09 September 2026
AI Training for Teachers: Designing homework in the age of AI CAS event
If you were unable to join us for AI Training for Teachers: Designing Homework for the Age of AI online community meeting, don't worry! You can catch up on all the content and a recording of the session below.
The session, led by BCS Computing subject expert Ben Davies, focused on how homework can continue to develop genuine understanding when pupils have ready access to tools such as ChatGPT, Claude, Copilot and Gemini.
Designing Homework for the Age of AI: Keeping the Thinking with the Learner
Key Takeaways
- The finished piece of homework can no longer be assumed to provide reliable evidence that the thinking behind it was carried out by the pupil.
- Homework design should begin with the learning and thinking we want pupils to practise, rather than simply the product we want them to submit.
- AI does not need to be either banned or embraced wholesale: its appropriate role will depend on the purpose of each task.
- AI can support learning as an organiser, editor, thinking partner or generator, provided the pupil retains responsibility for judging and using its output.
- Clear boundaries, evidence of the learning process and explicit AI literacy should become part of our thinking about homework design.
Rethinking what homework tells us
The session began with a useful reality check. This was not going to be a webinar about detecting AI-generated homework or finding tasks that AI cannot complete. As Ben pointed out, attempting to remain permanently one step ahead of rapidly developing technology is unlikely to be sustainable. Instead, the question was: how can we redesign homework around the technology pupils already have access to, while keeping them actively engaged in learning?
That distinction matters because many familiar homework instructions now look remarkably like basic AI prompts. Write, research, summarise, compare and explain can all be copied into a chatbot, which can produce a polished response almost instantly. A pupil can then ask the system to alter its vocabulary, tone or level of formality before submitting the result.
The issue is therefore not simply whether AI has been used. It is whether the pupil has done the thinking the task was intended to develop.
Traditionally, we have often treated the final product as a proxy for learning. A strong explanation suggested that a pupil understood the concept; a correct answer suggested that they understood the process. The session challenged us to reconsider that assumption. Generative AI can now produce sophisticated-looking work without the learner necessarily understanding very much of it at all.
For Computing teachers, this is particularly relevant. A functioning program, accurate explanation of an algorithm or convincing evaluation of a technology does not, by itself, tell us who did the reasoning. That raises a useful question whenever we set independent work: what evidence would actually demonstrate the learning we care about?
Start with the thinking, not the output
A recurring theme throughout the meeting was the purpose of homework. Are pupils retrieving knowledge? Practising a process? Connecting concepts? Applying understanding to a new situation? Making a choice and justifying it? Evaluating an answer? Improving something? Challenging an assertion?
Once that purpose is clear, the task can be designed around the thinking pupils need to do rather than simply the artefact they need to produce.
This also gives us a more useful way to decide whether AI belongs in a particular activity. There will still be perfectly good reasons to specify no AI. If pupils need to strengthen recall, develop fluency, practise coding independently or rehearse a process, using an AI tool could remove exactly the practice the homework is supposed to provide.
At other times, AI could contribute without taking over. The session described this as a continuum rather than a simple choice between "AI allowed" and "AI banned".
Four useful roles for AI
Ben suggested four ways in which a chatbot could be incorporated into homework while leaving intellectual responsibility with the pupil.
AI as an organiser can help pupils structure material they have already produced. Their notes, examples or initial ideas come first; the AI helps them consider how those ideas might be arranged.
AI as an editor comes later in the process. The pupil produces a first draft and asks the AI for feedback or possible improvements. Crucially, the pupil then decides which suggestions are useful rather than automatically applying everything the system recommends.
AI as a thinking partner can challenge a pupil's position, suggest an alternative explanation or ask questions they have not considered. This can require the learner to defend a decision and apply their subject knowledge rather than simply retrieve an answer.
Finally, AI as a generator may sound counter-intuitive, but the important distinction is what it generates. Rather than producing the homework to submit, AI might produce an answer, argument, program or explanation that the pupil must evaluate, critique, correct or improve.
