VIEW THIS AS

Auto mode follows the Route Engine until you choose a viewpoint.

YOU ARE HERE

ROUTE CHECK

CONNECTED TO

WHAT NEXT

Use the canonical route for this room, or HELP if you are unsure.

How to be Good at Research

Three secondary students review open books and notes together in a classroom.

How to be good at research? Start with the most useful question in the whole process: What would count as knowing?

Research is not collecting many tabs.

It is not copying quotations into a document.

It is not finding one source that agrees with you.

Research is a disciplined way of moving from a question to a justified answer while leaving enough evidence for another person to inspect the route.

That is the gold standard: not knowing everything, but knowing what the evidence supports, what remains uncertain, and why your conclusion deserves confidence.


Did You Know? Research Begins by Shrinking Ignorance

A huge topic is not yet a research question.

“Artificial intelligence” is a topic.

“How does retrieval-augmented generation reduce unsupported answers in a school knowledge system, and where can it still fail?” is closer to a research question.

“Singapore history” is a topic.

“How did the development of public housing alter everyday access to schools, transport and town services across different planning periods?” creates something that can be investigated.

Good research turns curiosity into a bounded uncertainty.


The Gold-Standard Research Loop

A practical research loop is:

  • Question — define what you are trying to know.
  • Scope — decide what is inside and outside the inquiry.
  • Map — identify concepts, variables, actors, time periods and competing explanations.
  • Search — find relevant evidence systematically.
  • Evaluate — judge source quality and fit.
  • Extract — capture claims, data, methods and limitations accurately.
  • Compare — put sources into conversation rather than stacking them.
  • Reason — decide what the evidence supports.
  • Verify — challenge the conclusion.
  • Write — show the reader the route from evidence to claim.
  • Update — revise when stronger evidence appears.

Research quality lives in the connections between those steps.


Step 1: Turn the Topic Into a Researchable Question

A strong question is specific enough to investigate but open enough to require thinking.

Useful question patterns include:

  • What causes X under Y conditions?
  • How does X differ across A and B?
  • What changed after X?
  • Which explanation best fits the evidence?
  • What trade-offs appear when X is implemented?
  • How reliable is X for a particular purpose?
  • Where does a commonly accepted claim stop being true?

Avoid questions that are so broad that every source becomes ‘relevant’.

Scope is not a weakness.

Scope is what makes evidence comparable.


Step 2: Define the Terms

Many research arguments are actually definition arguments in disguise.

What counts as success?

What counts as learning?

What counts as a civilisation?

What counts as a ‘good school’, ‘rich country’, ‘safe system’ or ‘effective intervention’?

Define important terms early.

If the definition is contested, show the alternatives and explain which operational definition you are using.

This is one reason vocabulary sits at the centre of eduKate’s knowledge graph: words define the handles by which we operate complex ideas.


Step 3: Build a Question Map

Before searching, decompose the question.

For example, a question about whether a study method improves learning might contain:

  • learner age;
  • subject domain;
  • study method;
  • comparison condition;
  • outcome measure;
  • delay before testing;
  • classroom or laboratory setting.

Now search becomes structured.

You are no longer searching the internet for ‘everything about studying’.

You are searching for evidence connected to specific parts of the model.


Step 4: Search in Layers

Strong researchers do not expect one search query to do all the work.

Use layers:

  • Orientation search — learn the vocabulary of the field.
  • Authority search — find official, scholarly or primary sources.
  • Specific search — locate evidence for individual claims.
  • Contradiction search — actively look for disagreement or failure cases.
  • Update search — check whether newer evidence changes the picture.

Search terms should evolve as your understanding improves.

The first search helps you learn what to search for next.


Step 5: Prefer Primary Sources When the Question Needs Them

A secondary summary can be excellent for orientation.

But some claims should be checked against the primary source.

  • government rules — official agency;
  • scientific findings — original paper;
  • company results — filed report or official release;
  • historical document — archive or original text;
  • statistics — the agency or dataset that produced them;
  • school policy — the school or Ministry page.

