produce work you can take full responsibility for yourself
• S5: examine AI output critically, learn AI’s characteristics and limitations, and explore and create new questions, insights, and approaches to problems

Messages to Instructors

• Instructors and students alike share concerns about the negative effects of over-reliance on AI (loss of the capacity for critical thinking, loss of motivation to learn, and so on) (survey)

• Given that the principal goal of learning is “growth (change) in preparation for future contributions,” one might well think that it would generally be better not to use AI at all, and ask why we do not simply say so

• Here we need to face squarely the reality of learning without AI: that “almost every student gets stuck somewhere” and that “students do not have enough opportunity or time to ask questions even when they want to,” and that learning opportunities are sometimes lost as a result

• However much instructors welcome questions and answer them earnestly, there is a limit to the time available; there are therefore questions that never get asked, and students get stuck there and lose the learning opportunities that lie beyond --- we believe this happens often

• At such moments, AI --- which can be asked questions at any time, for as long as one likes, and which returns feedback immediately --- seems useful in many situations as a tool for moving learning forward

• Even before AI, especially on elementary and basic material, it was common to look at a worked solution instead of struggling for too long, and to learn through repetition; AI use seems to have a similar (or equal or greater) effect

• Beyond learning by studying the answer, it also becomes easy to obtain feedback on where one’s own approach went wrong or whether there is a better approach, and to pick up the background knowledge that should have preceded the problem

• It is important to take these points into account when deciding what AI use to permit. That is, we need to consider that the goal of an assignment is to enhance learning, not assessment

• Furthermore, in order to enhance learning, it is also necessary to reconsider the content taught in class and the design of assignments themselves --- where we want the student’s central experience in a given assignment to lie

• According to the survey, nearly 60% of the instructors who responded said that they “feel a need to substantially rethink existing course content and forms of assessment, now that AI has to be assumed”

• We take this to mean that many instructors are asking themselves not merely about the immediate problem --- that assignments AI can easily solve no longer work for assessment --- but a deeper question: if knowledge in the sense of a body of methods and know-how can be externalized to AI, what knowledge and abilities should human beings then acquire?

• Those would presumably include the capacity to scrutinize the claims of AI (and of other people), such as critical thinking and the ability to verify the truth and reliability of information for oneself; the integrative judgment of what to choose in light of the various facts and opinions available, AI output among them; responsibility for that judgment; and the ethical sense that underpins it

• It goes without saying that actually putting these into practice requires having mastered and internalized the fundamentals of a field and having experienced studying it deeply

• How to change education and learning in a world that has to assume the existence of powerful AI --- what to change and what not to change --- is not something for which anyone has already provided a model answer. It is a question we should keep exploring by bringing together the wisdom of instructors with deep expertise in their individual fields

• We intend to continue thinking about this and to create venues for sharing good examples of practice and for continuing the discussion