What changes when a child asks why a machine stopped? An answer may lead her to look at a part she had not noticed and ask another question. That conversation lets us follow how a doubt becomes a search.
An answer can change what we ask
If inspecting the machine reveals a loose cable, the question changes from ‘Why did it stop?’ to ‘What made the cable come loose?’ The answer has not merely filled a gap; it has changed what is worth looking for next. Tuval Raz and Yoed N. Kenett review research on this everyday ability and ask what happens between noticing something missing and choosing a question that might resolve it. They focus mainly on questions used to seek knowledge, while recognizing that people also ask to build relationships, request help or manage a conversation.
The questioner needs some starting knowledge. To ask how a machine works, one must know enough to recognize that an explanation is missing and to expect something more useful than a label. Someone who has never seen its parts may ask a broad question; after learning about the cable, a narrower one becomes possible. This apparent paradox—needing knowledge to ask about what one does not know—is central to the review.
The authors do not claim that more questions are automatically better. A question has a purpose, a cost and a context. It can narrow uncertainty, expose a contradiction or help plan an action. Asking about every unknown detail would overwhelm any conversation. The important issue is how people select the gap worth investigating now.
Some questions eliminate more possibilities
Older theories examined how people choose, form and voice a question. Newer accounts ask how much information an answer may provide. Raz and Kenett bring those approaches together. Imagine trying to identify an object hidden in a box. ‘Is it a carrot?’ tests one possibility. ‘Is it a vegetable?’ can rule out a much larger set. The second question may be more informative, depending on what is already known. This does not mean every question should be broad: a broad answer can be unhelpful when the task calls for a precise next step.
Information theory provides a useful language for this problem: a question can reduce uncertainty among several possible explanations. But the most informative question on paper may be difficult to invent, tiring to evaluate or socially awkward to ask. The review discusses a trade-off between speed and quality of inquiry. People do not always have the time and resources to calculate which question would eliminate the most possibilities, and idealized models of experiment design do not describe every real conversation.
Question quality therefore depends on the goal. If a bus passenger only needs to know whether this is the right stop, a simple yes-or-no question is enough. If an engineer needs to understand repeated failures, a question about the conditions under which a component fails may open a more productive investigation. Calling one universally ‘better’ would ignore what the person is trying to learn.
The authors' model is a loop, not a checklist
Raz and Kenett propose a framework called QuInS, short for Question Asking in Information Seeking. It begins when a surprising event, inconsistency or missing piece exposes a gap in understanding. Curiosity can make that gap worth pursuing. The person uses what they already know to form a question, decides whether and to whom it can be asked, evaluates the answer and, if necessary, asks again or investigates in another way. The key idea is that an answer changes the next question. Inquiry is an unfolding activity rather than a single sentence ending in a question mark.
The six proposed stages should not be mistaken for a rigid route. After hearing an incomplete answer, someone may go straight back to formulation and evaluation without repeating every earlier step. The model also includes a social decision that abstract accounts of information often miss: perhaps the person asked does not know, is busy, has a reason to withhold an answer or could be embarrassed. A question's usefulness depends partly on the person and setting in which it is voiced.
QuInS is a synthesis of earlier theories and studies, not a finding that every mind passes through six neatly observed stages. The review's value is to make the moving parts visible so researchers can test them separately: noticing a gap, creating a question, expressing it, judging the reply and deciding what to do next.
Curiosity starts the search; complexity changes it
Curiosity helps turn a missing piece of knowledge into an active search. In one task reviewed by the authors, participants read statements containing deliberate gaps, wrote their own questions and could then choose which answers to uncover. People who generated more high-quality questions reported greater curiosity, explored more information and remembered more of what they learned. The result links question generation with learning, but it does not show that any instruction to ‘be more curious’ will produce the same effect in every setting.
The review distinguishes the complexity of a question from simply having many questions. Bloom's familiar educational taxonomy runs from remembering a fact through understanding, applying, analyzing and evaluating to creating something new. In two studies using an alternative-questions task, more complex questions about ordinary objects were also judged more creative. That is a relationship in particular tasks, not proof that every elaborate question is useful. Bloom's categories themselves have been criticized for blurring distinctions and for treating learning as too rigid a ladder.
A question can be original yet irrelevant, or simple yet exactly what the situation needs. What matters in creative work is often whether it opens a useful route beyond the obvious: ‘What else could this object do?’ may change the search more than another request for its name. Complexity becomes interesting when it helps reorganize a problem, not when it merely sounds impressive.
Learning changes which questions become possible
In classrooms, students often receive questions already written by someone else. The review discusses another approach: give them a broad prompt, such as ‘What are the strengths and weaknesses of this explanation?’, then ask them to write specific questions about what they have just studied. Classroom work described by the authors found that guided student-generated questions encouraged fuller explanations and better understanding. The structure matters because a blank demand to ‘ask something’ may leave a novice with no path, while a ready-made answer leaves no room to search.
