A question is a request for missing information, explanation, judgement, prediction, choice or action.
“What is this?”, “Why did it happen?”, “What should I do?”, “How likely is it?”, “Which option is better?”, “What caused the error?” and “What would happen if…?” are not the same kind of question. They demand different evidence, reasoning and answer forms.
Quick answer: how should questions be categorised?
- Purpose: factual, explanatory, diagnostic, predictive, evaluative, creative, procedural?
- Scope: narrow, bounded, broad, systemic?
- Answer type: fact, list, explanation, estimate, recommendation, decision?
- Evidence: what sources or observations are required?
- Difficulty: recall, synthesis, inference, modelling, judgement?
- Openness: one answer or many defensible answers?
- Uncertainty: known, estimated, disputed, unknowable?
- Audience: who needs the answer and at what level?
- Domain: what body of knowledge governs?
- Actionability: does the answer lead to a next step?
This article applies the general framework from How to Categorise Anything to questions as structured information needs.
1. Start with what is missing
A question exists because some information, relationship, explanation or decision is unresolved.
2. Factual questions ask what is true
They usually seek names, dates, quantities, definitions or directly verifiable states.
3. Explanatory questions ask why or how
They require mechanisms, causes, relationships or models rather than isolated facts.
4. Diagnostic questions ask what is wrong
They compare symptoms or observations against possible causes.
5. Predictive questions ask what may happen
They require models, assumptions, uncertainty and time horizon.
6. Procedural questions ask how to act
They seek steps, sequence, tools, prerequisites and checkpoints.
7. Evaluative questions ask what is better
They require explicit criteria, trade-offs and context.
8. Decision questions ask what should be chosen
They combine alternatives, constraints, evidence, stakes and preferences.
9. Creative questions ask what could be made
They allow multiple valid outputs and are judged by fit, originality, coherence and constraints.
10. Counterfactual questions ask what if
They require a model of how changing one condition affects the rest of the system.
11. Narrow questions reduce search space
A bounded question can often be answered directly from a small evidence set.
12. Broad questions require decomposition
“How does civilisation work?” contains many subquestions about institutions, resources, culture, technology and time.
13. Scope should be explicit
Time period, geography, population, domain and assumptions define the question boundary.
14. Ambiguous questions need interpretation
One sentence may contain several possible meanings. Resolve by context or surface the main interpretations.
15. Closed questions constrain answer form
Yes/no, multiple choice and exact-value questions have narrow valid outputs.
16. Open questions permit multiple defensible answers
They still need evidence and criteria; openness does not mean absence of standards.
17. Answer type should match question type
A recommendation is not a substitute for a factual answer, and a fact list is not enough for a causal explanation.
18. Evidence demand varies
Some questions need one authoritative source; others need multiple independent sources or direct measurement.
19. Historical questions need source context
Primary, secondary and retrospective evidence may answer different parts of the same question.
20. Scientific questions need testability
Questions about mechanism or effect should identify what observations could support or challenge the claim.
21. Normative questions require values
“What should society do?” cannot be answered from facts alone; values and trade-offs must be named.
22. Difficulty is not the same as breadth
A narrow mathematical proof can be difficult; a broad descriptive question can be easy.
23. Recall questions test retrieval
They ask for known information with little transformation.
24. Synthesis questions combine sources
They require integrating evidence across documents, observations or domains.
25. Inference questions require reasoning beyond explicit text
The answer follows from evidence but may not be stated directly.
26. Modelling questions require representation
They ask for simplified systems capable of explaining or predicting behaviour.
27. Judgement questions require criteria
Quality depends on how explicitly alternatives are compared.
28. Uncertainty belongs in the question record
Some questions have known answers, some estimates, some disputed answers and some no currently knowable answer.
29. Unknown is not unanswerable forever
New evidence or methods can move a question from open to answerable.
30. Audience changes explanation depth
The same question may need a Primary 4 answer, university answer and specialist answer at different resolutions.
31. Domain determines vocabulary and standards
Medicine, law, science, history and mathematics evaluate evidence differently.
32. Cross-domain questions need decomposition
A policy question may contain scientific, economic, ethical and legal subquestions.
33. Actionable questions lead to decisions or tasks
They are often better when the requested next step, constraints and stakes are explicit.
34. Diagnostic follow-up questions reduce uncertainty
A good next question is one that most efficiently distinguishes among plausible explanations.
35. AI can classify question intent
Models can route questions toward search, calculation, explanation, recommendation or specialist review.
36. Intent classification can be wrong
Preserve ambiguity when a user may be asking several things at once.
37. A practical question record
- question ID;
- purpose;
- scope;
- answer type;
- domain;
- audience;
- evidence requirement;
- reasoning demand;
- uncertainty;
- constraints;
- actionability;
- related questions;
- status;
- taxonomy version.
38. Question classification should improve routing
If question type does not change which evidence, expert, tool or reasoning process is used, the category may be too weak.
39. Better questions reduce wasted search
Clarifying scope and answer type often solves half the information problem before research begins.
40. The deeper idea
A question is a map of what is missing. Categorising the question reveals what kind of bridge must be built to reach an answer.
The form of the question determines the form of evidence and reasoning needed to answer it well.
Final answer
Categorise questions by purpose, scope, answer type, evidence demand, reasoning difficulty, openness, uncertainty, audience, domain and actionability. Separate factual from explanatory, diagnostic, predictive, evaluative and decision questions, and route each toward the evidence and reasoning it actually needs.