Asking Better Questions

A better question is one whose answer is already half-formed in the other person’s head, and loose enough that they pick which part to hand over. Specific on the target, open on the response. That property, not cleverness, is how you ask questions people actually want to answer. Vague questions make people compose from nothing.

Key Takeaways:

• Two properties decide whether a question gets answered: how much of the answer is already stored, and how much freedom the other person has over which part of it to give
• Category questions ask for composition, instance questions ask for retrieval, and retrieval is roughly free while composition costs the answerer four separate operations under time pressure
• Stacking two questions in one breath deletes the first; people answer the last thing they heard, which is almost always the easier one
• The fear that a specific question intrudes runs ahead of the evidence: participants in Kardas, Kumar and Epley’s 2022 experiments forecast more discomfort from a substantial conversation than the conversation actually produced
• The question that lands hardest is usually the one that proves you kept a detail from three weeks ago, and it takes no imagination at all

The Test That Sorts Good Questions From Bad Ones

Ask yourself where the answer is going to come from. If the other person already has it sitting there, filed as an episode or a fact or an opinion they have held for a decade, the question is cheap and you will usually get more than you asked for. If they have to build the answer from scratch out of everything they know about themselves, the question is expensive, and expensive questions get the shortest legal reply.

Watch what happens with “what do you do for fun.” Nothing in a human being is stored under that heading. The person has to search across every hobby, side project and Saturday habit they own, rank those by how presentable they are, decide which version of themselves they want in the room, and compress the winner into one sentence. That is four operations in two seconds, performed for somebody whose opinion of them is not yet settled. “Not much, really” is the correct answer to a question that hard. It is the same reason “how was your day” kills conversations: the day is a category, not a stored episode.

Now ask what they ended up doing last Sunday. One day, already indexed, no ranking required. Even a dull Sunday comes back with texture attached, because the texture was stored with the memory instead of being assembled on demand.

The second half is where most advice stops early. Once you have pointed at something retrievable, leave the shape of the answer alone. “What did you do Sunday” lets them hand you the hike, the fight in the parking lot, four hours face down on the couch, or a shrug that turns into a story two sentences later, and which one they pick tells you more than the content does. “Did you get outside on Sunday” points at the same day and takes the choice away, so you learn the weather report instead of the person.

Locate the answer before you open your mouth. If it is not already somewhere in their head, you are asking a stranger to do unpaid work, and unpaid work at exchange three gets declined politely.

How to Ask Better Questions Under Real Conditions

Four moves cover most of it, and not one of them requires you to be interesting.

Anchor in time rather than in category. Not what music someone likes, but what was playing in the car this morning. Not whether they travel, but the last trip that was worth what it cost. A date points at a file. A category asks for a summary of a life, delivered on request, to someone who might not deserve it yet.

Go for the last instance instead of the average. Nobody keeps an average. Ask a nurse what her shifts are like and you get a shrug about how it depends; ask what happened on the last one and a specific person walks into the story, usually with a name.

Ask about the exception rather than the norm. “What does everyone get wrong about that job” is the whole move. The exception is the part that is filed sharply, with the irritation still attached, which is why people answer those questions at length and then apologize for going on. The most rewarding deep questions to ask people work on this exact principle.

Then send one question at a time. Two questions in one breath is the most common way a good first question gets destroyed, and it happens because the pause after a real question feels enormous to the person who asked it. It is about two seconds. Feels like ten. Count to three, and let them be the one who fills it.

Of the four, anchoring in time is the one to test first. The difference in reply length shows up on the very first try, which leaves you nothing to talk yourself out of.

Four Ways a Question Kills Its Own Answer

Most bad questions are not vague at all. They are over-specified in exactly the wrong place, which closes the answer down while pretending to open it.

The stack is the first one. “How was the move? Did you find a place near work, or are you still looking?” The move question is gone. Nobody will return to it, and you traded a story for a logistics update.

Second is the question with the answer already inside it. “That must have been exhausting, right?” Now there are two available responses: agree, or correct someone who was trying to be kind. Most people agree. You have just written their line and handed it to them, and you will never know what they would have said. The same trap wrecks asking for advice and actually getting honesty: load the question and you get agreement, not the truth.

