English | 日本語
LAB OS — PART II DESIGN | APPENDIX 03-A

Question Archetypes

Question Archetypes — Seven archetypes, and what must be decided for each
Even across fields, the structure of questions is broadly shared. Check which "archetype" your research belongs to, and whether the elements that archetype requires have been decided.

What "must be decided" means

The "things that must be decided for the question to stand" listed for each archetype are not items you write in the paper. They are elements you decide before turning the question into one sentence. For estimation, if "what to estimate" and "from what" are not decided, the question cannot be written.

Once decided, two things follow automatically. One is the content of the Core Question (Chapter 03, Step 2-3). The other is the data you need — what to observe, under which conditions (Chapter 03, "Once the question is decided, the required data are decided"). Start experiments before the elements are decided, and you find out later that the data you took cannot answer the question.

A|Estimation|Inference / Estimation

"Can we infer X from Y?"

A question that estimates structures or parameters not directly visible, from observed data.

What must be decided for the question to stand: what is X (the thing to estimate) / what is Y (the observable) / is it identifiable / how is uncertainty handled

B|Mechanism|Mechanism / Causality

"What controls X?"

A question that reveals the causal structure governing a phenomenon.

What must be decided for the question to stand: how "controls" is defined / how causation is distinguished from correlation / what the alternative explanations are

C|Validity|Validity / Boundary

"Under what conditions is the model/assumption valid?"

A question about the range of validity of a model or assumption.

What must be decided for the question to stand: what the assumptions are / under what conditions they break down / compared with what you call it valid

D|Uncertainty|Uncertainty / Sensitivity

"How sensitive is the prediction to uncertainty?"

A question about how much a prediction or decision is affected by uncertainty.

What must be decided for the question to stand: which uncertainty is treated (structural / parameter / data) / how it relates to decision-making / whether it ends as a mere variance evaluation

E|Scale|Scale-linking / Generalization

"How does micro affect macro?"

How small-scale processes affect large-scale behavior.

What must be decided for the question to stand: which scales are connected / what is preserved and what is lost / whether it generalizes

F|Dynamics|Dynamics / Evolution

"How does X evolve over time?"

A question about change over time or phase transitions.

What must be decided for the question to stand: what the state variables are / whether there are stability and critical points / dependence on initial conditions

G|Design|Design / Intervention

"How should we design a system/process to achieve X?"

A question that steps into design on the basis of structural understanding.

What must be decided for the question to stand: what the goal state is / what the design principles are / where the boundary lies between operation and manipulative steering

Entry and archetype

The entry — hypothesis-first or data-first (Chapter 02) — does not decide which archetype you choose. What it decides is how you answer that archetype's question. For estimation: set a forward model and invert it (hypothesis-first), or learn a mapping from observations to the estimate from data (data-first). For mechanism: set the mechanism as a hypothesis and test it, or search the data for the controlling factors. Under either entry the question is the same, but the content of "what must be decided" — especially the handling of identifiability and uncertainty — and the form of the data you need change. The exceptions are validity and uncertainty: both presuppose the existence of a model, so hypothesis-first is natural.

The entry is decided at the stage where you choose the archetype and fill in "what must be decided." If you decide the entry first, from the data or models you happen to have, the question shrinks to the range that tool can answer.

How to use

  1. Try writing your question in one sentence
  2. Choose the archetype above that it is closest to
  3. Check whether that archetype's "things to decide" are decided
  4. If any element is undecided, decide it, then rewrite the sentence

If you cannot decide the archetype, the question is still vague. If you cannot decide the elements, go back to Step 1 (survey the field).

Examples by field

Examples for understanding. Think in terms of archetypes A–G above.

What is decided once the archetype is decided

What to write / which journal to submit to / which data are needed. Being conscious of archetypes is not about following a trend; it is about making the structure of your own question clear.

Finding and managing literature (the practice of Step 1)

What to collect: the major review papers (the field's beliefs), five or more representative papers of the last 3–5 years (question archetypes and implicit premises), 3–10 similar papers from candidate journals (patterns of writing)

How to search: Consensus (search with a question sentence; try it first) / Google Scholar (citation counts; operators such as Geothermal AND "fluid transport") / Web of Science and Scopus (follow citation links) / arXiv (preprints). For positioning, following the References and Cited-by of papers you found useful works best.

Put it into the Comparison Table: prior study / their claim / hidden assumption / limitation / our update.

Managing: Zotero (free, browser extension, good for sharing) or Mendeley (PDF organization, Word/LaTeX integration). Leave one line per paper: "how it relates to my question."

Keep the target journal in mind: understand the journal's scope and narrow candidates with a journal finder. For experiences such as review times, see SciRev.