Across all four approaches, the learner remains responsible for the final judgement. AI may contribute suggestions; it does not become the decision-maker.
Making the learning process visible
Another useful part of the session was the emphasis on evidence. This was not about requiring pupils to provide an exhaustive audit trail to prove that they had not cheated. Instead, teachers might ask for a small piece of evidence that reveals something about the learner's decision-making.
That might be a first draft alongside selected AI feedback, a question generated while using AI as a thinking partner, or several suggested structures together with an explanation of which one the pupil chose and why.
A simple homework design process therefore emerged:
What subject learning should take place?
What thinking must remain with the pupil?
Could AI support that thinking without replacing it — perhaps as an organiser, editor, thinking partner or generator?
What small piece of evidence would make the pupil's thinking visible?
What will the pupil learn about the responsible use and limitations of AI in the process?
The practical challenge given to attendees was to take an existing homework activity and redesign it using those questions. Even a small change — such as adding an explicit boundary around acceptable AI use or asking pupils to justify one decision — could be a useful starting point.
AI literacy and safeguarding belong in the conversation too.
The session also stressed that encouraging purposeful use of AI comes with responsibilities. Pupils need an accurate understanding of what these systems are doing. Fluent, confident language does not mean that a chatbot "knows" the answer, and the session included an example in which an AI system confidently misidentified and miscounted animals in an image.
Teachers were also reminded to consider the information pupils enter into AI systems, particularly personal or sensitive information, and to avoid language that unnecessarily anthropomorphises AI by presenting it as if it thinks, knows or understands in a human sense. Safeguarding considerations and responsible use therefore need to sit alongside the pedagogical questions.
This is particularly timely because the 2026 Keeping children safe in education guidance came into force on 1 September 2026, and the Department for Education also provides specific guidance on safe use of generative AI in educational settings.
Supporting our own AI confidence
The final part of the meeting highlighted AI Confidence, a series of free, self-paced BCS CPD modules for people working in education. The modules use school-based scenarios and explore both opportunities and limitations rather than presenting AI as a solution to every problem. The session highlighted modules covering getting started with AI, AI and the world of work, school leadership and inclusion.
The programme remains free and does not require BCS membership; the current BCS site describes modules as taking roughly 45–60 minutes and awarding a digital badge following successful completion.
Perhaps the most useful message to take back to our Computing departments is that the future of homework does not necessarily depend on finding increasingly elaborate ways to prevent pupils using AI. A more sustainable question is: how can the task make the valuable thinking clear enough that pupils understand why it is worth doing themselves?
Next Steps
As you review homework across your own classes, you might ask yourself:
- What learning am I actually hoping this homework will secure, beyond producing something for me to mark?
- Which part of the thinking does the pupil need to practise independently?
- Could my current instructions be pasted almost unchanged into a chatbot?
- If AI is appropriate, which role — organiser, editor, thinking partner or generator — would support rather than replace the intended learning?
- What could pupils submit that gives me useful evidence of their reasoning without creating an excessive marking burden?
- Are my expectations about when AI may and may not be used clear to pupils?
- What does the task teach pupils about evaluating AI output, rather than simply accepting it?
For Computing lessons, some possible activities might include asking pupils to debug an AI-generated program, explaining each correction rather than simply producing working code; giving pupils two AI-generated solutions to the same algorithmic problem and asking them to compare correctness, efficiency and readability; providing an AI-generated trace table containing deliberate or naturally occurring errors for pupils to identify and correct; asking pupils to write an explanation of a concept such as iteration or abstraction independently before using AI as an editor and recording one suggestion they accepted and one they rejected; or asking AI to propose a solution to a cyber-security or networking scenario and requiring pupils to critique the proposal using knowledge from the lesson.
In each case, it is worth identifying in advance exactly where the pupil's thinking must remain visible.
Further Resources
The following links either featured in the session or provide useful follow-on material for CAS members:
AI and Agency: Helping Young People Think for Themselves — CAS recap and recording
AI in education: a practical starting point for teachers — CAS recap and recording