Primary does not automatically mean perfect.

It means closer to the original event, measurement or decision.


Step 6: Evaluate Sources by Function

Do not ask only, ‘Is this source trustworthy?’

Ask, ‘Trustworthy for what?’

A source may be excellent for one claim and weak for another.

Evaluate:

  • Authority — who produced it and with what expertise?
  • Proximity — how close is it to the original evidence?
  • Method — how was the conclusion produced?
  • Currency — does the claim depend on current conditions?
  • Scope — does the evidence match the population, place and period?
  • Incentives — what interests or constraints may shape presentation?
  • Transparency — can you inspect data, method or references?

This avoids a common mistake: treating source reputation as a substitute for source fit.


Step 7: Extract Claims Precisely

When taking notes, separate:

  • what the source directly reports;
  • what the authors infer;
  • what you infer;
  • what remains uncertain.

Record page numbers, table numbers, dates and URLs while you are reading.

Future-you will be grateful.

A research notebook should make it possible to reconstruct why a claim entered the draft.


Step 8: Compare Sources, Do Not Stack Them

Weak research often looks like this:

Source A says X. Source B says Y. Source C says Z.

Then the paragraph ends.

Strong research asks how the sources relate.

  • Do they measure the same thing?
  • Do they study the same population?
  • Do they use different methods?
  • Do they disagree?
  • Could both be true under different conditions?
  • Which evidence is stronger for the specific question?

Research becomes valuable when sources begin talking to each other.


Step 9: Distinguish Correlation, Mechanism and Causation

If two things change together, that does not automatically mean one caused the other.

Ask:

  • What alternative explanations exist?
  • What happened first?
  • What mechanism is proposed?
  • Was there a comparison group?
  • Could selection effects explain the pattern?
  • Could a third variable influence both?

This habit is essential in education, economics, health, technology and social research.

Causal language should be earned.


Step 10: Pay Attention to Denominators

Numbers can look impressive while hiding the base.

A 50% increase from 2 to 3 is real.

It is also different from a 50% increase from 20,000 to 30,000.

Always ask:

  • 50% of what?
  • How many observations?
  • Over what period?
  • Compared with what baseline?
  • Absolute number or relative change?

Good research makes quantities interpretable.


Step 11: Look for Missingness

Evidence is shaped by what is recorded.

Ask what might be absent:

  • people who never responded;
  • events that were never documented;
  • failed experiments that were not published;
  • data excluded by collection rules;
  • populations not represented;
  • records lost over time.

Missing data does not automatically invalidate a conclusion.

It changes what the conclusion can safely claim.


Step 12: Search for Counterevidence

Confirmation is comfortable.

Research is stronger when it survives an attempt to disconfirm it.

Try queries that oppose your current view.

Look for failed replications, alternative explanations, critical reviews and boundary conditions.

If the conclusion remains strong after that, confidence becomes better justified.


Step 13: Use Tables to Think

A comparison table can reveal contradictions faster than paragraphs.

Useful columns might include:

  • source;
  • year;
  • population;
  • method;
  • sample size;
  • main finding;
  • limitation;
  • relevance to your question.

The table is not just presentation.

It is a reasoning tool.


Step 14: Separate Evidence Quality From Conclusion Strength

You can have a strong conclusion from several consistent high-quality sources.

You can have a tentative conclusion from limited evidence.

You can have an interesting hypothesis with almost no evidence yet.

Use language that matches the evidence.

  • shows;
  • suggests;
  • is consistent with;
  • may indicate;
  • does not establish;
  • remains uncertain.

Good research is precise about confidence.


Step 15: Build a Citation Trail

A citation is not decoration.

It lets the reader inspect the evidence.

Cite claims that are:

  • factual and non-obvious;
  • quantitative;
  • historical;
  • attributed to a source;
  • dependent on a study;
  • current or changeable.

Use the smallest set of strong sources that actually support the claim.


