How Scientific Communication Works | From Result, Uncertainty and Context to a Message That Does Not Distort the Science

How Scientific Communication Works | From Result, Uncertainty and Context to a Message That Does Not Distort the Science is a world-facing guide to science communication, communicating scientific uncertainty, scientific writing, communicating evidence, risk communication, public understanding of science, science journalism, visual communication, scientific consensus, audience analysis, plain language, research communication and scientific integrity. Scientific communication is not the decorative stage after the science is finished. It is the disciplined translation of a result, its uncertainty, its scope and its evidential context into a message another person can understand without being pushed toward a conclusion stronger than the science permits.

Effective science communication depends on the goal, the audience, the setting and the scientific state of the evidence. The National Academies has emphasised that communication is more complex than simply replacing technical words with simpler ones: people approach scientific information with different knowledge, goals, values, beliefs and social contexts, while scientific findings themselves may be uncertain, incomplete, contested or applicable only to specific populations. A message can therefore be technically accurate sentence by sentence and still distort the science by hiding uncertainty, omitting boundaries, exaggerating consensus, compressing a correlation into causation or presenting a preliminary finding as settled knowledge.

This guide focuses on that translation problem within the eduKateYishun scientific-method library. How Scientific Observation Works focuses on what was detected and recorded; How Scientific Evidence Works focuses on how observations bear on claims; How Scientific Inference Works focuses on the reasoning from evidence to conclusion; How Scientific Uncertainty Works focuses on what is known, unknown and probabilistic. This guide follows the communication chain: scientific result → communication goal → audience → core claim → magnitude → uncertainty → scope → context → format → comprehension check → revision.

---

Choose your route

• Basic route: communicate the result without changing its scientific meaning.
• Uncertainty route: explain ranges, confidence, probability and unknowns without either hiding or dramatizing them.
• Audience route: adapt language and examples without changing the underlying claim.
• Visual route: graphs, tables, diagrams and images that clarify rather than mislead.
• Public route: press releases, social media, interviews and public-facing summaries.
• Technical route: abstracts, results sections, methods, supplementary material and scientific presentations.
• Risk route: communicate probability, severity, absolute risk and decision relevance.
• Advanced route: framing, trust, uncertainty, controversy, misinformation, consensus and corrections.

1. Scientific communication has a scientific job

The job is not merely:

make the science sound simple.

The job is:

make the science understandable while preserving what the science actually says, what it does not say, and how strongly it says it.

This is harder.

2. Communication can distort science without containing a factual lie

Imagine a study finds:

• small association;
• wide uncertainty;
• one population;
• observational design.

A headline says:

“Scientists discover that X causes Y.”

Every noun may refer to a real study.

The message is still scientifically wrong because:

• association became causation;
• magnitude disappeared;
• uncertainty disappeared;
• scope disappeared.

3. Communication is compression

A paper may contain:

• years of work;
• thousands of observations;
• dozens of analyses;
• many limitations.

A public message may contain:

• one sentence;
• one chart;
• one minute.

Communication is therefore a compression problem.

Good compression preserves the information needed for the audience’s task.

4. Compression always loses something

When a 10,000-word paper becomes a 50-word summary, information disappears.

The ethical/scientific question is:

Which information can be removed without changing the meaning of the scientific conclusion?

5. The irreducible scientific core

For many empirical claims, preserve at least:

• what was studied;
• what was found;
• how large the result was;
• how uncertain it was;
• who/what it applies to;
• whether the design supports association or causation.

6. Goal comes before format

Science communication can aim to:

• share discovery;
• increase understanding;
• support a decision;
• explain uncertainty;
• correct misinformation;
• enable scrutiny.

Different goals need different formats.

7. “Inform” and “persuade” are different goals

If the goal is:

“Help a patient understand treatment evidence,”

the communicator should clarify benefits, harms and uncertainty.

If the goal is:

“Convince the patient to choose Treatment A,”

the communication becomes advocacy.

Scientific communication should make that shift visible rather than disguise persuasion as neutral explanation.

8. Science communication and advocacy can coexist—but should not be confused

A scientist may advocate:

• vaccination;
• climate action;
• research funding.

The scientific evidence can be part of the argument.

But policy/value judgments should not be presented as if science alone mechanically determines them.

9. Audience is part of the communication system

The same scientific result may be communicated differently to:

• specialist researcher;
• doctor;
• parent;
• student;
• policy maker;
• journalist.

Changing language is acceptable.

Changing the claim is not.

10. Audience adaptation versus scientific distortion

Acceptable adaptation:

replace “95% CI” with “the likely range is still fairly wide” for an introductory audience, while retaining the interval elsewhere.

Distortion:

remove the interval and say “scientists are certain”.

11. The deficit model is incomplete

A common communication assumption is:

If people disagree with the science, they simply need more facts.

Research reviewed by the National Academies shows this is too simple.

People also bring:

• goals;
• values;
• beliefs;
• social identity;
• trust judgments.

Communication should understand the audience rather than merely increase information volume.

12. More information can reduce understanding

A message overloaded with:

• technical qualifiers;
• statistics;
• acronyms;
• background;

can hide the main finding.

Clarity is not the enemy of accuracy.

13. Clarity means organised accuracy

A clear scientific message answers in order:

1. What is the question?
2. What did we find?
3. How large is it?
4. How sure are we?
5. Who/what does it apply to?
6. What does it not establish?

14. Plain language is not childish language

Plain language uses:

• shorter sentences;
• defined technical terms;
• concrete verbs;
• logical order.

It does not require removing important concepts.

15. Define before simplifying

Some technical terms are worth teaching because they are precise.

Examples:

• randomisation;
• confidence interval;
• correlation;
• causal effect.

Define the term once, then use it consistently.

