A student types a question into Google.
Ten tabs open.
Then twenty.
Then an AI system summarises them.
A document begins filling with quotations, links, notes and facts.
The student has found a great deal of information.
Have they researched?
Not necessarily.
This is one of the defining educational problems of the modern information environment.
Access has become cheap.
Search is fast.
Generative AI can produce an apparently coherent background brief before a student has worked out what question they are actually asking.
The bottleneck has moved.
Research is no longer mainly:
Can I find information?
It is increasingly:
Can I construct a trustworthy route from a meaningful question to an answer that remains accountable to evidence?
That route has several distinct jobs.
- What am I trying to find out?
- What do I already know?
- What am I merely assuming?
- Which evidence would actually discriminate among the possibilities?
- Which search terms describe the concept rather than merely repeat my wording?
- Who produced this source?
- What original evidence sits underneath the summary?
- Do several sources genuinely corroborate one another, or are they all descendants of the same claim?
- Where does evidence end and my inference begin?
- What remains uncertain?
And finally:
Could another person reopen my research route and see why I concluded what I did?
That is research.
Not link collection.
Not summarisation.
Not quotation harvesting.
Not asking AI to “do deep research” and then accepting the prose.
Research is a disciplined journey from uncertainty toward better-grounded understanding.
This article continues eduKateSengkang’s Top 10 … Skills Worth Learning series alongside Studying Skills, Questioning Skills, Verification Skills, Synthesis Skills and Critical Thinking.
The Wintour House question is deliberately durable:
If a learner became excellent at ten research operations, which ten would still matter when Google, databases, libraries and AI systems changed?
Before the Top 10: Research Begins With Uncertainty, Not Search
Suppose a student is investigating:
Does background music improve studying?
A weak research route begins:
background music studying benefits
The search engine returns supporting material.
The student concludes:
music helps.
A better researcher asks what the question contains.
- What counts as studying?
- Reading?
- Memorisation?
- Mathematics?
- Writing?
- What kind of music?
- With lyrics?
- Instrumental?
- Preferred music?
- At what volume?
- Compared with silence?
- For which learners?
- Which outcome?
- Immediate performance?
- Delayed memory?
- Attention?
- Enjoyment?
Already the question has changed.
Research has started before the first search.
The learner is reducing ambiguity.
That matters because search engines are excellent at finding text that matches words.
They cannot automatically decide which distinction the learner actually needed.
1. Learn to Turn a Topic Into a Researchable Question
“Artificial intelligence.”
Topic.
“AI and education.”
Still a topic.
“Does using generative AI while practising Mathematics improve students’ ability to solve related problems independently after the AI support is removed?”
Now we have a research question.
It identifies an intervention, a learning context, an outcome, and an important condition—independent performance after support disappears.
A researchable question does not have to be experimental.
- History: Why did interpretations of this event change after new archival evidence became available?
- Literature: How does the writer use changes in narrative viewpoint to alter what the reader knows?
- Science: Which variable best explains the difference between these two observations?
Research becomes far more efficient when the learner knows what an answer would need to explain.
A useful opening sentence is:
I am trying to determine whether, why, how, under what conditions, or to what extent…
That produces a job.
Worth learning because: a precise question gives every later search, source and note a criterion for relevance.
2. Learn to Separate What You Know, What You Assume and What You Need Evidence For
Before searching, create three mental columns.
KNOWN
ASSUMED
NEEDS EVIDENCE
Suppose the learner researches screen use.
Known: The device was used for two hours.
Assumed: Two hours is excessive.
Needs evidence: How duration, type of activity, age, context and outcomes relate.
This simple division prevents assumptions from entering research disguised as facts.
It also prevents wasting time researching things already securely established.
A good researcher is not a blank slate.
Prior knowledge helps formulate better questions.
The danger occurs when prior knowledge silently becomes evidence.
Ask:
- Which part of my current model would I still believe if I had to show its source?
- Which part is merely plausible?
- Which part is my interpretation?
