You ask someone what they like, make a prediction, and miss spectacularly. Five minutes later, they explain their reasoning and suddenly the whole situation makes sense. How can someone be terrible at guessing other people's preferences yet quick to understand their problems?
Because predicting an answer before you have information and making sense of an answer after you hear it are different skills. Confusing the two makes social insight look like mind reading. It is usually something more ordinary—and more useful.
A preference is a clue, not the whole map
Imagine an invented conversation. A friend once said they love having free time. Months later they announce that their ideal vacation is a fully planned tour. Contradiction?
Instead of accusing them of inconsistency, ask: “You like freedom in everyday life, but you prefer a schedule when traveling. What's different?” They might explain that researching transport and hotels is exhausting; outsourcing those decisions makes the holiday feel freer.
That is a more accurate picture of their preferences than either statement alone. Their values didn't necessarily flip. The conditions changed.
Remembering earlier answers helps here. The first question gathers a fact. The later question checks how that fact fits with another. Conversation becomes a growing model of one person rather than a collection of isolated trivia.
Sometimes you have a theory. Sometimes you don't.
One person described asking questions even when they already had a plausible answer. Other times they genuinely had no idea. When someone tried to force this into an either-or choice, the reply was effectively: “Obviously both. If I already knew every answer, why would I be talking to people?”
The interesting moment comes when a hypothesis fails. Instead of defending it, you can say, “Oh! I thought it was for a completely different reason.” Now the unexpected answer becomes the reward.
A question asked to learn leaves room for surprise. A question asked only to prove your theory tends to close that room.
Understanding a problem is not predicting it
There is another important distinction. You might be unable to guess what somebody is worrying about. Their situation, priorities, and private experiences are genuinely unknown.
Yet when they tell you the story, you may quickly spot the friction: two expectations don't match; a decision criterion has shifted; a task has a hidden dependency; nobody agreed who should decide.
That rapid reconstruction is structural understanding, not proof that you predicted the person's feelings. You can check its accuracy by asking, “Is that the part that bothered you?” Their correction matters more than the elegance of your explanation.
The question that moves the entire problem
Consider a playful example about facial hair. Someone hopes a beard will suit them better when they get older. A friend asks: “Does getting older actually make a beard suit you better?”
The original speaker wanted hope. The question accidentally started an investigation. The reaction might be: “Can you not destroy the dream for five seconds?”
That little “wait, why are we assuming this?” also works outside conversation. If mistakes keep happening at work, the default question is often “Who needs more training?” A different question is “Can we redesign the steps so this mistake is harder to make?”
Neither question automatically wins. The second changes the level of analysis—from the individual to the system.
Then someone hands the same brain an AI content factory
Now imagine using that habit to automate content production. Instead of obsessing over how quickly an article can be drafted, you ask why a finished draft sits unpublished, why one failed check blocks unrelated work, or why publishing more pages doesn't automatically mean more readers.
An operator describes arranging AI-assisted work to run around the clock. The practical opportunity is that automated jobs can continue while a human thinks about the next bottleneck. But a twenty-four-hour schedule is not a twenty-four-hour stream of verified results. Drafts, approved work, published pages, and actual readership are different measurements.
And an endless supply of ideas can cause rework if the system changes faster than anyone can evaluate it. The trick is to preserve the speed while checking what actually improved.
Useful curiosity beats pretend telepathy
In a 2017 study, Huang and colleagues found that asking questions—especially follow-up questions—was linked to being liked more by conversation partners. A 2025 correction to the article left its main conclusions intact. Kardas and colleagues found that people often overestimate how awkward deeper conversations will feel. Neither finding turns an individual questioner into a certified genius.
The real takeaway is more fun: you can't predict every person's preferences, so you ask. When the reasons finally appear, you connect the pieces. When a new answer clashes with an old one, you get a new question.
People are wonderfully bad at fitting into a single template. Fortunately, that means there is always another interesting thing to ask.
Sources: Huang et al. (2017), https://doi.org/10.1037/pspi0000097 ; correction https://doi.org/10.1037/pspi0000491 ; Kardas et al. (2022), https://doi.org/10.1037/pspa0000281