16. Avoid synonym drift

Scientific writing can become confusing when the same concept is called:

• risk;
• chance;
• probability;
• likelihood;

without defining whether the words mean the same thing.

Consistency supports comprehension.

---

17. Communicating uncertainty

Uncertainty is part of scientific knowledge.

It can arise from:

• sampling;
• measurement;
• model assumptions;
• future outcomes;
• limited evidence;
• unknown mechanisms.

The communication task is not to eliminate uncertainty from the message.

It is to explain the right uncertainty at the right level.

18. Hiding uncertainty can backfire

If a communicator initially sounds certain and later evidence changes, audiences may interpret scientific updating as incompetence or dishonesty.

Better:

state where uncertainty exists from the beginning.

19. Exaggerating uncertainty can also distort

If evidence is strong, saying “we know nothing for sure” creates false ambiguity.

Uncertainty should be proportional.

20. Three useful questions for uncertainty

1. What is uncertain?
2. How uncertain is it?
3. Does the uncertainty change the decision or conclusion?

21. Numerical uncertainty

Example:

Estimated effect = 12% reduction, with interval compatible with roughly 6% to 18% reduction.

This communicates:

• direction;
• magnitude;
• precision.

22. Structural uncertainty

Example:

“The estimate assumes unmeasured confounding is limited.”

This uncertainty may not appear inside the confidence interval.

23. Consensus uncertainty

Sometimes scientists disagree about:

• magnitude;
• mechanism;
• interpretation.

Communication should identify what is agreed and what remains disputed.

24. Future uncertainty

Forecasts are distributions over possible futures, not guarantees.

Communicate:

• central estimate;
• range;
• conditions/scenarios.

25. Uncertainty is not ignorance

“We estimate 70% probability”

contains information.

It is not equivalent to:

“We have no idea.”

26. Ranges need interpretation

A range without explanation can confuse.

Say what the endpoints represent:

• confidence interval;
• credible interval;
• scenario range;
• min/max.

27. Avoid false precision

Do not communicate:

73.2841%

when the evidence supports only roughly:

70%.

Precision in digits should reflect precision in knowledge.

28. Avoid vague uncertainty words without context

Words such as:

• unlikely;
• possible;
• probable;

can mean different things to different audiences.

Where useful, pair words with numbers or examples.

---

29. Communicating magnitude

Scientific importance depends on how large an effect is, not only whether it exists.

30. Absolute risk

Suppose risk changes:

2 in 1,000 → 1 in 1,000.

Absolute reduction:

1 in 1,000.

31. Relative risk

The same change is a 50% relative reduction.

Both are correct.

Using only the relative number can exaggerate perceived impact.

32. Use absolute and relative risk together

Better:

Risk fell from 2 in 1,000 to 1 in 1,000—an absolute reduction of 1 in 1,000 and a relative reduction of 50%.

33. Natural frequencies

Many people understand:

“10 out of 1,000”

more easily than:

“1%”.

Natural frequencies can improve risk comprehension.

34. Use consistent denominators

Do not compare:

• 1 in 100;
• 8 in 10,000.

Convert to the same denominator when comparison matters.

35. Baseline matters

A 20% relative increase can mean:

• 1% → 1.2%;
• 50% → 60%.

Communicate baseline risk.

36. Number needed to treat

For some medical contexts, number needed to treat can help communicate how many people must receive an intervention for one additional beneficial outcome on average.

It depends on time horizon and baseline risk.

37. Effect size should be meaningful to the audience

A standardized effect of 0.3 may be useful to researchers.

Parents may understand:

• months of learning;
• score change;
• probability of reaching a threshold.

Translate carefully without inventing equivalences.

38. Small effects can matter at scale

A small individual effect can produce large population impact.

Communication should distinguish:

• individual magnitude;
• population magnitude.

39. Large percentage change can be trivial

If the baseline event is extremely rare, large relative changes may still have small absolute impact.

40. Magnitude belongs near the headline

“Drug lowers risk” is less informative than:

“Drug lowered 1-year risk from 10% to 8% in the trial.”

---

41. Audience analysis

Before writing, ask:

• What does the audience already know?
• What do they need to do with the information?
• Which terms will be unfamiliar?
• Which misconceptions are likely?
• Which values or goals matter to their decision?
• How much time or attention do they have?

42. Audience analysis is not audience manipulation

The purpose is to improve comprehension.

It should not be used to:

• hide inconvenient evidence;
• select only emotionally persuasive facts;
• change scientific meaning for different groups.

43. One result can have several legitimate messages

For a specialist:

report method, estimator, interval and assumptions.

For a patient:

report expected benefit, harms, absolute risk and uncertainty.

For a student:

teach the concept and evidence chain.

The scientific core remains stable.

44. Build from the audience’s task

If the audience needs to decide:

“Should I take this treatment?”

they need:

• baseline risk;
• benefit;
• harm;
• uncertainty;
• relevance to people like them.

They do not necessarily need every technical detail first.

45. Build from prior knowledge

An introductory audience may need:

• what a randomized trial is;
• what risk means.

An expert audience can move directly to:

• estimand;
• effect modification;
• sensitivity analysis.

46. Jargon can be useful when it compresses shared meaning

Among specialists, one term can replace a paragraph.

Jargon becomes a problem when:

• the audience does not share the definition;
• the term hides an assumption;
• the term substitutes for explanation.

47. Explain the concept before the acronym

Better:

“We used a randomized controlled trial (RCT), meaning participants were assigned by chance to treatment or control.”

Then use RCT later.

48. Concrete examples reduce abstraction cost

Instead of:

“A 0.02 absolute risk difference,”

add:

“about 2 fewer events per 100 people over the study period.”

49. Analogy can help—and mislead

Analogies can clarify:

• scale;
• mechanism;
• probability.