Worth learning because: research becomes more honest when the learner knows which parts of the starting model are evidence-backed and which are merely provisional.
3. Learn to Search Concepts, Not Just Sentences
Students often search the sentence already inside their head.
- “Why homework is bad.”
- “Proof AI helps students.”
- “Reasons uniforms improve discipline.”
The search inherits the conclusion.
A stronger researcher translates the question into conceptual search terms.
Suppose the question concerns whether practice spread across several days improves retention.
Useful concepts may include:
- spaced practice,
- distributed practice,
- retention,
- delayed test,
- learning.
Not: “why studying every day is better.”
This matters because research communities may use vocabulary different from everyday language.
“Carelessness” might map onto attention lapse, execution error, response monitoring, or speed–accuracy trade-off.
“Remembering where a fact came from” may map onto source monitoring.
“Checking websites properly” may map onto lateral reading, source credibility, or civic online reasoning.
Search skill therefore includes vocabulary expansion.
- What would a researcher call this?
- What is the broader term?
- The narrower term?
- A competing term?
- The technical construct?
And once results arrive:
Which new vocabulary improves the next search?
Research search is iterative.
Search 1 teaches Search 2 how to ask a better question.
Worth learning because: the best source may never contain the learner’s original wording even though it directly investigates the underlying concept.
4. Learn to Identify Who Is Behind a Source Before Reading It as Evidence
A professional page can still be weak evidence.
A plain page can contain excellent evidence.
So the researcher asks:
- Who produced this?
- What is their relationship to the claim?
- What expertise do they possess?
- What evidence do they have access to?
- What incentives matter?
- What is the publication context?
Recent intervention research supports explicit instruction in this skill. A 2025 meta-analysis synthesising 64 controlled studies found that source-credibility interventions improved assessment skill on average, with lateral reading showing the largest effects; substantial heterogeneity remained, so no one intervention should be treated as universally effective. Read the study.
A 2024 review likewise found that explicit lateral-reading instruction—from elementary age through adulthood—can improve digital source evaluation. Read the review.
The important move is lateral:
leave the source; investigate the source elsewhere.
Do not allow a website to provide all the evidence for its own authority.
This remains distinct from Verification.
Research uses source evaluation to decide what deserves entry into the evidence set.
Verification decides whether a specific claim ultimately survives the checks.
Worth learning because: source quality changes how much evidential weight a piece of information deserves before it enters the research model.
5. Learn to Move Toward the Original Evidence
A blog says: “A study found…”
Find the study.
A news article quotes government statistics.
Find the official release or dataset.
A review summarises an experiment.
When the precise result matters, inspect the original study.
A social post shows a screenshot.
Find the original page.
An AI system cites a paper.
Open the paper.
This is not because original sources are always perfect.
They are not.
It is because each retelling adds another opportunity for selection, compression, rounding, framing, error and loss of conditions.
Research becomes stronger when important claims can be traced toward the evidence object that first carried them.
Sometimes the secondary source is the correct object.
If you are researching how newspapers framed an event, the newspaper article is primary evidence for that research question.
“Primary” is therefore relative to the job.
What source is closest to the evidence needed for this particular claim?
Worth learning because: important claims are easier to evaluate when the learner can inspect what was originally measured, written, observed or recorded.
6. Learn to Read Laterally Before You Read Deeply
There is a strange research habit that feels responsible but can waste enormous time:
reading a questionable source very carefully.
Line by line.
Annotating.
Highlighting.
Only later discovering that the source should never have been admitted.
Lateral reading reverses the order.
- Who is this?
- What do credible external sources say about it?
- Is this the kind of source that can know the thing it claims to know?
Then, if it passes:
read deeply.
This is one reason lateral reading is so efficient.
It allocates deep reading to sources that have earned it.
But lateral reading is not a replacement for close reading.
Once the source is admitted, its actual wording and evidence matter.
LATERAL FIRST → CLOSE SECOND
Worth learning because: research time is better spent deeply reading sources after their basic relevance and credibility have survived an external check.
