The Question Behind the Question

P60.answer-architecture.01 · Audience: guest, it-ml, language-pro

Capstone: threat-model an AI application STAR — and When It Creaks

LLM grading (this page):Real gradingOff — no LLM calls

"Tell me about a time you disagreed with a colleague." A candidate — a school teacher interviewing for a coordinator role — answers with a vivid five-minute account of an argument over the exam calendar: who said what, how tense the staff room got, how it was eventually settled by the head. Every word true, well told, and useless. The interviewer was not collecting anecdotes about calendars; they were probing whether this person can hold a professional disagreement without damaging the relationship — and the story, told as a story, never showed that. This module is about hearing the question behind the question, so the answer proves the thing actually being tested.

Step 1 / 5 — The anecdote is not the answer

A behavioural question looks like a request for a story, so most people answer it with one: setting, characters, plot, resolution. But the interviewer is not a story collector. They ask "tell me about a time…" because past behaviour is the best available evidence of a capability — can you handle conflict, own a mistake, deliver under pressure — and a specific incident is harder to fake than a claim like "I'm a team player". The story is the evidence format; the capability is the question. An answer succeeds when the listener can tick the capability box, not when the plot resolves.

ⓘ Concept: A behavioural question requests evidence, not entertainment

behavioural question = claim under test + request for one specific incident as evidence

Why it matters — Once you see the question as an evidence request, the standard for a good answer changes completely. A vivid, true, well-paced story can still fail, because vividness is not the test — the test is whether your telling demonstrates the capability the question probes. Conversely, a modest, low-drama incident can be a superb answer if it shows you doing the thing: naming the disagreement, making the call, absorbing the feedback. This reframe is the single most useful idea in the discipline, because every later skill — structure, calibration, story choice — exists to serve it. You are not performing your past; you are entering evidence for a claim about what you can do.

🗣️ From a linguist's perspective: The exam question behind the essay prompt

A language teacher grading an essay on "describe your favourite city" is not assessing the city. The prompt is a vehicle; the real test is range of vocabulary, control of tense, paragraph structure. A student who writes a heartfelt but grammatically flat piece has answered the prompt and failed the question. Behavioural interviews work the same way: the incident is the prompt's vehicle, and the capability is what gets graded.

✅ Key takeaways

A behavioural question is an evidence request, not a story request. 'Tell me about a time X' asks you to prove a capability with one specific incident; the plot is the vehicle, the capability is what gets graded.
Decode before you answer: strip the scenario, ask what a good hire would do in situations like this, and keep the verb — push back, own, prioritize, translate. That verb is the real question, and the list of verbs is short and stable across fields.
Answer the real question explicitly. Choose the incident that most clearly shows the verb (not the most dramatic memory), name the capability out loud, and let the story serve as evidence rather than hoping the listener infers the point.
No cinematic story is required. Small, true incidents with specific decisions are stronger evidence than grand vague ones — and a low-stakes story is always a legitimate choice. Never invent: fabrication collapses under follow-up questions.
The half-second habit: before answering, ask 'what is this really testing?' Structure and calibration (the next modules) organize an answer that aims at the right target; nothing rescues one that aims at the wrong target.

Next: P60.answer-architecture.02 — STAR — and When It Creaks

Whatever your field — teaching, interpreting, engineering, care work — the questions behind the questions are the same short list, and your own working life already contains the evidence. Step back any time to the pillar overview to see how this track pairs with the story-bank track, where that evidence gets mined and shaped.

Ask the mentor about this module

Ask a question about this content. The mentor explains and grounds its answer in what you are studying; asking is recorded as a learning signal, not a grade.

Images, PDF or text. Kept on this device only.

Keeping your files on this device

Keep a copy of the files I attach, in this browser

Off by default. The mentor always gets your file; this only decides whether your own copy stays here. Copies live in this browser only - they do not follow you to another device, and clearing site data removes them.

Ctrl/Cmd + Enter to send

🎓 Practice ladder

2 graded rungs · ~20 min

Two mentor-graded craft rungs. Rung 1 gives you three behavioural questions to decode into the capability each one tests. Rung 2 gives you a rambling answer — true but aimless — to rewrite so it answers the real question. You write prose; the mentor grades the decoding and the telling, never your history.

Rung 1 — Decode the real question (explorer)

Loading exercise…

Rung 2 — Rewrite to answer it (practitioner)

Loading exercise…

Try it yourself

A scratch console for this page's ideas — ungraded, nothing you run here is recorded.

Scratch console

A scratch console with the scientific stack (pandas, numpy, scikit-learn). Runs on the server — no network, resource-limited and measured.

Output appears here.

My notes on this module

Loading your notes...

Where next?

Continue

STAR — and When It Creaks →

Answer Architecture · P60

Later in Answer Architecture

Go up a level