But every analogy breaks somewhere.

State the boundary when it matters.

50. Narrative can improve memory

Stories can make science:

• concrete;
• memorable;
• human.

But one vivid story should not replace population evidence.

---

51. Framing

Framing is the selection and emphasis of aspects of information.

All communication has some frame because not everything can be said at once.

52. Frame choice can change perception

Compare:

• 90% survive;
• 10% die.

These can describe the same outcome.

People may respond differently.

53. Balanced framing

Where framing affects risk perception, show both:

• survival;
• mortality.

or:

• benefit;
• harm.

54. Gain/loss framing

Messages can emphasize:

• benefits gained;
• losses avoided;
• risks incurred.

Choose a frame that serves comprehension rather than covert persuasion.

55. Emphasis framing

A communicator can truthfully emphasize:

• economic cost;
• health outcome;
• environmental effect.

Different emphases can lead to different impressions.

Important omitted dimensions should remain accessible.

56. Framing should not change evidential weight

A preliminary finding should remain preliminary whether the communication goal is:

• education;
• fundraising;
• media coverage.

57. Headlines are high-risk frames

Many readers see only:

• headline;
• thumbnail;
• first sentence.

The headline should not contain a claim stronger than the body.

58. Headline integrity test

Ask:

If someone reads only the headline, will they leave with the same scientific direction, magnitude and certainty as the study supports?

59. Avoid causal verbs for associative studies

Weak headline:

“Social media causes anxiety.”

Better:

“Study finds higher anxiety among heavier social-media users.”

Causal interpretation can be discussed separately.

60. Avoid universal verbs for local studies

Weak:

“Children learn better with Method X.”

Better:

“In this trial, students using Method X scored modestly higher on the delayed test.”

---

61. Visual communication

Charts and diagrams can reveal:

• magnitude;
• uncertainty;
• trend;
• comparison.

They can also distort.

62. Axis choices matter

A truncated axis can make a small difference look enormous.

Sometimes truncation is justified for resolution.

If used, it should be visually obvious and scientifically appropriate.

63. Show zero when zero matters

For bar charts representing magnitude from zero, a zero baseline is often important.

For line charts showing small changes around a large baseline, a non-zero axis may be reasonable.

Match design to the question.

64. Uncertainty should be visible

Possible visual elements:

• confidence intervals;
• credible intervals;
• prediction bands;
• scenario ranges.

65. Error bars need labels

Error bars can mean:

• standard deviation;
• standard error;
• 95% interval.

Readers should know which.

66. Avoid decorative precision

Too many gridlines, decimals and categories can make a chart look scientific while reducing understanding.

67. Use direct labels

Label lines or bars near the data when possible instead of forcing readers to repeatedly consult a legend.

68. Colour carries meaning

Colour should:

• distinguish categories;
• preserve accessibility;
• avoid emotional coding that exaggerates risk.

69. Red does not always mean danger

If a neutral scientific category is coloured alarm-red, the design can add interpretation not present in the data.

70. Maps can mislead

A choropleth map can make large geographic areas dominate visual attention even when population is small.

Communicate denominators and rates.

71. Area and volume are hard to judge

Bubble charts can exaggerate differences if radius is scaled instead of area.

Use perceptually honest encodings.

72. 3D charts usually add distortion

Perspective can make values difficult to compare.

Use 2D unless the third dimension represents actual scientific structure.

73. Tables remain powerful

A table is often better when readers need:

• exact values;
• many categories;
• reference lookup.

74. Visual hierarchy should match scientific hierarchy

The most visually prominent element should usually be the most scientifically important message.

75. Figure captions should stand alone

A strong caption explains:

• what is shown;
• sample/context;
• uncertainty;
• important caveat.

---

76. Communicating risk

Risk combines:

• probability;
• consequence;
• time horizon.

A risk message should specify all three where relevant.

77. “Doubles the risk” needs baseline

If risk doubles:

• 1 in 1,000 → 2 in 1,000;
• 1 in 2 → nearly certain.

Very different practical meanings.

78. Time horizon matters

“10% risk” is incomplete without:

• over one year?
• over a lifetime?

79. Compare like with like

Risk comparisons should use:

• same denominator;
• same time horizon;
• similar population.

80. Avoid fear-based denominator tricks

“1,000 people affected” sounds large.

If the population is 100 million, the rate is different from a population of 10,000.

81. Communicate benefits and harms symmetrically

If benefit is reported as relative risk and harm as absolute risk, perception can become biased.

Use comparable formats.

82. Risk ladders

Risk ladders compare probabilities against familiar reference risks.

They can help scale understanding.

References should be relevant and not emotionally manipulative.

83. Uncertain risk

When probability itself is uncertain, say so.

Example:

“Best estimate is 5%, but plausible values range from 2% to 10%.”

84. Scenario risk

Some risk depends on future choices or conditions.

Communicate scenario assumptions explicitly.

---

85. Technical scientific writing

Scientific papers communicate to readers who may want to:

• evaluate;
• replicate;
• reuse;
• challenge.

Technical communication therefore values precision and auditability.

86. Abstract

A strong abstract should contain:

• question;
• design;
• main result;
• magnitude;
• uncertainty;
• bounded conclusion.

87. Avoid abstract overclaiming

Many readers see only the abstract.

Limitations central to interpretation should not exist only in the final paragraph of the full paper.

88. Methods communication

Methods should be detailed enough that readers can understand:

• sample;
• measurement;
• design;
• analysis.

89. Results versus discussion

Results primarily report what the analyses found.

Discussion interprets meaning, alternatives and implications.

Keeping the distinction visible reduces interpretive drift.

90. Limitations section

Limitations should identify issues that can change:

• magnitude;
• confidence;
• scope;
• causal interpretation.