7. Learn to Separate Claim, Evidence and Inference in Your Notes
Suppose a paper reports:
Students receiving feedback improved more than controls.
Your note says:
“Feedback is effective.”
Perhaps.
But the paper may concern one population, one task and one feedback type.
A better research note has layers.
SOURCE CLAIM:
Students in this intervention improved on outcome X.EVIDENCE:
Design, sample, comparison, measure.MY INFERENCE:
This supports using this feedback approach under similar conditions.LIMIT:
Does not establish that all feedback improves all writing.
This is a powerful habit because it prevents later writing from accidentally upgrading the strength of the evidence.
It also links naturally to MindOS Source-Monitoring State and the broader Critical Thinking architecture.
The research notebook should preserve enough structure that future you can tell what the source said, what the evidence was, and what you concluded from it.
Worth learning because: research becomes traceable when the learner does not silently turn their own interpretation into something “the source says.”
8. Learn to Compare Sources for Independence, Agreement and Conflict
Two sources agree.
Why?
Independent evidence?
Or copying?
Two sources disagree.
Why?
- Different population?
- Definition?
- Time period?
- Measurement?
- Method?
- Assumption?
Research is not finished when several sources have been collected.
The learner needs to model the relationships among them.
A useful matrix might ask:
Source A — claim — method — population — result.
Source B — same.
Then:
- Where do they genuinely converge?
- Where do they conflict?
- Which difference may explain the conflict?
- Which source is stronger for this particular question?
This is the clean handoff to Top 10 Synthesis Skills Worth Learning.
Research owns the end-to-end inquiry route.
Synthesis owns the construction of the larger model once distributed evidence exists.
Worth learning because: several sources become research evidence only when the learner understands how their evidential routes relate rather than merely counting how many agree.
9. Learn to Carry Uncertainty Instead of Hiding It
Good research often finishes with less certainty than weak research.
That sounds backwards.
It is not.
Weak research begins certain.
Finds support.
Ends certain.
Strong research may discover the evidence is mixed, the mechanism remains contested, the population is narrow, the effect is smaller than expected, or the original question was too broad.
Excellent.
The model improved.
Use phrases accurately:
- well established,
- supported,
- consistent with,
- suggestive,
- uncertain,
- contested,
- insufficient evidence.
Uncertainty is not a defect to remove before submission.
It is information about the state of the evidence.
This is exactly where Top 10 Verification Skills Worth Learning supplies a neighbouring HOLD state.
Research may conclude:
We do not yet have enough evidence to resolve this question.
That can be the correct answer.
Worth learning because: research is stronger when the confidence of its conclusion matches the evidence rather than the learner’s desire for closure.
10. Learn to Leave a Source Trail That Another Person Can Reopen
Research should be reversible.
Someone should be able to ask:
- Where did this number come from?
- Which article supports this statement?
- Which edition did you use?
- What date?
- What search led you here?
- Why did you exclude the competing source?
A source trail does not need to record every click.
It needs enough provenance to reopen consequential claims.
For school work: author, title, publisher/site, date, URL where appropriate, access date where needed.
For serious research: full citation, DOI or stable identifier, specific page/table/figure where material, notes about the evidential role.
For AI-assisted research: AI may help search, summarise or formulate queries. But the final source trail should lead to the underlying evidence—not only to an AI conversation.
The test is simple:
Could a future reader reproduce enough of my route to understand why I trusted this conclusion?
If yes, the research has a receipt.
Worth learning because: research becomes durable when important claims remain traceable after the browser tabs, memory and AI conversation are gone.
The Top 10 Research Skills as One System
QUESTION → KNOWN/ASSUMED → CONCEPTUAL SEARCH → SOURCE ACTOR → ORIGINAL EVIDENCE → LATERAL CHECK → CLAIM/EVIDENCE/INFERENCE → SOURCE COMPARISON → UNCERTAINTY → SOURCE TRAIL
The quieter Wintour House version:
Ask a question precise enough to fail. Find the evidence that could answer it. Know who produced that evidence. Keep source, claim and inference separate. Compare independent routes. Preserve uncertainty. Leave a path back.
