Before asking AI, write down the task in your own words, what you know, what remains unknown, what you attempted, and what the evidence showed.
This record protects independent reasoning. It also gives the AI less room to invent the problem, select hidden assumptions, or replace the student’s next step.
Start with the task and policy
Record the actual objective:
Task: Explain why my loop skips the last element.
Course boundary: AI may explain errors, but may not write assignment code.
Private material: Do not share the full assignment or student data.
Course rules differ. A tutoring interaction allowed in one course can be prohibited in another. If the boundary is unclear, stop and ask the instructor.
Do not upload unpublished assessment content, classmates’ work, credentials, or restricted institutional material merely to improve context.
Separate facts from assumptions
A fact in this record is supported by an observation or authoritative source:
Known:
- The list has five elements. Verified by printing its length.
- The loop runs four times. Verified with a counter.
- The course requires an index-based loop. Verified in the instructions.
An assumption is plausible but not yet verified:
Assumptions:
- I think the final index should equal the list length.
- I suspect the loop condition is wrong.
This separation matters because an AI response may accept a false premise and build a coherent answer around it.
Name one unknown
“I do not understand loops” is too broad for one useful response. Find the first specific uncertainty:
Unknown:
- I do not understand the relationship between a five-element list's length
and its final valid index.
If several unknowns exist, order them. Resolve the earliest dependency first. The later question may disappear.
Preserve the attempt
Record your reasoning, not only the final code:
Attempt:
1. Changed the condition from `<` to `<=`.
2. Predicted that the fifth element would be included.
3. Observed an out-of-range error.
4. Restored the original condition.
The failed attempt provides evidence. It also prevents the AI from recommending the same unchecked change as though it were new.
Include only the smallest permitted excerpt needed to explain the question. Use placeholders for private names and values.
State the allowed help
Tell the system what not to do:
Help boundary:
- Do not write or rewrite the loop.
- Ask one question that helps me connect length to valid indices.
- Wait for my response.
- Mark uncertainty and cite official language documentation for factual claims.
This boundary is a request, not an enforcement mechanism. The model may still give too much. Do not use prohibited output merely because it appeared.
OpenAI’s current Study Mode guidance recommends providing the learner’s level, topic, goal, and existing understanding when using that ChatGPT feature. The known–unknown–attempt method is broader: it is an editorial recommendation for preserving evidence, not a claim that one prompt formula guarantees a better response.
Use the template
## Before asking AI
Task in my words:
Known facts and evidence:
-
Attempt:
1.
Observed result:
First unknown:
Allowed help and verification:
A preflight gate comes before the five fields: check the course AI rule and remove private or restricted information. Stop if the use is not permitted or the information cannot be shared safely. Record policy details only when a special condition needs to be preserved.
A focused question might be:
Without writing code, ask me one question that helps me derive the final valid index from the list length.
Evaluate the response against the record
Do not replace your notes with the AI response. Compare:
- Did it use the stated facts?
- Did it treat an assumption as proven?
- Did it respect the help boundary?
- Did it cite a real primary source when needed?
- Does its explanation predict the observed result?
- Can you state the next step yourself?
Revise the attempt using evidence. Record which suggestion you accepted, rejected, or still need to test.
A non-code example
For a biology concept:
Task: Explain the direction of water movement in the example.
Known: The diagram labels side A as having a higher solute concentration.
Assumption: I think water moves toward lower solute concentration.
Unknown: Which gradient determines net water movement?
Attempt: Applied my assumption and got the opposite answer from the key.
Allowed help: Ask me to identify water concentration on each side. Do not state
the final direction.
Verification: Check the textbook definition and explain a new example.
The structure applies across subjects because it exposes the student’s model.
Common mistakes
- Writing conclusions as facts. Attach evidence.
- Hiding the failed attempt. It narrows the next question.
- Listing many unknowns without order. Identify the first dependency.
- Sharing the entire assignment. Use the smallest permitted context.
- Assuming the AI will obey the boundary. Ignore prohibited output.
- Accepting a response that contradicts observations. Test the discrepancy.
- Keeping no independent record. Preserve your reasoning outside the chat.
Do this now
Complete the template for one current question. Ask for one hint, diagnostic question, or explanation rather than a final answer. Verify the response with course material, a test, or instructor feedback.
Log what you learned
The completed before-AI record is the learning log. Add the response decision and observed verification after you evaluate the answer.
Next, apply the template to ask AI to explain an error without solving the assignment.