Do not fill the section with ritual caveats that have no connection to the conclusion.

---

91. Public-facing summaries

A public summary should not merely remove equations.

It should reorganise the science around the audience’s needs.

92. Public-summary core

1. What question was studied?
2. What did researchers do?
3. What did they find?
4. How large was it?
5. How certain is it?
6. What does it mean?
7. What does it not mean?

93. Press releases

Press releases often shape downstream headlines.

They should avoid:

• causal overstatement;
• relative-risk-only framing;
• ignoring study limitations;
• quotes stronger than the paper.

94. Quotes can distort the paper

A cautious abstract followed by a dramatic researcher quote still creates overclaiming.

Quote integrity matters.

95. Embargoes do not change evidential status

A result does not become more certain because it is newsworthy or embargoed by a major journal.

96. Science journalism

Journalists should ask:

• study design?
• sample size?
• effect size?
• absolute risk?
• replication?
• conflicts?

97. One-study news is inherently provisional

Place a new study in the existing evidence base.

Ask whether it:

• confirms;
• extends;
• contradicts;
• stands alone.

98. Preprints

Preprints can accelerate communication.

They have not completed journal peer review.

That status should remain visible.

99. Conference findings

Early conference results may be incomplete.

Communicate them as preliminary where appropriate.

100. Retractions and corrections should remain attached to the story

If a result changes materially, public communication should update.

---

101. Cross-Domain Science Communication Casebook

Case 1 — Medical trial

Study result: treatment reduces event risk from 10% to 8%.

Poor communication: “New drug cuts risk by 20%.”

Better communication: “In the trial, risk fell from 10 in 100 to 8 in 100 over one year—a 2 percentage-point absolute reduction and 20% relative reduction.”

Case 2 — Screening test

Poor communication: “The test is 95% accurate, so a positive result means you almost certainly have the disease.”

Better communication: explain sensitivity, specificity and disease prevalence.

A highly accurate test can still produce many false positives when a condition is rare.

Case 3 — Vaccine adverse-event signal

Observation: reports increased after rollout.

Poor communication: “Vaccine causes the event.”

Better communication: “A safety signal has appeared and is being investigated; reports alone do not establish causation.”

Case 4 — New cancer treatment

Poor communication: “Breakthrough cures cancer.”

Better communication: specify cancer type, trial phase, response rate, duration and follow-up.

Case 5 — Nutrition study

Study: observational association between food and health outcome.

Poor communication: “Eating Food X prevents disease.”

Better communication: “People reporting more Food X had lower disease rates; the study cannot prove Food X caused the difference.”

Case 6 — Exercise study

Result: moderate improvement on one fitness measure.

Poor communication: “Exercise programme transforms health.”

Better communication: name the measured outcome and magnitude.

Case 7 — Psychology finding

Result: small behavioural effect in one sample.

Poor communication: “Scientists reveal how the human mind works.”

Better communication: “The study found a small effect on Task T in this sample; replication is needed.”

Case 8 — Education intervention

Result: students score 3 points higher after eight weeks.

Poor communication: “Method X makes children smarter.”

Better communication: “Students using Method X scored about 3 points higher on the tested assessment after eight weeks.”

Case 9 — Retrieval practice

Result: better delayed recall.

Poor communication: “Testing is always the best way to learn.”

Better communication: explain what retrieval practice improved, compared with what, under which conditions.

Case 10 — AI benchmark

Result: model achieves higher benchmark score.

Poor communication: “AI now reasons better than humans.”

Better communication: “Model scored higher on Benchmark B under the published evaluation; broader reasoning ability is a separate claim.”

Case 11 — AI hallucination study

Result: factual-error rate under one prompt set.

Poor communication: “Model hallucinates 20% of the time.”

Better communication: “On this benchmark and prompt protocol, 20% of evaluated responses contained at least one factual error.”

Case 12 — Climate projection

Poor communication: “Scientists predict exactly what the temperature will be in 2100.”

Better communication: explain scenarios, model ranges and physical constraints.

Case 13 — Extreme weather attribution

Poor communication: “Climate change caused this storm.”

Better communication: “Attribution analysis estimates how climate change altered the probability or intensity of events like this.”

Case 14 — Biodiversity decline

Result: monitored populations decline.

Poor communication: “Nature has collapsed everywhere.”

Better communication: specify taxa, geography, monitoring coverage and trend.

Case 15 — Air pollution

Result: higher exposure associated with higher hospitalisation.

Poor communication: “Air pollution sends people to hospital.”

Better communication: explain study design, effect size and causal evidence separately.

Case 16 — Astronomy discovery

Observation: candidate exoplanet signal.

Poor communication: “Second Earth discovered.”

Better communication: “A planet candidate of approximately X size has been detected; atmosphere and habitability are not yet known.”

Case 17 — Gravitational-wave candidate

Poor communication: “Einstein proved right again” before signal validation.

Better communication: distinguish candidate, detection significance and later confirmation.

Case 18 — Archaeology

Observation: unusual structure at a site.

Poor communication: “Lost civilisation discovered.”

Better communication: identify the structure, dating evidence and competing interpretations.

Case 19 — Genetics

Result: variant associated with disease risk.

Poor communication: “Scientists find the gene for Disease D.”

Better communication: “Variant V is associated with a modest change in disease risk; many genetic and environmental factors contribute.”

Case 20 — Neuroscience

Result: brain region shows activity difference.

Poor communication: “Scientists locate love in the brain.”

Better communication: “Region R showed greater signal during the task; this does not mean the region uniquely causes the experience.”

Case 21 — Engineering safety

Result: component failure rate estimated at 1 in 10,000 hours.

Poor communication: “System is 99.99% safe.”