That is research.
Research Is Not the Same as Search
Search finds candidates.
Research decides which candidates deserve entry, how they relate, what they establish and what remains unresolved.
MindOS Search State therefore keeps the search owner.
Research consumes search.
It does not collapse into it.
Research Is Not the Same as Verification
Verification tests whether one claim has earned acceptance.
Research has a wider route: question → evidence discovery → source admission → comparison → synthesis → conclusion.
Verification is one gate inside the process.
Research Is Not the Same as Synthesis
Synthesis integrates distributed information.
Research includes synthesis but also owns question formation, search strategy, evidence acquisition, source admission and source trail.
A learner can synthesise a provided source packet without having done the research that assembled it.
Research Is Not the Same as Critical Thinking
Critical Thinking evaluates reasoning.
Research uses that capability across an evidence-seeking process.
One can think critically about a single argument without conducting research.
One can also gather enormous amounts of information while thinking critically very little.
Different job.
For Primary Students
Primary research can be wonderfully small.
Question: Why do some materials float?
Before searching: What do we already know? What do we think? What could we test?
One short book.
One experiment.
One teacher-approved digital source.
Then: What did each source add? Did anything disagree? Can we explain our answer? Where did we get the information?
That is research.
Children need not begin with fifty websites.
They need the structure:
question → evidence → answer → source.
For Secondary Students
Secondary students should learn purposeful online search.
They should be able to reformulate questions, use technical vocabulary, investigate unfamiliar sources laterally, trace important claims, distinguish original evidence from commentary, compare several evidence routes, and cite material well enough to reopen it.
They should also learn to stop.
More tabs do not automatically mean better research.
Once the evidence is sufficient for the current question, move from acquisition to synthesis.
For JC Students
JC research should become much more epistemic.
- What design produced the evidence?
- What population?
- How was the construct measured?
- Which competing explanation survives?
- Is the source original, review-level or commentary?
- Does a later study change the conclusion?
- What is consensus and what remains disputed?
The learner should become comfortable stating:
This conclusion is well supported for X but not established for Y.
That is research maturity.
Research in the Age of AI
AI dramatically lowers the cost of the first-pass research brief.
That is useful.
It also creates a new failure: research without provenance.
The system produces ten elegant points.
The learner cannot say which came from which source.
So use AI carefully.
Ask it to help generate search terms, identify candidate sources, suggest competing hypotheses, locate terminology, and compare provided sources.
Then leave the AI and inspect the actual evidence.
Use the machine to widen and organise the search space. Do not let the machine become the final source trail.
The Wintour House Test
Suppose search engines become perfect.
Suppose AI can locate every relevant paper instantly.
Research skill still survives because someone must decide which question matters, what would count as evidence, which source can know, whether two sources are independent, where conditions differ, what remains uncertain, and what conclusion the evidence actually warrants.
The scarce resource becomes judgement.
That is why Research belongs in the Skills Worth Learning series.
Research Anchors
A 2025 meta-analysis of 64 controlled experimental studies found that interventions to improve source-credibility assessment were effective on average (g = .42), with lateral-reading approaches showing the largest effects. The prediction interval was wide, which is an important warning that outcomes vary across contexts; repeated practice and open-internet tasks appeared useful moderators. Read the meta-analysis.
McGrew’s 2024 review of lateral-reading interventions similarly concluded that targeted explicit instruction can improve online source evaluation across learners from elementary school through adulthood. Read the review.
These findings do not establish one universal ten-step research method. They support the narrower proposition that important parts of digital evidence evaluation are teachable. Wintour House therefore treats the ten skills here as an editorial end-to-end research architecture whose individual components remain linked to specialist owners.
The strongest defensible conclusion is:
Research is the disciplined construction of a traceable route from a meaningful uncertainty to a proportionate conclusion: formulate the question, search conceptually, evaluate who is behind the evidence, return toward original sources, separate claims from inference, compare independent evidence, preserve uncertainty and leave a trail that can be reopened.
