On October 23, 2026, MIT News reported on legal scholar Martha Minow's fall Compton Lecture at MIT. Her discussion connected democracy with the ways people use AI, including its place in education. She argued for tools that support learners through questions and hints, and challenged her audience to improve assessment. The practical question for a student is what that approach would look like during an ordinary study session.
A warning about dependence on AI can leave you uncertain about what to do differently tomorrow. Treating every chatbot interaction as either helpful or harmful is too blunt a starting point. Instead, examine the work that remains yours: identifying the question, attempting a solution, evaluating evidence, and explaining a conclusion. Those are observable actions you can record. The following audit is our suggested exercise inspired by the lecture, not an assessment instrument tested or endorsed by Minow or MIT.
Begin with a small task you can check against reliable course material. A practice problem with a worked solution, or a concept covered in your notes, is more useful than an unfamiliar topic whose answers you cannot verify. Write down your own first attempt before opening the tool. Keep both versions so you can distinguish what you already understood from what the interaction added. Do not upload private information, classmates' work, or material you lack permission to share.
Minow's call for stronger assessments also raises a useful question for learners: how could you make your own reasoning visible? As a practical suggestion, keep a short record of the step you found difficult and the explanation you eventually accepted. A polished answer alone does not tell you how it was reached. You might ask a teacher whether an explanation of the method, an oral follow-up, or a revision note would fit the task. These are possible learning activities, not a description of requirements at MIT or any other institution.
For a student, this reframing turns a vague anxiety into a practical design challenge: what would it actually mean to use AI so that deep thinking, memory, and curiosity are strengthened rather than bypassed? And, just as practically, how can you tell whether a given tool passes that test?
Here is a simple audit you can run on an AI tool you use for permitted study tasks. Our practical question, inspired by Minow’s discussion of tutoring, is whether the interaction asks you to think before revealing an answer. Look for questions and hints, then add your own explanation of the reasoning.
First, look at the default behavior. When you paste in a problem set, does the tool immediately produce a finished solution? If so, it is an answer dispenser. A tutor-style tool — or a tutor-style prompt you write yourself — would ask what you have tried so far, which part of the problem confuses you, or what concept the problem is testing. The delay is the point: the gap between your question and the answer is where your own thinking happens.
Second, check whether the tool invites self-explanation. Good tutoring does not end at the correct answer; it asks you to say why the answer is correct, to restate the reasoning in your own words, or to predict what would change if one condition changed. If a tool never asks you to explain anything back, it is treating you as a delivery address, not a learner.
Third, notice what you can recall afterward. A week later, can you reconstruct the solution without the tool? Try a comparable problem and record where you still need help. This is a personal check, not proof that a particular prompt caused a change in memory; task difficulty, prior knowledge, and practice may differ. If you repeatedly cannot explain a step, return to your course material or ask a teacher. Save the question you need answered rather than only saving the chatbot's solution.
Fourth, consider whether the interaction opens a useful question for further study. After using the tool, write one question you could investigate with your notes, a textbook, or a teacher. You might compare two approaches, ask why a condition matters, or look for a counterexample. Check suggested connections rather than assuming they are accurate. The useful result is a next step you can explain and verify, not simply a longer conversation or an impressive-looking answer.
None of this requires new software. Much of it can be achieved by changing how you prompt the tools you already have. Which leads to this week's concrete action: pick one workflow where you currently use AI as an answer dispenser — a homework problem, a draft essay, a coding exercise — and replace it with a tutor-style interaction. Before asking for any answer, ask the tool for hints and counterquestions first: give me three hints without revealing the solution; ask me two questions that test whether I understand the underlying concept; after I answer, ask me to explain my reasoning back to you. Then compare how much you retain a few days later compared with the old workflow.
Apply the same questions when comparing courses, programs, or workshops. Read the actual AI policy and ask which forms of assistance are allowed for each activity. Permission to use hints on practice work does not imply permission to use AI in an assessed submission. Ask how feedback and assessment help students show their reasoning: possible examples to discuss include oral explanations, process notes, or drafts with visible revision. Do not infer teaching quality from a single policy feature or assume that monitoring rules settle the question. Compare the support offered, the expectations for independent work, and your own learning needs. Where a rule is unclear, ask the instructor before using the tool.
For your next study session, make the experiment modest enough to complete. Choose one permitted practice task, save your unaided attempt, request a hint, and explain your next step in your own words. Check the result using reliable material, then note what you still cannot explain. Return to that gap later without the chat open. The exercise does not promise better grades or establish that AI is the right tool for every task. It gives you a concrete record of whether you did the reasoning you intended to practise, and what support you should seek next.
中文摘要
MIT News于2026年10月23日报道了Martha Minow在MIT的演讲，她主张教育AI支持学生思考，并呼吁改进评估。本文据此提出个人学习练习，而非经过MIT测试的方法：在规则允许的任务中，先独立尝试，再索取提示、解释推理、用可靠资料核对，并在稍后不借助工具自查。比较课程时应了解具体AI政策、反馈和独立完成要求；不要上传私密材料，不把练习效果或政策特点当作成绩与教学质量的保证。