Better communication: specify failure definition, exposure duration, uncertainty and operating conditions.

Case 22 — Autonomous vehicle

Result: lower collision rate in one monitored fleet.

Poor communication: “Self-driving cars are safer than humans.”

Better communication: compare exposure, road type, driver supervision and uncertainty.

Case 23 — Software performance

Result: algorithm 30% faster on one workload.

Poor communication: “New algorithm is 30% faster.”

Better communication: “On Workload W using Hardware H, median runtime was 30% lower.”

Case 24 — Cybersecurity

Observation: unusual traffic pattern.

Poor communication: “Massive cyberattack confirmed.”

Better communication: “An anomalous pattern is under investigation; attack attribution is not yet established.”

Case 25 — Economic study

Result: association between policy and employment.

Poor communication: “Policy destroys jobs.”

Better communication: report estimated effect, design, assumptions and uncertainty.

Case 26 — Survey

Result: 52% support with sampling margin.

Poor communication: “Most people support Policy P.”

Better communication: “In this survey, estimated support was 52%, with uncertainty that overlaps a closely divided population.”

Case 27 — Rare adverse event

Result: risk estimated as 4 per million.

Poor communication: “Treatment linked to deadly condition.”

Better communication: present absolute risk, baseline comparison and seriousness.

Case 28 — Disease outbreak

Early data: small number of severe cases.

Poor communication: either “Nothing to worry about” or “Catastrophe certain”.

Better communication: explain what is known, what is unknown, and what would change the assessment.

Case 29 — School performance data

Result: one school’s average score rises.

Poor communication: “School quality improved.”

Better communication: discuss cohort differences, test changes and relevant indicators.

Case 30 — Small-group tuition

Observation: one student’s marks improve.

Poor communication: “Our programme guarantees results.”

Better communication: describe the student’s observed trajectory without claiming universal or guaranteed outcomes.

Case 31 — Weather forecast

Forecast: 70% chance of rain.

Poor communication: “It will rain.”

Better communication: explain probability in the forecast context.

Case 32 — Hurricane cone

Poor communication: interpret cone boundary as storm-size boundary.

Better communication: explain what the cone represents and what it does not.

Case 33 — Earthquake risk

Estimate: probability over 30 years.

Poor communication: “Major earthquake expected soon.”

Better communication: include time horizon and probabilistic nature.

Case 34 — Food safety

Result: contaminant detected below regulatory limit.

Poor communication: “Toxic chemical found in food.”

Better communication: report concentration, threshold and uncertainty.

Case 35 — Environmental chemical

Hazard evidence: high-dose animal toxicity.

Poor communication: “Chemical causes cancer.”

Better communication: distinguish hazard from exposure-dependent risk.

Case 36 — Meta-analysis

Result: small average effect with heterogeneity.

Poor communication: “Science settles the debate.”

Better communication: report average, uncertainty, heterogeneity and publication-bias considerations.

Case 37 — Failed replication

Poor communication: “Study debunked.”

Better communication: compare effect estimates, uncertainty and total evidence.

Case 38 — Retraction

Poor communication: “All related science is false.”

Better communication: explain why the paper was retracted and whether independent evidence remains.

Case 39 — Scientific consensus

Poor communication: “97% agree, therefore no uncertainty exists.”

Better communication: explain what the consensus concerns and which details remain uncertain.

Case 40 — New preliminary result

Poor communication: “Scientists have discovered…”

Better communication: “A preliminary study reports…; independent confirmation is still needed.”

---

102. Communication Failure Atlas

Failure 1 — Association becomes causation

Repair: use association language unless causal design supports more.

Failure 2 — Relative risk without absolute risk

Repair: provide both.

Failure 3 — No baseline

Repair: tell readers where the risk started.

Failure 4 — No time horizon

Repair: state over what period risk applies.

Failure 5 — Uncertainty hidden

Repair: show interval, range or caveat near the claim.

Failure 6 — Uncertainty exaggerated

Repair: distinguish remaining uncertainty from established core.

Failure 7 — One study becomes consensus

Repair: place study in evidence base.

Failure 8 — Consensus becomes certainty

Repair: state what is agreed and what remains open.

Failure 9 — Preliminary becomes established

Repair: label preprint, conference or pilot status.

Failure 10 — Headline stronger than paper

Repair: headline integrity test.

Failure 11 — Quote stronger than abstract

Repair: align public quote with paper’s inferential strength.

Failure 12 — Magnitude omitted

Repair: report effect size.

Failure 13 — Denominator omitted

Repair: give rates, not only counts.

Failure 14 — Inconsistent denominators

Repair: standardise comparison.

Failure 15 — Tiny decimal precision

Repair: round to meaningful precision.

Failure 16 — Jargon dump

Repair: define technical terms and prioritise core message.

Failure 17 — Oversimplification removes mechanism or boundary

Repair: simplify language, not scientific structure.

Failure 18 — Analogy treated literally

Repair: state where analogy stops.

Failure 19 — Anecdote replaces evidence

Repair: use story as illustration, not population estimate.

Failure 20 — Graph axis exaggeration

Repair: honest scale and labels.

Failure 21 — Error bars unlabeled

Repair: define uncertainty representation.

Failure 22 — Colour implies danger

Repair: neutral palette unless danger is scientifically relevant.

Failure 23 — Decorative 3D distortion

Repair: use 2D.

Failure 24 — Map denominator hidden

Repair: use rates or population context.

Failure 25 — Cherry-picked time window

Repair: show relevant longer context.

Failure 26 — Cherry-picked subgroup

Repair: label exploratory and show overall result.

Failure 27 — Values disguised as science

Repair: separate empirical findings from normative recommendation.

Failure 28 — “Scientists say” without source

Repair: identify evidence or authoritative synthesis.

