Close your eyes and picture a classroom from your childhood. Your teacher has just announced a pop quiz or handed out an assignment. What is almost certainly on the page in front of you?
A question.
That is how most of us learned, and, perhaps unsurprisingly, it is how many of us now teach. It occurred to me recently that one of the most consequential things I do when creating an assignment happens before my students ever see it: I choose the question.
I decide what problem deserves attention and what a successful answer should look like. Then I hand the whole package to students with a rubric that accounts for every possible point.
For generations, that division of labor made sense. Teachers asked; students answered. From our earliest years in school, we have been rewarded for raising our hands and giving the right answer.
Artificial intelligence changes that division of labor. Give AI a clear question and enough context, and it can generate a string of answers before you’ve had your first sip of coffee.
If answer production is changing, our assignments should change with it. As answers become easier to produce, it may be more valuable to know what question is worth asking.
The intervention can be modest: every so often, make students do some of the intellectual work that usually happens before the assignment begins. Here is what it might look like.
1. Remove the question
Give students a situation instead. Imagine a short video or pamphlet showing that food waste has risen sharply in the campus dining hall. Include a few data points, student comments, menu changes, and operating information. Then stop.
Do not ask, “How should the university reduce food waste?”
Instead ask: “What do you need to know before you can decide what to do?”
This example is intentionally ordinary. A chemistry professor could substitute an unexpected lab result. A political scientist could present a policy outcome. A business professor could use a struggling organization.
The point is not the topic. It is that students encounter the situation before someone else defines the problem for them. One of their first tasks may even be deciding whether there is a problem at all.
2. Build a question funnel
Have students generate questions on their own, perhaps five each, then put them in groups and pool the results. A group of five now has 25 questions, which is where the interesting work starts.
Have them combine, cluster, and rank. Which questions could change how they define the problem? Which need to be answered first because others depend on the answer? Which sound interesting but probably will not change the decision?
The goal is not to produce 25 clever questions. It is to learn why three matter more than the other 22.
3. Test competing explanations
Before students recommend anything, ask them to explain the situation in three different ways.
Maybe food waste increased because portions grew. Maybe schedules changed and students have less time to eat. Maybe food quality declined.
Or perhaps the increase reflects something else entirely: another campus dining hall temporarily closed, sending more students to this one.
That last possibility matters. Not every troubling data point represents a problem that needs fixing. Sometimes the most important conclusion students can reach is that their original interpretation was wrong.
Then ask what evidence would help distinguish among those explanations, or my favorite question: What would have to be true for your current explanation to be wrong?
Now students have to investigate rather than simply defend their idea.
4. Map the process
Finally, ask students to make their thinking visible with a one-page inquiry map built around five questions: What did we notice? What might be going on? What matters most? What would change our minds? Where did we land?
Keep it to one page. The purpose is to show how their thinking changed, not to archive every thought or 30 pages of notes.
AI belongs in this process. It can help surface additional questions and counterarguments or reveal overlooked explanations. But students remain responsible for deciding what belongs on the map and what changed their thinking.
Now we can assess the path, not just the destination.
Only then, start solving
Once students have framed the problem and tested competing explanations, the familiar work of answering can begin. Problem-solving still matters. But so does the intellectual work of problem-finding.
We have spent generations teaching students how to answer the questions we give them. AI makes it increasingly important that they also learn how to decide which questions deserve answering.