Step 16: Write the Argument After the Evidence Map

Do not force the evidence into the outline you imagined before researching.

Let the evidence change the structure.

A strong research paragraph often follows:

  • claim;
  • evidence;
  • interpretation;
  • limitation or contrast;
  • connection to the larger question.

The reader should be able to see how the conclusion was built.


Step 17: Research With AI Without Losing Source Control

AI can accelerate orientation, query generation, summarisation and comparison.

But fluent synthesis can hide source errors.

Use AI carefully:

  • ask for search terms and competing hypotheses;
  • request a source table;
  • verify every important source independently;
  • open the original document;
  • check whether quotations and numbers actually exist;
  • separate AI synthesis from source evidence;
  • do not cite a source you did not inspect.

The gold-standard workflow uses AI to expand research capacity while preserving provenance.


Research in Singapore’s Knowledge Graph

Singapore’s libraries, universities, government agencies, archives and open-data systems form a research substrate.

The National Library Board provides access to collections, reference resources and digital services. Government agencies publish policy, statistical and programme information. Universities generate scholarly work. Archives preserve historical evidence.

A strong researcher learns which layer owns which kind of truth.

That is the same one-owner principle eduKate uses in its content graph: one canonical owner for an intent, with supporting pages linking into it rather than competing blindly.


A 30-Day Scaffold for Becoming Better at Research

Week 1: Questions

  • Turn four broad topics into researchable questions.
  • Define the key terms.
  • Build question maps.
  • Write scope boundaries.

Week 2: Sources

  • Find primary and secondary sources.
  • Create an evidence table.
  • Record provenance carefully.
  • Search for disagreement.

Week 3: Reasoning

  • Compare methods.
  • Check denominators.
  • Separate causation from correlation.
  • Write confidence-qualified conclusions.

Week 4: Synthesis

  • Draft an argument from the evidence map.
  • Add citations.
  • Run a contradiction search.
  • Revise the conclusion if necessary.
  • Create a short research brief from the longer work.

Common Research Traps

The First-Page Trap

Search ranking is not evidence ranking.

The One-Source Trap

One source can orient you, but important claims often require comparison or primary verification.

The Quote-Collection Trap

A document full of quotations is not yet an argument.

The Confirmation Trap

If every source agrees with your starting position, you may not have searched widely enough.

The Recency Trap

Newer is not automatically better for historical or foundational claims. Current claims, however, do require current evidence.

The AI-Citation Trap

Never assume an AI-generated citation is real or supports the stated claim.


Frequently Asked Questions

How many sources do I need?

There is no universal number. Use enough strong, relevant sources to support the claims, represent material disagreement and show the evidence base without padding.

Is Wikipedia useful for research?

It can be useful for orientation and terminology, but important claims should generally be traced to stronger primary or scholarly sources where available.

How do I know whether a source is biased?

All sources have perspective and constraints. Examine incentives, method, omitted information and whether claims can be independently checked. Bias is not solved by simply labelling a source; it is managed through comparison and evidence.

Should I use old research?

Foundational work can remain important. For claims that depend on current conditions, technology, policy or population, check newer evidence.

Can AI do my literature review?

AI can assist discovery and synthesis, but source verification, inclusion criteria, interpretation and final responsibility should remain with the researcher.

What makes research ‘good’?

A clear question, appropriate evidence, transparent method, careful reasoning, honest uncertainty and a conclusion proportionate to the evidence.


Helpful Reading Inside eduKate


Public Research Resources


How to Be Good at Research

Research is curiosity with an audit trail.

Ask a question that can be investigated.

Define the terms.

Map the uncertainty.

Search in layers.

Inspect primary evidence.

Compare sources.

Challenge the conclusion.

Write what the evidence supports — no more and no less.

The gold standard is not sounding certain.

It is knowing why your level of certainty is justified.

Continue the series with How to be Good at Exams, How to be Good at Coding and How to be Good at Negotiation.

Properly taught kids shine a bright light into the future.