If AI can write the homework, change the homework

Whenever I talk to teachers these days, the same worry comes up: "Students are getting AI to write their assignments. How are we supposed to tell what is really their own work?" It is a fair concern. But the first reflex, which is to ban the tools and rely on detection software, leads in the wrong direction.

Reliably identifying machine-generated text remains genuinely difficult. Detection tools produce false results and can cast unfair suspicion on a student who wrote every word themselves. Building an education system around an alarm that cannot give a definite answer is fair to neither students nor teachers.

The real question: what does the assignment measure?

Perhaps what this debate really reveals is that some assignments were never measuring learning in the first place, only the ability to produce a certain kind of text. "Write a two-page summary of this topic" has been vulnerable to copy and paste since the early days of the internet. AI has simply made the weakness impossible to ignore.

An assignment a machine can finish in seconds was never telling us much about what a student had learned.

Instead, we can design work that makes the process visible. Drafts that show how a student improved an argument after feedback. Ideas defended aloud in class. Projects that require investigating a local issue and observing it first-hand. These tasks are hard to outsource to a chatbot, and far closer to what learning actually is.

We can go further and invite the tools into the classroom. Asking students to critique a machine-written text, find its mistakes and question its sources may be one of the most relevant exercises in critical thinking available today. A generation that will grow up alongside these tools needs to learn how to interrogate them, not pretend they do not exist.

None of this is free. Process-based assessment places new demands on teachers who already face large classes and crowded curricula. That is exactly why the burden should not fall on teachers alone. Assessment itself needs rethinking at the level of education policy.

Every new technology has triggered a similar panic in schools: the calculator, the internet, the search engine. Each time, the winners were not those who banned the tool, but those who asked afresh what learning means. The question today is the same: what do we actually want our students to learn?