Should students be allowed to use AI in school?

The sensible position is task-specific permission rather than a blanket rule — because the honest answer depends entirely on what's being learned and what stage the student is at. The principle that resolves most cases: **protect the struggle that builds the skill; allow the assistance that doesn't.** Learning happens through effortful retrieval and problem-solving. Any use that removes the effort removes the learning, even when the output is excellent. Any use that adds support around the effort is usually fine or actively good. Applied concretely: **Restrict, especially early:** - Foundational skills being built for the first time. Writing a first essay, learning arithmetic, learning to construct an argument. The clumsy early attempts *are* the mechanism. - Assessments measuring whether the student can do it unaided. **Allow, and often encourage:** - Explanation and tutoring. A patient explainer available at 11pm is a genuine equity improvement — it's most valuable to students without educated parents or paid tutors, and that's not a small consideration. - Practice question generation and self-testing. - Feedback on a draft the student wrote. - Language support for students working in a second language, which is a levelling of an unfair playing field rather than an unfair advantage. - Anything where the AI is the object of study — evaluating its output, finding its errors, which is itself a critical skill. **The design change that matters more than any policy**: assessments that are resistant by construction. In-class writing, oral explanation of submitted work, projects tied to specific class discussion, drafts and process portfolios, and 'critique this AI-generated answer' tasks. Detection is a losing arms race — detectors are unreliable and falsely accuse non-native speakers disproportionately — while assessment redesign works permanently. **On the argument that bans are simplest**: they don't hold. Students use it anyway, secretly, without guidance, and learn the worst possible use pattern. A ban converts a teachable moment into a hidden habit. It also leaves students unprepared for workplaces where these tools are standard. **What students most need taught explicitly**, and rarely are: that fluent explanations create a strong feeling of understanding that isn't understanding; that verification is their responsibility; and that the check is whether they can close the tab and explain it. Teaching that metacognitive skill is more valuable than any rule about when the tool is permitted. The practical recommendation for a course: state per-assignment whether AI is permitted, permitted-with-disclosure, or prohibited, and say why. Ambiguity is what produces both anxious honest students and casual cheating.