{"course":{"slug":"ai-for-teachers","title":"AI for Teachers","url":"https://addaly.com/learn/ai-for-teachers","level":"beginner","track":"work"},"lesson":{"slug":"rethinking-homework","title":"Rethinking homework","url":"https://addaly.com/learn/ai-for-teachers/rethinking-homework","position":40,"of":57,"minutes":9,"answer":"Homework that asks for a product a model can produce (summaries, comprehension answers, definitions, essays) has stopped measuring anything; keep homework that is retrieval, practice with self-checking, reading with a one-line response.","keyIdea":"Homework that asks for a product a model can produce (summaries, comprehension answers, definitions, essays) has stopped measuring anything; keep homework that is retrieval, practice with self-checking, reading with a one-line response, or evidence from the student's own home and street."},"body":{"format":"markdown","text":"## What homework was for\n\nStrip away the tradition and homework has a few defensible purposes: practice of a skill taught in class, so that it becomes fluent; retrieval of material learned earlier, so that it stays; reading or preparation that a lesson will build on; and, occasionally, work that can only be done outside school: an observation, an interview, a measurement at home. The research on homework has always been lukewarm about anything else. Large reviews find little effect in primary school and a modest one in secondary, and the effect that exists comes from practice and retrieval, not from producing things.\n\nThat matters now because the model has made one whole category of homework (producing things) pointless as a measure of anything, and the lukewarm evidence says that category was never where the value was.\n\n## The tasks that have stopped working\n\nBe honest about the list. If a task is on it, the excellent homework coming back tells you nothing, and the check question's falling scores are the consequence.\n\n- **Summarise the chapter.** The model summarises better than any student.\n- **Answer the comprehension questions at the end of the chapter.** The chapter is in the training data; so, often, are the questions.\n- **Define these terms.** Instant.\n- **Write an essay on...** The assignment-redesign lesson covered this for assessed pieces. For homework it is worse, because there is no in-class paragraph to anchor it.\n- **Research X and write a report.** Search plus summarise, in one prompt.\n- **Translate this passage.** Any phone.\n\nSetting these still, with a ban attached, produces a class where the compliant students do a task that no longer teaches them anything a lesson would not, and the rest paste. Neither group learns from the homework. Stop setting them.\n\n## What replaces them\n\n**Retrieval, answered in the next lesson.** The homework is \"revise these three topics\". The check is five questions on paper at the start of the next lesson, answered from memory. There is nothing to generate, the learning happens in the revising, and you find out in five minutes who did it. This is the single best replacement and it costs no marking beyond a glance.\n\n**Practice with a self-check.** Twenty problems with the answers at the bottom of the page. The task is not the answers: they are given. The task is the working, and the student marks their own. A student who copies the answers has gained nothing and knows it; a student who works them has practised. Pair it with a starter next lesson on the same skill, and copying becomes visible without any policing.\n\n**Reading with a one-line response.** \"Read pages 40 to 44. Bring one sentence you did not understand.\" Or one question you would ask the author. The model can read for them; it cannot tell you what they did not understand. And the sentences they bring are the starter for the next lesson.\n\n**Evidence from home.** Measure the rainfall, interview a grandparent about prices, count the vehicles on the road at 8am, photograph three shop signs in two languages. The assignment-redesign lesson called this anchoring to the room; for homework, the anchor is the street. A model cannot do it and a student who does it has data nobody else has.\n\n**Preparation, then discussion.** Watch, read or listen to something short; the lesson is a discussion of it. Students who did not prepare are visible in the first two minutes and no product was ever asked for.\n\n## What about the practice essay\n\nOlder students preparing for written exams do need to practise writing under conditions. The honest answer is that this practice now belongs in class, timed, on paper, or in the version-history document the process lesson described. Homework can be the plan for it (five bullet points, a thesis sentence) and the in-class session is the writing. That is not a loss. Exam writing practised in exam conditions is practised properly.\n\n## Less of it\n\nA consequence of all this is less homework, and better. Teachers who make the change report that they set roughly half what they used to and mark a quarter of it, because retrieval checks mark themselves and self-checked practice needs a glance. Parents sometimes read less homework as less rigour. Explain the change in a letter: what was dropped and why, what replaced it, and what they will see instead: a child revising for a quiz rather than typing a summary.