Narrowing and Sharpening PBL Project Topics with AI

In project-based learning, the place the most time leaks away is, unexpectedly, choosing the topic. Hand out something vague like "environmental problems in our neighborhood" and students spend a full week searching without finding a direction. On a four-week project, it's common to burn an entire week just settling the topic. A topic that's too big yields no depth; one that's too small has nothing to investigate. A good project starts not from a good topic but from a good question. AI can be a sparring partner that provokes student thinking in the work of narrowing that vague topic into an investigable question. But guard from the outset against letting AI settle the question too, because then the owner of the project shifts from the student to the AI.

Steps for Narrowing a Big Topic into an Inquiry Question

Rather than leaving topic selection entirely to student autonomy, have them narrow it down through the following steps. At each step AI's role is to ask back, not to answer.

1. Spread out the areas of interest: From the big lump of "the environment," have students write the sub-areas that pull at them (trash, fine dust, water).
2. Get questions from AI: For the area they chose, ask AI for "five questions a middle school student could investigate on this topic firsthand." Here, don't let them use AI's answers as they are; have them pick only the one they like.
3. Check whether it can be investigated: With the chosen question, have them ask themselves, "Can I gather data myself? Is the conclusion not already fixed to one answer?"
4. Sharpen the question: Revise it so that comparison and testing enter in, moving from "why is it like that?" toward "which of these works better?"

If an inquiry question is answered by a single search, that's not a project; it's homework. Inquiry begins only when the answer doesn't come easily.

Safeguards That Use AI While Protecting Student Thinking

Bring AI in and the risk of students outsourcing their thinking always follows. Use these safeguards to keep the balance.

• AI output is material; the decision is the student's: The act of choosing from the list of questions AI produced, and the act of writing one line on why, must always be done by the student.
• Leave a trace of use: Have students record "what I asked AI and what I changed" in a single field. That record is itself metacognitive training.
• A teacher check gate: Before a topic is finalized, the teacher runs a 5-minute conference to check, "can this question hold up for four weeks?"
• Prevent overlap between groups: When questions come out similar, suggest a different angle to one side to keep some variety.

With these safeguards in place, AI works not as a substitute for a student's thinking but as a foothold that widens its range. If a vague interest in "reducing waste" narrows, through a back-and-forth with AI, into a testable question like "to cut leftover food from our school lunches, which works better, an advance sign-up system or smaller portions?", that narrowing process itself becomes the fuel for four weeks.

Key Takeaways

The first button on a PBL project isn't the topic but the inquiry question, and that question doesn't emerge from a vague interest in one go. Let AI provoke student thinking at the stage of spreading out areas and throwing out candidate questions, but leave the choosing, sharpening, and deciding in the student's hands. On a first project, allocate a full period to topic selection alone and walk slowly through the four steps above. A solid opening question keeps the following four weeks steady.