Interview Question Development

How do I ask "why" in an interview without getting a rationalised answer?

Route around it. Direct "why" questions return the most available, most socially acceptable explanation, produced on the spot by someone who has usually never been asked before, and it arrives in a form indistinguishable from insight. Use sequence instead ("what happened just before that?", "and then what?"), consequence ("what did that mean for you?"), and contrast anchored to a real event ("how was that different from the time before?", "what would have had to be different?"). If a direct why is used at all, use it last, on an episode already reconstructed, and treat the answer as a claim to weigh against the sequence rather than as the finding.

Why is "tell me about the last time" better than "what do you usually do"?

Three reasons. The participant is retrieving rather than composing, so the answer is a memory rather than an on-the-spot summary weighted to whatever is most recent and most vivid. The answer contains detail nobody thought to ask for, which is where unexpected findings come from. And an episode is an analysable unit, whereas a generalisation is the participant's own analysis, performed badly, with the working thrown away. The exception is when the participant's summary judgement is genuinely the object of study, such as their stated position on an issue or their sense of a relationship over time.

What questions should never be asked in a qualitative interview?

Four families. Anything asking a participant to predict their own future behaviour, because stated intention is systematically generous. Anything asking them to explain their own unconscious motivation, because that is an analyst's inference from evidence, not something a person has access to. Anything asking them to rank or rate things they have never compared, because it manufactures a preference at the moment of asking. And anything asking them to count a routine, low-salience behaviour over a long window, because the answer is rounded to salient numbers and anchored to whatever the question suggests.

What is laddering and when does it stop being evidence?

Laddering climbs from a stated attribute to the consequence it delivers and the value beneath it, one rung at a time, using the participant's own words at each step. It stops being evidence at the rung where the answers become abstract virtues that could apply to anyone: "I want value", "I want the best for my family". Everything above that point is the participant supplying the culturally expected top of the ladder. Two or three rungs is usually the honest limit, and the output is one person's articulated account of a preference, not proof of a hidden hierarchy.

What is critical incident technique?

A protocol for reaching what actually happens at a moment that matters. Define the incident class precisely, have the participant select a specific instance and anchor it in time and place, elicit the sequence in the order it happened without interrupting, walk it again for detail, elicit the outcome and its consequence, and only then ask what made the difference. Two or three incidents per interview is realistic; budget 8 to 12 minutes each. It produces behaviour-level detail no attitude question reaches, and its cost is time.

How do you ask about embarrassing or socially undesirable behaviour?

Design the permission into the question rather than retrofitting an assurance. Normalise first ("some people manage to keep on top of this and plenty of people do not"). Load the question toward the undesirable side, so the presupposition is easier to correct than to confess. Ask about the behaviour, not the identity, because people will describe missing three payments and will not accept being someone who misses payments. Where the direct route is closed, ask in the third person about what people they know say, and label the answer at analysis as evidence about the norm rather than about the individual. Never pair a confidentiality assurance with a question that makes the honest answer sound bad; the assurance signals that the topic is shameful.

How do you ask about something that happened months ago?

Anchor to an event rather than a period: "the last time" beats "in the past six months", and a landmark ("since you moved", "since the baby") beats a calendar date. Ask what happened before you ask what they thought, because an opinion given first gets retrofitted into the account. And never ask someone to explain their past self's reasoning as though they had access to it. Ask what they knew at the time, what options were in front of them and what they did, and let the analyst reconstruct the decision.

What is a grand tour question?

A question that asks the participant to walk you through a domain they know and you do not: "talk me through what happens on a normal Tuesday from when you get in". It works because it hands the expertise to the participant, sets the expected answer length at several minutes rather than a sentence, and surfaces vocabulary, sequence and actors no direct question would have thought to ask about. Anchor it in a time and a place, or the participant will not know where to start, and follow it with mini tours into the parts that matter.

How can I tell whether an interview question is any good before I use it?

Three tests. Write down the three answers you expect: if you can predict them, the question is closed, leading, or collecting culture rather than this person's life. Ask whether someone with nothing to say could still answer it fluently: if yes, it will not discriminate. Then cognitively test it on two or three people from the target population, asking them afterwards what they thought the question meant and how they arrived at their answer. That last test takes under an hour and catches ambiguous terms, unanswerable recall windows and imported vocabulary that no amount of desk redrafting will find.

Can AI write interview questions?

It can draft them quickly, and its characteristic failures are specific. Generated question sets are symmetric, with the same shapes in every section and nothing removed. They are frequently full of well-worded "why do you feel that way" questions, which read as good practice and are the core defect, because the problem is the mental act being demanded rather than the phrasing. They invent category vocabulary that was never supplied. And the most dangerous habit is generating example answers or "what participants are likely to say" alongside the questions, which travel into analysis documents and become indistinguishable from real evidence. Require a rewrite log naming the defect in every question changed, and never accept anticipated answers in a design document.