Third is the verdict request. “Was it good?” asks for a rating, and ratings terminate. Once a number is on the table there is nowhere to go except a second number. Compare it with “what was the part you did not expect,” which asks for the same experience without demanding a score first.

Fourth, the audit. “So what’s next for you?” and “when are you going to start looking?” read as evaluation no matter how warmly they are delivered, because they ask the person to justify their current position. The answer comes back defensive and short, and the conversation goes cold two exchanges later without either of you knowing why.

Reread your last three questions to someone you care about. If any of them contained a second question, a built-in adjective, or a request for a score, that is the one that flattened the reply.

Why the Interesting Question Underperforms

The question people remember is almost never the clever one. It is the one that proves you kept a detail. “How’s the ankle” from someone who heard about the ankle three weeks ago beats anything ever printed on a list of conversation prompts, and it requires no imagination whatsoever. That is a follow-up question doing what follow-ups do, proving you were listening.

Retrievability, plus evidence that you were paying attention. That is the entire mechanism, and it is so unglamorous that people keep hunting for better material instead of using it.

Where This Stops Working

Specific questions stop working under acute stress, and that boundary is worth respecting. Someone who got bad news that morning cannot produce the Sunday story, because retrieval flattens when a person is holding something heavy, and a precise question lands as pressure. Say something true instead and leave a gap. “That sounds like a lot” does more at that moment than any question in this article.

Question lists are the other failure. Aron and colleagues built a 45-minute structured procedure in 1997, with two people who had agreed to the exercise and a sequence that escalated disclosure step by step. Lift those questions out of that frame, drop them into a dinner nobody signed up for, and they read as an audition. The frame was carrying more of that procedure than the questions were. Inside the right frame, though, deep questions to ask your best friend do exactly what Aron’s list did.

Power changes the object too. “What did you do this weekend” from a manager is not the same question as the identical sentence from a neighbor, because one of them can affect your income and both of you know it. The answer will be presentable rather than true, and that is a reasonable thing for the other person to do.

The usual worry about specific questions is that they intrude. That worry has been measured, and it runs high: Kardas, Kumar and Epley put participants into more substantial conversations in 2022 and found their forecast of discomfort came in above what they reported afterward. One limit travels with the finding. Every one of those conversations was agreed to in advance, which is precisely the condition a dinner party does not hand you.

FAQ

What are good questions to ask someone you just met? Ask about something in the room or something in the last 24 hours, because both are already loaded and neither is personal. “How did you get roped into this” works at exchange two. Hold back anything that asks for a summary of their personality until they have volunteered some detail you never requested, which is the signal that they have decided you are worth the effort.

Why do my questions get one-word answers? Usually because the question asked for a summary or a score rather than an episode. “How’s work” and “was it good” both have correct one-word answers, so a one-word answer is not rudeness, it is accuracy. Swap in a question about a specific day or a specific surprise and the reply length changes immediately.

Is it rude to ask someone a personal question? It depends on whether the question is specific about the topic and loose about the answer. “Are you seeing anyone” demands a status; “what has your year been like” points at the same territory and lets them decide how far in to go. The rude version is the one that leaves no graceful way to give a small answer.

How do I think of questions in the moment? Take the last concrete noun the other person said and ask about that, or use one of two standing sources: the room you are both in, and the previous day. Nobody generates good questions out of nothing while also monitoring how they are coming across. The nouns are free, and they are already lying on the table.

Most advice about questions is written as if the goal were to be interesting. The goal is to be easy to answer, which is a different and much lower bar, and the two are frequently at odds. A person handed a question they can answer without effort will talk for four minutes and walk away thinking the conversation went well, and they will not remember what you asked. That is the correct outcome. The question is a door, not an exhibit, and nobody admires a door. One way to learn which questions open a door is to keep asking them where the answers keep coming. Lona gives you that: an AI you talk to daily, it retains what you asked it last week, and it lets you feel the difference between a stored answer and a composed one. It is where you practice for the real doors. It is not one of them.

Sources

• Aron, A. et al., “The Experimental Generation of Interpersonal Closeness,” 1997
• Kardas, M., Kumar, A. & Epley, N., “Overly Shallow?: Miscalibrated Expectations Create a Barrier to Deeper Conversation,” JPSP, 2022