Failure 29 — False balance

Repair: represent evidence weight, not equal airtime.

Failure 30 — Appeal to consensus without explaining evidence

Repair: consensus is useful context, not substitute for reasoning.

Failure 31 — Correction buried

Repair: update the original communication visibly.

Failure 32 — Retraction context omitted

Repair: explain reason and downstream evidence.

Failure 33 — Audience blamed for misunderstanding

Repair: test message comprehension.

Failure 34 — Information volume mistaken for clarity

Repair: prioritise and layer information.

Failure 35 — Emotional framing overwhelms magnitude

Repair: restore quantitative context.

Failure 36 — Technical caveat placed where nobody sees it

Repair: put interpretation-changing limitations near main claim.

Failure 37 — Misinformation repeated without correction structure

Repair: lead with accurate fact, explain misconception briefly, reinforce accurate model.

Failure 38 — Old result communicated as current

Repair: date and update evidence status.

Failure 39 — Statistical significance used as headline

Repair: communicate magnitude and practical relevance.

Failure 40 — Communication goal remains hidden

Repair: state whether the message aims to inform, support decision or advocate.

---

103. Science Communication Practice Bank: 30 reusable tasks

Practice 1 — Preserve the claim

Take one technical result and rewrite it for a 12-year-old without changing:

• direction;
• magnitude;
• uncertainty;
• scope.

Practice 2 — Headline integrity

Write a headline that remains accurate even if the reader sees nothing else.

Practice 3 — Add the baseline

Rewrite a relative-risk statement to include starting risk.

Practice 4 — Add the denominator

Turn “1,000 cases” into a population rate.

Practice 5 — Add the time horizon

Turn “10% risk” into a complete statement.

Practice 6 — Association versus causation

Rewrite a causal headline for an observational study.

Practice 7 — Preliminary status

Rewrite a press release so a preprint remains visibly preliminary.

Practice 8 — Uncertainty without paralysis

Explain a 70% probability without saying either “certain” or “we know nothing”.

Practice 9 — Magnitude before significance

Summarise a study without using the words “statistically significant”.

Practice 10 — Absolute and relative risk

Present both in one sentence.

Practice 11 — Natural frequency

Convert a percentage into “x out of 1,000”.

Practice 12 — Audience shift

Write the same finding for:

• researcher;
• parent;
• policy maker.

Practice 13 — Jargon audit

Underline every technical term and decide:

• keep;
• define;
• replace.

Practice 14 — Analogy boundary

Use one analogy, then write where it fails.

Practice 15 — Graph honesty

Redesign a truncated-axis chart to preserve magnitude honestly.

Practice 16 — Error-bar label

Add exactly what the uncertainty band represents.

Practice 17 — Caption stand-alone test

Give a figure to someone without the paper. Can they understand what it shows?

Practice 18 — Press-release quote

Write a researcher quote no stronger than the abstract.

Practice 19 — Limitation placement

Move one interpretation-changing caveat near the main result.

Practice 20 — Consensus statement

Explain what scientists broadly agree on and what remains uncertain.

Practice 21 — Misinformation correction

Lead with the accurate model, mention the false claim briefly, explain why, repeat the accurate model.

Practice 22 — Correction notice

Write an update that states:

• what changed;
• why;
• whether conclusion changed.

Practice 23 — Retraction communication

Explain why one paper’s retraction does or does not change the broader evidence base.

Practice 24 — Study-in-context paragraph

Place a new finding alongside earlier studies rather than treating it as isolated.

Practice 25 — Policy boundary

Separate:

• scientific evidence;
• policy recommendation;
• value judgment.

Practice 26 — Comprehension test

Ask a reader to explain the finding back in their own words.

Practice 27 — Recall test

After ten minutes, ask what the reader remembers. Did they remember magnitude and uncertainty or only the dramatic phrase?

Practice 28 — Visual memory test

Show the chart without caption and ask what conclusion it implies. Does that match the science?

Practice 29 — One-sentence core

Compress a result into one sentence without losing design, magnitude or scope.

Practice 30 — Full communication cycle

Result → goal → audience → core claim → uncertainty → format → comprehension test → revision.

---

104. Scientific communication through the school years

Primary 3 — Say what happened clearly

P3 pupils can learn to separate:

• observation;
• conclusion.

Communication begins with accurate description.

Primary 4 — Use comparison language

Students can communicate:

• more/less;
• faster/slower;
• same/different.

They should point to the evidence.

Primary 5 — Explain with evidence

Use the pattern:

result → scientific concept → explanation.

Do not skip from answer to unsupported reason.

Primary 6 — Communicate scope

Students can learn phrases such as:

• in this experiment;
• under these conditions;
• the results suggest.

This teaches bounded claims.

Secondary Science — Data communication

Students can learn:

• tables;
• graphs;
• uncertainty;
• result versus discussion.

JC and advanced study — Scientific argument

Students can communicate:

• methodological limits;
• statistical evidence;
• alternative interpretations;
• technical citations.

105. School Science communication ladder

1. Describe accurately.
2. Compare clearly.
3. Use evidence.
4. Explain mechanism.
5. Qualify scope.
6. Represent uncertainty.
7. Respond to alternative interpretation.

106. Scientific vocabulary should increase precision

Students should not use technical words merely to sound scientific.

Terms such as:

• evaporation;
• adaptation;
• energy transfer;

should compress a precise concept.

107. Answer architecture is communication architecture

A strong scientific answer often follows:

evidence → concept → mechanism → outcome.

This is both reasoning and communication.

---

108. The Science Communication Notebook

A communication notebook keeps the message tied to the science.

Field 1 — Source result

What exactly did the study or synthesis conclude?

