Purposeful homework teacher judgement and Teaching Artistry

An AI tool can generate a worksheet in twenty seconds. The difficult question comes next: does this worksheet belong to this lesson, these learners and this moment in their learning?

During an ITI Teachers Sharing Practice workshop, Saeid Farid demonstrated a homework generator designed to turn the content of a completed lesson into targeted consolidation. Participants entered the language focus, learner level, age group, intended homework time and preferred task types. The tool then produced a sequence of activities and an answer key.

The speed was impressive. Yet the most important part of Saeid’s process appeared in a small optional field: What happened in class?

That question takes us directly into Teaching Artistry. Useful AI-supported homework begins with a teacher who has noticed learners closely enough to describe what they attempted, where they hesitated and what they may be ready to do next.

Speed solves only one problem

Teachers often rely on generic homework because creating tailored material takes time. A workbook exercise may practise the same grammar but have little connection with what happened in the room. One task may be too demanding for some learners and repetitive for others. Even when teachers want to respond more precisely, the next class is already approaching.

Saeid proposed four steps for using AI more purposefully:

Identify the learning focus from the completed lesson.

Describe the learners and the intended level of challenge.

Prompt the AI with clear constraints and suitable task types.

Review the output before assigning it.

The sequence is practical, but its quality depends on the human decisions inside each step. AI can accelerate production. It cannot decide by itself which moment from the lesson deserves to continue.

The lesson leaves a trace

Saeid used the phrase ‘Every lesson leaves a trace.’ This trace is more than the item written in a plan or the page completed in a coursebook. It includes what learners actually did with the material.

Perhaps several learners used should to instead of should plus the base form. Perhaps they could complete a gap-fill but avoided the target language when discussing a real problem. Perhaps the topic produced unusual energy, or one example confused the group. These are traces of learning that a responsive teacher can carry into the next task.

The teacher therefore prompts from observation rather than from the syllabus alone. ‘Create A2 homework on giving advice’ may produce something plausible. A richer prompt explains that learners practised should and shouldn’t, that several added to after should, and that the next task should help them move from choosing a form to using it in a short exchange.

This is pedagogic noticing translated into material design. The technology starts working only after the teacher has interpreted the room.

A level label cannot describe a class

CEFR levels can help an AI system estimate vocabulary, sentence length and task complexity. They cannot capture the variation inside a real group. As one participant observed, an A2 class may contain learners near the beginning of the band and others already approaching B1.

Difficulty also comes from more than language. A familiar topic can make a demanding structure easier to use. Complex instructions can make a simple exercise inaccessible. A writing task may require ideas, cultural knowledge or digital access that the teacher never intended to test.

A stronger prompt therefore describes support as well as level. It can specify language learners already know, examples they have seen, the time available and the kind of response expected. It can request an accessible starting point and an optional extension rather than producing one supposedly perfect level for everyone.

Review is an act of teaching

Saeid’s final step was review and assign. This deserves more attention than the generation itself.

An AI output may look polished while drifting away from the lesson. A worksheet intended to consolidate should and shouldn’t may quietly introduce several new advice phrases. An answer can be grammatically possible but unnatural. A health recommendation may be questionable, or an apparently simple question may depend on a cultural assumption unfamiliar to learners.

Reviewing therefore involves professional judgement, not just proofreading. Before assigning an AI-generated task, a teacher can ask:

Does every activity serve the learning focus?

Does the task recycle the language learners met, or introduce unplanned demands?

Are the instructions easier to understand than the activity itself?

Could learners with different levels of confidence make a meaningful start?

Are the content, examples and advice accurate, respectful and appropriate for this group?

What will I learn from their responses, and what will I do with that information?

The last question reconnects material design with the teacher’s purpose. A worksheet becomes educationally valuable when it helps the teacher and learners see what is developing.

Homework should return to the classroom

One of the strongest suggestions in Saeid’s materials was to bring the homework back into the next lesson. Learners might compare answers in pairs, perform one dialogue, notice a common difficulty or improve one response. The teacher might use two successful examples as the next warmer.

This creates continuity. The previous lesson informs the prompt, the prompt produces a draft, the teacher reshapes it for the group, and learners’ responses influence what happens next. Homework no longer disappears between classes. It becomes part of an ongoing relationship between teaching and learning.

This also changes the meaning of feedback. If many learners struggle with the same item, the result may tell us something about the original explanation, the task or the assumptions built into the material. AI does not remove the need for teacher reflection. It can provide another surface on which the consequences of our teaching become visible.

Teacher artistry extends beyond the live lesson

Teaching Artistry is often associated with what a teacher does in the room: voice, gesture, presence, responsiveness and the ability to work with emerging interaction. Saeid’s workshop reminds us that artistry also appears in what happens after the class.

The teacher selects the trace worth following. The teacher decides how much challenge the group can use, what support will preserve agency and which generated material should be rejected. The teacher also protects learners by avoiding personal data in prompts and by questioning confident but unsupported AI output.

AI can make production faster. Purpose still comes from a human educator who has paid attention.

Join the conversation

Teaching Artistry explores the human, embodied and relational dimensions of teaching, including the professional judgement teachers bring to new technologies. Join the free Performative ELT community: https://performativeelt.com/free-community-6779