Knowledge and questioning can reinforce each other. In a study following university students through a semester, their ability to ask questions in the course's subject improved as they learned more about it, even though their general question-asking scores did not show the same rise. More domain knowledge did not close inquiry; it gave students better material for seeing gaps. That is why an expert can ask a precise question a beginner has not yet learned to imagine.
The same study found a revealing tension in assessment. Original, complex questions predicted stronger performance in an open research project but were negatively associated with multiple-choice exam scores. The authors do not conclude that testing is useless or that questioning always raises grades. They show that different forms of assessment reward different kinds of thinking. If we claim to value inquiry, we should notice what our tests actually invite.
Before solving a problem, decide what the problem is
Many accounts of problem solving begin after a problem has been defined. Raz and Kenett draw attention to the earlier work of finding and shaping it. A vague complaint—‘the machine keeps failing’—can lead to a rush of solutions. Questions about when it fails, what changed just before the failure and which parts remain unaffected can turn the complaint into something testable. The question does not merely collect a missing fact; it helps construct the problem that the later solution will address.
The review relates question complexity and originality to performance in open-ended problem-solving tasks, while noting that closed puzzles can demand something different: narrowing possibilities toward one correct answer. A study comparing people and GPT-4 found that more original and complex questions accompanied stronger solutions on open tasks in both groups, but not the same pattern on closed tasks. The work is preliminary, and comparing model outputs with human creativity requires care. Its useful point is that the form of the problem changes what a good question can do.
A question can also reveal an assumption hidden in the first solution. If a team asks ‘How do we make people read this notice?’, it has already decided the notice is the remedy. ‘What information are people missing when they choose?’ may lead to a different design. This is an editorial illustration of problem framing, not a result from the experiment.
Children learn to ask within a social world
Children do not begin as perfect information seekers. The review describes a developmental shift from testing one possible answer at a time toward questions that can eliminate several alternatives. In studies of children aged four to six, older children asked more of these constraint-seeking questions, and their use helped predict success in solving the task. Development involves recognizing that information is missing, finding someone who might know, deciding how to ask and checking the reply. It is not simply a rise in the number of question marks spoken each day.
The social part develops too. In a study using stories about a child with a question, younger children were more willing than older children and adults to endorse asking even when an adult was busy. Learning when a question is welcome is not the same as losing curiosity. It shows that inquiry must negotiate attention, roles and norms. Sometimes the available expert is not approachable; sometimes a child has to wait, watch or ask someone else.
The evidence across ages and settings is uneven, so the review does not offer one complete developmental timetable. It does, however, resist a common mistake: treating questions as a private skill detached from the people who can answer them. From the beginning, learning to ask includes learning how another person's knowledge and availability affect the search.
Good questions also depend on what happens between people
An adult may avoid asking a sensitive question for fear of embarrassing someone. The review cites experiments in which the people receiving such questions reacted more positively or neutrally than the askers expected. But that is no permission slip to ask anything at any time. Other work shows that withholding a question can protect privacy and trust. Asking for help also changes who controls the exchange; help offered before it is requested can threaten a recipient's sense of competence, especially when it comes from someone with more status.
New AI tools create another setting for inquiry. A model can propose questions, score their complexity or help someone explore a problem, but rapid, plausible answers can also narrow a group's ideas into similar forms. The review reports early studies relating question complexity to open-ended problem solving in humans and models, while recognizing that the comparison remains preliminary. The important human skill is not merely writing a clever prompt; it is judging whether the response really reduced uncertainty and what should be asked next.
The article ends with open questions about how this skill develops, differs among people and might be trained. Its enduring insight is more modest and more useful than ‘ask better questions’: inquiry is a loop in which existing knowledge makes a question possible, an answer changes that knowledge, and the new picture lets us ask something we could not ask before.
What to keep
- A question makes a knowledge gap actionable.
- Questions help direct learning and problem solving.
- Their role changes across developmental and social contexts.
The nuance
What this story does not let us claim
Asking more questions is not automatically better; their usefulness depends on goals, context and the information available.
Go deeper: concepts and optional reading practice
Reversible cards
Test the concept
Topic assessment
Check whether you can explain the idea, not merely recognize it.
Every answer is grounded in the explanation above. Some questions introduce new cases to test whether you can apply and analyze the ideas, not merely recall them.
Source & scope
The Psychology of Question Asking: A Comprehensive Review. Tuval Raz & Yoed N. Kenett. Annual Review of Psychology, vol. 78 (2027). First published as a Review in Advance on August 7, 2026.
This editorial guide is an independent explanation. Consult the original publication for the studies, references and full context.
Open the original review ↗