\n\n## The honest limit\n\nRetrieval and practice do not cover everything. The extended piece of writing, the sustained project, the thing that takes a student a weekend to make: those are real and they still belong somewhere, and the somewhere is now the classroom, the portfolio and the supervised session rather than the kitchen table. The tradeoff is class time, and the earlier lessons on generating materials are partly about winning back the time to afford it."},"check":{"question":"A teacher sets \"read chapter 5 and write a 300-word summary\" every week. Since AI arrived, the summaries are excellent and test scores on the chapters have fallen. What is the mechanism, and what should replace the task?","options":[{"text":"Students have got lazier since the tools arrived, and a stricter penalty for undeclared AI use would restore the learning.","correct":false,"tempting":"Treats a design problem as a discipline one; the task rewards a product a model produces well, and no penalty makes it measure reading again."},{"text":"The task rewards a product the model makes well, so it no longer causes reading; replace it with in-class retrieval.","correct":true},{"text":"Summaries are a poor task in themselves and should be replaced with longer essays that demand more of the students.","correct":false,"tempting":"A longer essay is even more the model's strength; the fix is a task whose value is in the doing, not a bigger product."},{"text":"The chapter is too hard for the class, and the summaries were always being copied from somewhere before AI existed.","correct":false,"tempting":"Copying existed before, but the sharp change follows a specific new capability; the task needs redesign regardless of the chapter."}]},"prev":{"slug":"handling-a-case-fairly","title":"Handling a case fairly","url":"https://addaly.com/learn/ai-for-teachers/handling-a-case-fairly"},"next":{"slug":"citing-and-declaring-ai-use","title":"Citing and declaring AI use","url":"https://addaly.com/learn/ai-for-teachers/citing-and-declaring-ai-use"},"authorship":"Founder and series editor: Dr Mustafa Mun. Written by AI models to a written standard. Exam keys checked blind by a second model: 3,665 of 3,666 agreed. Academic review invited.","review":{"status":"Draft, open for academic review"},"cite":"Addaly (2026) 'Rethinking homework', in AI for Teachers. Edited by M. Mun. Addaly. Available at: https://addaly.com/learn/ai-for-teachers/rethinking-homework (Accessed: 13 October 2026).","licence":{"name":"CC BY-NC-SA 4.0","url":"https://creativecommons.org/licenses/by-nc-sa/4.0/"},"reader":{"schema":1,"title":"Rethinking homework","courseTitle":"AI for Teachers","module":{"index":5,"slug":"integrity-and-ai-written-work","title":"Integrity: AI-written work and what to do about it"},"answer":null,"keyIdea":"Homework that asks for a product a model can produce (summaries, comprehension answers, definitions, essays) has stopped measuring anything; keep homework that is retrieval, practice with self-checking, reading with a one-line response, or evidence from the student’s own home and street.","hero":{"kind":"plate","image":null,"alt":null,"caption":null},"blocks":[{"t":"h2","id":"what-homework-was-for","runs":[{"text":"What homework was for"}]},{"t":"p","runs":[{"text":"Strip away the tradition and homework has a few defensible purposes: practice of a skill taught in class, so that it becomes fluent; retrieval of material learned earlier, so that it stays; reading or preparation that a lesson will build on; and, occasionally, work that can only be done outside school: an observation, an interview, a measurement at home. The research on homework has always been lukewarm about anything else. Large reviews find little effect in primary school and a modest one in secondary, and the effect that exists comes from practice and retrieval, not from producing things."}]},{"t":"p","runs":[{"text":"That matters now because the model has made one whole category of homework (producing things) pointless as a measure of anything, and the lukewarm evidence says that category was never where the value was."}]},{"t":"h2","id":"the-tasks-that-have-stopped-working","runs":[{"text":"The tasks that have stopped working"}]},{"t":"p","runs":[{"text":"Be honest about the list. If a task is on it, the excellent homework coming back tells you nothing, and the check question’s falling scores are the consequence."}]},{"t":"ul","items":[[{"text":"Summarise the chapter.","b":true},{"text":" The model summarises better than any student."}],[{"text":"Answer the comprehension questions at the end of the chapter.","b":true},{"text":" The chapter is in the training data; so, often, are the questions."}],[{"text":"Define these terms.","b":true},{"text":" Instant."}],[{"text":"Write an essay on...","b":true},{"text":" The assignment-redesign lesson covered this for assessed pieces. For homework it is worse, because there is no in-class paragraph to anchor it."}],[{"text":"Research X and write a report.","b":true},{"text":" Search plus summarise, in one prompt."}],[{"text":"Translate this passage.","b":true},{"text":" Any phone."}]]},{"t":"p","runs":[{"text":"Setting these still, with a ban attached, produces a class where the compliant students do a task that no longer teaches them anything a lesson would not, and the rest paste. Neither group learns from the homework. Stop setting them."