Field 2 — Study status

Preprint, peer-reviewed study, replication, meta-analysis, guideline?

Field 3 — Communication goal

Inform, explain, support decision, correct misinformation, announce?

Field 4 — Audience

Who are they and what do they need?

Field 5 — Core claim

One sentence that preserves scientific meaning.

Field 6 — Magnitude

What number matters?

Field 7 — Uncertainty

What range or unknown changes interpretation?

Field 8 — Scope

Who/what/when does this apply to?

Field 9 — What it does not show

Which common overinterpretation must be prevented?

Field 10 — Format

Article, chart, slide, interview, video, thread?

Field 11 — Comprehension test

How will you know the audience understood?

Field 12 — Update rule

What will happen if the science changes?

109. Communication versioning

If evidence changes, preserve:

• old message;
• new evidence;
• updated message;
• date.

This builds public scientific memory.

110. Communication provenance

A public claim should be traceable back to:

• study;
• data;
• guideline;
• review.

“Scientists say” without provenance is weak communication.

---

111. Misinformation and correction

Scientific misinformation can involve:

• false claims;
• misleading statistics;
• outdated results;
• context removal.

112. Correct with the accurate model first

Useful structure:

1. State the accurate fact.
2. Briefly identify the false or misleading claim.
3. Explain why it is wrong.
4. Restate the accurate model.

This reduces the risk that the correction amplifies the false claim.

113. Explain the mechanism of the misconception

If a misleading statistic uses relative risk without baseline, show how the framing changes perception.

Teaching the error pattern improves resilience.

114. Corrections should be visible

If an institution posts an inaccurate statement, correcting only a hidden archive page is insufficient.

Update the original channel where feasible.

115. Corrections should state whether the conclusion changed

Example:

• number corrected;
• main conclusion unchanged.

or:

• analysis error corrected;
• conclusion materially weakened.

116. Retractions

A retraction means the paper should not be relied upon as part of the scientific literature in the same way as a valid paper.

But one retraction does not automatically erase:

• independent studies;
• replications;
• mechanistic evidence.

117. Retraction reason matters

Reasons can include:

• honest error;
• unreliable data;
• ethical problems;
• misconduct.

Communication should not imply a reason that has not been established.

118. Debunking should not become ridicule

Ridicule can make communication adversarial and reduce trust.

Correct the claim and explain the evidence.

119. “Do your own research” problem

Independent inquiry is valuable.

But evaluating science requires:

• study design;
• statistics;
• evidence synthesis.

Help audiences understand how to evaluate sources rather than merely telling them whom to trust.

120. Source hierarchy is context-dependent

For a current scientific question, useful sources can include:

• systematic reviews;
• authoritative scientific reports;
• high-quality primary studies.

Popularity or search ranking is not scientific authority by itself.

---

121. Trust and credibility

Science communication is interpreted partly through judgments about the communicator.

Trust can depend on:

• expertise;
• transparency;
• honesty about uncertainty;
• conflicts of interest;
• track record.

122. Expertise is not enough

An expert can communicate badly.

A clear communicator can lack domain expertise.

High-stakes science communication needs both scientific competence and communication competence.

123. Transparency can support trust

State:

• funding;
• conflicts;
• uncertainty;
• limitations.

Transparency does not guarantee agreement.

It makes the basis of the message inspectable.

124. Overconfidence can damage trust

If communicators routinely oversell early findings, later corrections can look like reversals.

Calibrated uncertainty builds more durable credibility.

125. “Trust us” is weaker than “here is why”

Whenever possible, communicate:

• evidence;
• method;
• independent checks.

126. Institutional authority still matters

Institutions can aggregate:

• expertise;
• review;
• quality control.

But authority should remain connected to evidence.

127. Conflicts of interest should be visible

Funding does not automatically invalidate research.

Hidden conflicts can undermine credibility and prevent readers from evaluating incentives.

128. Correcting yourself can increase credibility

Science is expected to update.

A visible correction can demonstrate integrity rather than failure.

---

129. Scientific communication channels

The same result behaves differently across communication channels because each channel imposes different limits on:

• length;
• visuals;
• interaction;
• revision;
• audience attention.

130. Research papers

A research paper should optimise for:

• precision;
• auditability;
• methodological detail;
• bounded interpretation.

Its audience is expected to inspect the reasoning rather than only remember one headline.

131. Abstracts

Because abstracts are widely read in isolation, the main inferential boundary must appear there.

An abstract should not:

• say “caused” when the full paper says “associated”;
• omit the population that limits generalisation;
• replace magnitude with significance alone.

132. Graphical abstracts

Graphical abstracts compress even more aggressively.

They should preserve:

• comparison;
• direction;
• magnitude or central result;
• study type.

Decorative arrows should not imply causal mechanisms not established by the study.

133. Posters

A scientific poster should let a passer-by recover:

1. question;
2. method;
3. result;
4. conclusion;
5. uncertainty.

Dense walls of text reduce communication quality even when technically complete.

134. Oral presentations

Talks require temporal sequencing.

Listeners cannot scan backward as easily as readers.

Useful structure:

• question;
• why it matters;
• design;
• result;
• interpretation;
• limits.

135. Slides should support the speaker, not duplicate a paper

Slides are most useful when they:

• show key evidence;
• highlight comparisons;
• reduce cognitive load.

Do not place paragraphs on screen and read them aloud.

136. One scientific claim per slide is often enough

If a slide contains five conclusions, the audience may remember none of them accurately.

137. Visual pacing

Reveal a complex graph in stages if the audience needs to learn how to read it.

Explain:

• axes;
• groups;
• uncertainty;
• main pattern.

138. Conferences and Q&A

Questions can reveal where communication failed.

If several listeners ask:

“Was this causal?”

the talk may not have made the design boundary clear enough.

