Let AI analogize with something I already know
It doesn't know what you already know, so it pulls a source domain from the public question bank. Leave that slot empty in the question, fill it with something you already get — then the analogy actually lands.
“Even if the same payment request is submitted twice, the system should only record it once.” You stare at that sentence for two minutes. You know every word. How the system tells those two are the same payment — you can't picture it. It's 8:30 p.m. You're reading a note on stopping double charges.
You paste the question to AI and ask for an analogy. It comes back in three sentences:
Think of it as PUT and POST in HTTP. Send PUT a few times, the result is usually the same; each POST may create a new record.
The paragraph is short and lines up neatly. You've seen PUT and POST, so you treat them as two ready-made labels and jot down “one is safe to repeat, one creates an extra row”. You think the original question is done.
Ten minutes later a colleague follows up: “If the user taps pay twice, what does the system use to tell it's the same payment?” You can't answer. You open three more tabs on PUT and POST. You were one question short. Now you have two extra words, and why that first payment isn't recorded twice is still hanging.
The analogy itself isn't wrong. You nodded because two acronyms sat in a practiced contrast and looked like an answer. For you they're still material that needs explaining. AI only saw your question. It never saw which experience you already know.
Ask it differently. Hand it a reference you already know. If you've done delivery, have it use the same dispatch slip sent out twice. If you've handled checks, have it use the same check handed to the counter twice.
- First say what you already know, so the analogy starts from a move you've seen.
- Then say which concept to explain, so AI only maps the part that belongs to this question.
- Finally ask for a line-by-line map, so you can check whether each step lined up.
An analogy can catch you only if the thing it borrows is already in your head. Below, swap the same concept across three references, and see which one you can check yourself.
What actually changes the result in the demo is that extra slot in the question. Copy the two sentences below as a block — just swap what's in the brackets.
- Name a reference you already know
I'm trying to get 【the same payment request submitted twice is only recorded once】. Please analogize with something I already know: 【the same dispatch slip the system sent out twice】.
- Then ask AI to map each piece
Please line both sides up, item by item: what the payment request maps to, what the number the system uses to recognize the same payment maps to, and what ignoring the second submit maps to. One sentence per item.
In the first sentence, “something I already know: 【a concrete thing】” is the key. Without that half-sentence, AI picks the reference for you. What it picks may be accurate — or something you have to learn again. Fill the brackets with something you've done yourself and seen the outcome of. Hearing the name is not enough.
Whether an analogy helps depends on whether you can check each step — not on whether it sounds smooth. When a dispatch slip goes out twice, the same slip number should stop the second send, and the rider only makes one trip. You can judge those three outcomes from experience, so you can also turn back and check whether the payment request, the request ID, and the duplicate handling line up one to one.
“Please analogize with something from everyday life” still hands the choice to AI. It might pick library loans, airport security, or a bank transfer — any of which might happen to be a blank for you. Write “please analogize with the same dispatch slip sent out twice,” and it knows which stretch of experience to borrow.
Also, analogy only carries structure. It doesn't promise every detail matches. The second sentence asks for a line-by-line map, so you can see which parts actually lined up. If one item can only be filled in by imagining, swap in a more familiar reference and ask again.
This is usage experience. For where the reference sits in a full question, see Let AI explain it for my situation.