}]},{"t":"h2","id":"what-replaces-them","runs":[{"text":"What replaces them"}]},{"t":"p","runs":[{"text":"Retrieval, answered in the next lesson.","b":true},{"text":" The homework is “revise these three topics”. The check is five questions on paper at the start of the next lesson, answered from memory. There is nothing to generate, the learning happens in the revising, and you find out in five minutes who did it. This is the single best replacement and it costs no marking beyond a glance."}]},{"t":"p","runs":[{"text":"Practice with a self-check.","b":true},{"text":" Twenty problems with the answers at the bottom of the page. The task is not the answers: they are given. The task is the working, and the student marks their own. A student who copies the answers has gained nothing and knows it; a student who works them has practised. Pair it with a starter next lesson on the same skill, and copying becomes visible without any policing."}]},{"t":"p","runs":[{"text":"Reading with a one-line response.","b":true},{"text":" “Read pages 40 to 44. Bring one sentence you did not understand.” Or one question you would ask the author. The model can read for them; it cannot tell you what they did not understand. And the sentences they bring are the starter for the next lesson."}]},{"t":"p","runs":[{"text":"Evidence from home.","b":true},{"text":" Measure the rainfall, interview a grandparent about prices, count the vehicles on the road at 8am, photograph three shop signs in two languages. The assignment-redesign lesson called this anchoring to the room; for homework, the anchor is the street. A model cannot do it and a student who does it has data nobody else has."}]},{"t":"p","runs":[{"text":"Preparation, then discussion.","b":true},{"text":" Watch, read or listen to something short; the lesson is a discussion of it. Students who did not prepare are visible in the first two minutes and no product was ever asked for."}]},{"t":"h2","id":"what-about-the-practice-essay","runs":[{"text":"What about the practice essay"}]},{"t":"p","runs":[{"text":"Older students preparing for written exams do need to practise writing under conditions. The honest answer is that this practice now belongs in class, timed, on paper, or in the version-history document the process lesson described. Homework can be the plan for it (five bullet points, a thesis sentence) and the in-class session is the writing. That is not a loss. Exam writing practised in exam conditions is practised properly."}]},{"t":"h2","id":"less-of-it","runs":[{"text":"Less of it"}]},{"t":"p","runs":[{"text":"A consequence of all this is less homework, and better. Teachers who make the change report that they set roughly half what they used to and mark a quarter of it, because retrieval checks mark themselves and self-checked practice needs a glance. Parents sometimes read less homework as less rigour. Explain the change in a letter: what was dropped and why, what replaced it, and what they will see instead: a child revising for a quiz rather than typing a summary."}]},{"t":"h2","id":"the-honest-limit","runs":[{"text":"The honest limit"}],"role":"limit"},{"t":"p","runs":[{"text":"Retrieval and practice do not cover everything. The extended piece of writing, the sustained project, the thing that takes a student a weekend to make: those are real and they still belong somewhere, and the somewhere is now the classroom, the portfolio and the supervised session rather than the kitchen table. The tradeoff is class time, and the earlier lessons on generating materials are partly about winning back the time to afford it."}]}],"check":{"question":[{"text":"A teacher sets “read chapter 5 and write a 300-word summary” every week. Since AI arrived, the summaries are excellent and test scores on the chapters have fallen. What is the mechanism, and what should replace the task?"}],"options":[{"text":[{"text":"Students have got lazier since the tools arrived, and a stricter penalty for undeclared AI use would restore the learning."}],"correct":false,"tempting":[{"text":"Treats a design problem as a discipline one; the task rewards a product a model produces well, and no penalty makes it measure reading again."}]},{"text":[{"text":"The task rewards a product the model makes well, so it no longer causes reading; replace it with in-class retrieval."}],"correct":true,"tempting":null},{"text":[{"text":"Summaries are a poor task in themselves and should be replaced with longer essays that demand more of the students."}],"correct":false,"tempting":[{"text":"A longer essay is even more the model’s strength; the fix is a task whose value is in the doing, not a bigger product."}]},{"text":[{"text":"The chapter is too hard for the class, and the summaries were always being copied from somewhere before AI existed."}],"correct":false,"tempting":[{"text":"Copying existed before, but the sharp change follows a specific new capability; the task needs redesign regardless of the chapter."}]}],"order":[0,1,3,2],"why":null},"prev":{"slug":"handling-a-case-fairly","title":"Handling a case fairly","minutes":9},"next":{"slug":"citing-and-declaring-ai-use","title":"Citing and declaring AI use","minutes":8}}}