Build a Question Map Before You Search: A Better Research Starting Point

Direct answer

Before searching, place the decision you need to make at the center of a page and draw seven branches: definition, mechanism, comparison, context, evidence, uncertainty, and action. Turn each branch into answerable questions and mark which evidence type could answer each one. Search from the map, not from the topic label.

Why topics are weak research units

“Research remote work,” “learn about AI,” and “understand memory” are topics, not inquiries. They have no stopping condition and no rule for relevance. Almost any source can enter, so search ranking quietly decides what the project becomes.

A question map adds architecture before abundance. It does not freeze the inquiry; it makes early assumptions visible enough to revise. This is especially useful for independent researchers who lack a supervisor or review team to expose hidden scope.

Formal review methods begin with precise objectives, eligibility criteria, comparisons, outcomes, and study types because those choices govern searching and synthesis.^{cochrane-question, cochrane-search, acrl-inquiry} A personal research brief need not imitate a clinical review, but it inherits the same logic: an answer can only be evaluated against a question.

Put the decision at the center

Write an actual choice, not a theme:

• Weak: “AI and writing.”
• Better: “Should our editorial team use AI for first drafts?”
• Stronger: “For low-risk internal explainers, does an AI-assisted first-draft workflow reduce total production time without increasing factual corrections or eroding writers’ independent ability?”

The stronger question exposes task, population, comparator, outcomes, and risk. It also makes room for a legitimate “do not adopt” result.

The seven branches

| Branch | Ask | Output | |---|---|---| | Definition | What do the key terms include and exclude? | Operational definitions | | Mechanism | How might the effect occur? | Causal or process hypotheses | | Comparison | Compared with what realistic alternative? | Baseline and options | | Context | For whom, where, and under what conditions? | Scope boundaries | | Evidence | What observations would discriminate among claims? | Source and method plan | | Uncertainty | What is unknown, disputed, or changing? | Uncertainty register | | Action | What decision, threshold, or experiment follows? | Decision rule |

Add a red dot beside questions whose answers could reverse the decision. Those are the load-bearing branches. Research them before interesting but peripheral history.

Evidence snapshotHigh confidence

Evidence frameworks support defining the question and selection logic before searching, while information-literacy guidance treats research as iterative inquiry. A map operationalizes those principles for bounded desk research; its seven-branch design is an editorial framework, not an experimentally validated universal template.

Claim sources: ^{cochrane-question, cochrane-search, acrl-inquiry}

Separate question types

Different questions demand different sources:

• Definition: standards, statutes, taxonomies, foundational theory.
• Prevalence: representative datasets and surveys.
• Causality: randomized or credible quasi-experimental comparisons where feasible.
• Experience: qualitative studies, ethnography, interviews, and affected voices.
• Implementation: field studies, audits, cases, and operational records.
• Forecast: explicit models, assumptions, scenarios, and leading indicators.
• Value: ethical reasoning, stakeholder interests, and normative argument.

The ACRL framework’s “research as inquiry” frame matters here: answers generate new questions, and the scope of inquiry changes as understanding develops.^acrl-inquiry Revision is a feature, provided the history remains visible.

A worked map

Decision: “Should a university replace some language-practice sessions with an AI conversation tool?”

• Definition: What counts as practice, replacement, and success?
• Mechanism: Does the tool increase speaking time, feedback, motivation, or avoidance?
• Comparison: Existing class, tutor, peer exchange, or no practice?
• Context: Which language pairs, proficiency levels, access needs, and privacy regimes?
• Evidence: Delayed speaking performance, error patterns, attendance, learner experience, data flows.
• Uncertainty: Model changes, pronunciation feedback quality, novelty effects, teacher workload.
• Action: Pilot only if privacy is acceptable; proceed if speaking time rises without worse delayed performance or inequitable access.

Now the search contains multiple evidence jobs. A product page may document features, but it cannot establish learning effectiveness. A small trial may estimate one effect, but it cannot settle institutional governance.

Convert branches into a search sequence

Search in this order:

1. Vocabulary pass: definitions, synonyms, technical constructs.
2. Map pass: authoritative syntheses, major institutions, landmark debates.
3. Evidence pass: targeted searches for the red-dot questions.
4. Challenge pass: rival mechanisms, null findings, harms, and affected perspectives.
5. Decision pass: implementation constraints, current rules, cost, and local evidence.

Document queries and selection decisions proportionate to the stakes. Cochrane requires much more comprehensive, reproducible searching for formal reviews, including multiple databases and attention to missing studies.^{cochrane-search} Do not borrow the label “systematic” without doing that work.

Draw the seven-branch map

Take one live project:

1. Write the decision in one sentence.
2. Add the seven branch names without consulting sources.
3. Generate two to four questions per branch.
4. Tag each question with an evidence type.
5. Mark up to five red-dot questions.
6. Set exclusions and a time budget.
7. Search one red-dot question and revise the map.
8. Preserve version one so later certainty cannot erase your starting assumptions.

Store active maps inside Build a Personal Learning System. Use Critical Thinking: A Practical System for Claims and Evidence to audit conclusions and How to Make Decisions Under Uncertainty to set action thresholds.

Mapping errors that bias the search

• Putting a favored solution, rather than a decision, at the center.
• Asking only “does it work?” and omitting mechanism, context, and harm.
• Treating the current alternative as “nothing.”
• Defining success with a metric the intervention happens to improve.
• Adding branches indefinitely without ranking decision relevance.
• Hiding values—fairness, autonomy, privacy—as if they were technical side notes.
• Revising the map silently after seeing results.

What a question map cannot anticipate

Limits and counterevidence

The map reflects the mapper’s vocabulary, access, culture, and incentives. It can omit people affected by the decision, frame contested categories as neutral, or privilege measurable outcomes over important experience. High-stakes research should involve domain specialists, information professionals, and affected stakeholders, with a documented protocol and independent challenge.

A good search begins before the query box. It begins when a broad subject is transformed into a set of questions that evidence could genuinely change.

Named sources

Evidence and further reading

1. Cochrane Handbook Chapter 3 — Defining Criteria and Grouping Studiesofficial · accessed 2026-07-28
2. Cochrane Handbook Chapter 4 — Searching for and Selecting Studiesofficial · accessed 2026-07-28
3. Framework for Information Literacy for Higher Educationofficial · accessed 2026-07-28

Publication record

Published July 29, 2026. No substantive revision has been recorded. Evidence last verified July 28, 2026.