The Purpose of Homework In The Age of Artificial Intelligence

How teachers could set homework that deepens learning and understanding

Photo by Annie Spratt on Unsplash

Every day in classrooms all around the world, teachers set homework.

Now we have AI in our lives. Some people hate it, others welcome it, and others put up with it. It’s having a big effect on education and, in particular, on homework.

As a maths teacher of 30 years, it’s giving me lots to think about. Do I teach using AI or do I give my students the chance to use AI in their homework? I can’t be with them as they work on an assigned task, so I need to be able to trust them to use AI in a responsible way.

Over the last year I’ve been thinking about the purpose of homework now that AI is very much part of the educational landscape.

The purpose of homework in the 21st century

When I was at school and when I trained as a teacher, I believed that homework helped students to better understand what had been taught in a lesson, develop independent study habits, resilience, and perseverance.

I believed that students’ critical thinking, creativity and organisational skills were all helped by doing regular homework.

Now I think a bit differently.

I think homework should engage students in applying concepts, exploring new perspectives, and developing confidence in expressing their ideas. AI could be used to help them develop the skills they have learned using the AI as an assistant to explore a topic in more depth.

I know that there are people who hate AI and see it as something that robs people of their ability to think for themselves. I disagree. I see AI as a useful tool that can deepen understanding and learning if used in the correct way.

I encourage my students to use AI to help them think more deeply about a topic, to reflect on a topic, to follow up on feedback given on previous homework, and to organise all their homework in a way that helps them build on their skill set.

I remember back to my own days at school.

I was a student who was good at surface learning. There’s a lot of stuff I memorised, especially in subjects that I didn’t enjoy. I could pass tests and exams in those subjects, but my understanding of them wasn’t good. In the 1970s and 1980s AI didn’t exist so I had plenty of opportunites to develop my memory skills well.

How AI can help students move to deep learning

I teach maths and computer science to students aged 11 to 18.

AI tools like Copilot, ChatGPT, and others can write great answers to maths questions, the trouble is those answers often look right but 50% of the time the reasoning is wrong and so the answer is wrong.

So whilst the AI looks like it can be used to cheat, for me as a maths teacher it’s easy to see when students have tried to use AI in their maths homework.

I know that in other subjects AI can give great answers to questions, but as I’ve come across AI written essays so much over the last few years like most teachers, I find it easier to spot essays written about computer science topics using AI more easily.

I know the level of writing a student can produce independently through our discussions and going through questions in our lesson.

When working with my students, AI can be a support or obstacle in learning so it’s important I spend time advising my students on how to use it well and ethically in their work.

When AI gets maths wrong

AI might seem like the ultimate maths cheat but in fact it gets answers to maths questions wrong about 50% of the time.

That’s interesting considering all I’ve been told about AI. So I’ve dug deeper into that. I found that the generative models that AI uses, are not calculators. They don’t do maths like a human or computer algebra system.

They’re pattern recognizers.

This means that they often get multi-step problems or logic puzzles wrong. AI is poor at thinking in logically correct steps. I’m always astounded though at AI attempts to reason through incorrect solutions, setting out steps that to a non-maths person look impressive.

AI has no understanding of the context of the curriculum, what is required to be shown mathematically or the methods a student has been taught, so the solutions it provides can be way off beam.

Students who rely on AI for maths homework risk embedding misconceptions.

So instead I suggest to my students that they can use AI to solve half of the problems I set, then they go through checking the steps carefully. If they find that a solution is wrong, they write out the correct logical steps and state the correct answer. Then they put together a question and answer of their own similar to the question they’ve just completed.

Brainstorming with AI

In computer science when I set an essay type question, I ask my students to start by using AI to brainstorm ideas on the topic.

AI can start ideas flowing where students are stuck, so instead of staring at a blank piece of paper and wondering what to write, they can get started with a list of ideas that AI gives them.

Once they’ve sorted through those ideas, they could write the essay or they could ask AI to give them an outline for an essay.

Then I ask them to make notes on the ideas they’ve accepted, rejected, or changed, and why they made those decisions.

Some students might want to move on to get AI to write that essay for them, but I remind them that the essay they produce has to be their own work. There is no satisfaction or achievement they can call their own, to be had from having AI write the essay for them.

I know the style my students use to answer an essay type question. Whilst that changes over time, it’s easy for me to see when it’s been written with AI. So we talk about what they’ve written and I remind them that the task set was designed to help them learn.

If AI has written the essay for them they might have come across some new concepts but has their understanding of the topic improved as a result? If the answer is no, then I ask them to rewrite the essay in a lesson using their own words.

Used transparently and as a thinking aid rather than a ghostwriter, AI can support students in sorting out and organising their ideas, without replacing the cognitive effort needed to develop them.

Teachers aren’t powerless when it comes to working with AI in and out of the classroom. Homework can be redesigned with all the original purposes in mind so that AI shortcuts are much harder for students to do.

AI-Resistant Homework

Here are some other AI-resistant homework ideas I’ve come up with over the last two years.

Personalised reflections

What was the most challenging part of today’s lesson for you, and why? This requires personal insight, which AI can’t fake convincingly.

Process-first submissions

I ask for drafts, mind maps, or voice notes describing how a student approached the task.

Classroom anchoring

I connect homework to class experiences, for instance, “Based on our group discussion, what do you think about the advantages and disadvantages of…”.

Real-world tie-ins

I give tasks where students need to observe, experiment, or research in their local context. In computer science, I’ve asked them to look at different pieces of code that achieve a set task and to write a short essay comparing the efficiency of each piece of code.

Set a “Compare AI versus me” task

I ask students to ask AI to generate a response to a set question and then I ask them to criticise or improve on the response that AI has given them.

The goal is to develop metacognitive awareness, helping students think about how they think, and how AI can assist them responsibly.

Difficulties in following up on student’s work

I don’t think any of this is easy.

There will still be some students who will take opportunities to cheat no matter how much effort a teacher goes to to explain how AI can be used ethically and responsibly.

Teachers today have a heavy workload so trying to find which students have used AI to do the heavy lifting for them is often a big ask.

Homework is often invisible at home and teachers often have little insight into how a task was done , or who did it. Time pressures on teachers can make monitoring and following up on missed or inadequate submissions difficult and some students may have digital aids, parental help, or quiet study space and others don’t.

So homework, while potentially powerful, can unintentionally widen the achievement gap.

Addressing AI Misuse

Once homework has been handed in and marked, a teacher has the opportunity to work with students on pieces that may have stood out as possibly AI generated.

I use oral follow-ups or short in-class quizzes to confirm understanding and to see if students really understand concepts they’ve written about.

I engage students in talking about their work, and I ask them what they found difficult, what surprised them, how they tackled a problem.

I make integrity part of the conversation reminding students of the need for intellectual honesty and invite their views on this.

I hope that I’m helping students to become more responsible members of society where they do their best at all times to complete set work. I aim to get them to choose AI as a useful assistant rather than as a tool that they use to complete their homework for them.

Changing homework tasks so that students can work with AI

Now we’ve moved into a world where AI is increasingly integrated into our lives, the world of education needs to move with that as well.

AI has changed the rules. But it also offers us a moment to pause and think about what matters most in learning. If we believe education should teach students how to think, reflect, communicate, and question, then AI can complement older style homework tasks well with a bit of adaptation.

Educators, parents, and students can work together to reimagine how and why we ask students to complete homework. Homework can still be a space for deep thinking and further exploration of a topic and it can still give students the opportunity to master skills learned in class.