139. Interviews

Before an interview, prepare:

• one-sentence core finding;
• one magnitude statement;
• one uncertainty statement;
• one “what this does not show” statement.

140. Bridge without evasion

If asked an unrelated question, a communicator can bridge back to the evidence.

But bridging should not be used to avoid legitimate limitations.

141. Quotable does not mean exaggerated

A memorable sentence can still be accurate.

Example:

“The effect is real enough to measure, but too small and uncertain to justify the dramatic claims circulating online.”

142. Podcasts and long-form interviews

Long form allows:

• mechanism;
• history;
• uncertainty;
• alternative explanations.

Use the extra time to deepen context, not merely repeat the conclusion.

143. Video

Video combines:

• spoken explanation;
• visual evidence;
• emotion;
• editing.

Editing can change perceived certainty by removing pauses, caveats or context.

144. Short-form video

Short videos are high-risk for oversimplification.

A useful pattern:

1. accurate hook;
2. main result;
3. magnitude;
4. one essential caveat;
5. source.

145. Social media

Social media encourages:

• speed;
• shortness;
• conflict;
• novelty.

These incentives can conflict with scientific calibration.

146. Thread structure

A scientific thread can use:

1. main finding;
2. study design;
3. effect size;
4. uncertainty;
5. limitations;
6. link to source.

147. Do not hide the caveat in post 12

If a limitation changes the meaning of the headline, it belongs near the beginning.

148. Screenshots of graphs need context

A cropped graph without:

• axes;
• caption;
• sample;
• uncertainty;

can become misinformation even when the original figure was sound.

149. Social virality can detach claims from evidence

As content spreads, it can lose:

• source;
• qualifiers;
• date.

Build provenance into the graphic or caption where possible.

---

150. Policy communication

Policy makers need science that is:

• relevant;
• timely;
• decision-focused;
• explicit about uncertainty.

151. Policy brief core

A useful policy brief can separate:

• what science finds;
• what remains uncertain;
• which options exist;
• which trade-offs are values or policy choices.

152. Science does not choose every policy objective

Evidence can estimate:

• health effects;
• costs;
• risks.

Societies still decide:

• which goals matter most;
• which risks are acceptable;
• how benefits are distributed.

153. Communicate policy scenarios

Instead of:

“Science says choose Policy A,”

write:

“Under Policy A, models project X; under Policy B, Y, given these assumptions.”

154. Scenario is not prediction certainty

A scenario often asks:

“What happens if these conditions occur?”

It should not be communicated as a single inevitable forecast.

155. Policy uncertainty should be decomposed

Distinguish:

• scientific uncertainty;
• implementation uncertainty;
• behavioural response;
• value disagreement.

---

156. Crisis communication

Crises create pressure for rapid messages before evidence is complete.

Examples:

• outbreaks;
• industrial accidents;
• extreme weather;
• product contamination.

157. Early messages should separate known and unknown

Useful format:

• What we know now
• What we do not yet know
• What we are doing to learn more
• What people should do now, if anything

158. Timestamp crisis information

Fast-changing evidence should be labelled with:

• date;
• time;
• version.

159. Avoid premature reassurance

“There is no risk” is rarely justified early.

Better:

“We have not yet found evidence of X, but investigation is continuing.”

160. Avoid premature catastrophe

A cluster or preliminary signal is not automatically a worst-case scenario.

161. Explain why guidance changes

If recommendations change, say:

• what new evidence appeared;
• how it changed the assessment.

Updating is a feature of science, not inconsistency.

---

162. Communication ethics

Science communication has ethical responsibilities because messages can affect:

• health choices;
• risk perception;
• policy;
• reputation.

163. Accuracy

Do not knowingly communicate a stronger claim than the evidence.

164. Completeness

Completeness does not mean saying everything.

It means not omitting information that changes the likely interpretation.

165. Relevance

Include the information needed for the audience’s task.

166. Transparency

Disclose:

• conflicts;
• uncertainty;
• source.

167. Respect for audience agency

Communication should help people understand evidence rather than covertly manipulate them.

168. Avoid paternalistic certainty

“People cannot handle uncertainty” is often used to justify hiding it.

Better communication explains uncertainty clearly.

169. Avoid strategic ambiguity

Do not use vague scientific language to create plausible deniability.

170. Do not weaponise complexity

Technical detail can be used to overwhelm rather than inform.

Layer detail so readers can inspect it without being buried by it.

---

171. Communicating scientific consensus

Scientific consensus is not a vote that replaces evidence.

It is a description of how strongly a scientific community converges around a conclusion after:

• many studies;
• criticism;
• replication;
• evidence synthesis.

172. Communicate what the consensus is about

Consensus can exist about:

• direction of effect;
• existence of mechanism;
• broad theory.

while disagreement remains about:

• exact magnitude;
• timing;
• policy response.

173. Consensus is not unanimity

A mature field can have a few dissenting scientists.

Communication should represent:

• weight of evidence;
• distribution of expert views;
• nature of remaining disagreement.

174. Avoid false balance

If 95% of relevant evidence supports one conclusion and 5% supports another, giving equal space can misrepresent the scientific state.

175. Avoid consensus-as-authority only

Better:

“Multiple independent evidence streams support X, and the scientific community broadly agrees on that core conclusion.”

This connects consensus to evidence.

176. Consensus communication needs boundaries

Say:

“Scientists broadly agree that human activity is warming the climate.”

Then distinguish uncertainties about:

• regional impacts;
• exact future trajectory under different emissions scenarios.

177. Consensus can change

Scientific consensus is durable when evidence is mature, but it is revisable.

Communication should not imply infallibility.

---

178. Communicating controversy

Scientific controversy can arise from:

• limited data;
• different models;
• measurement disagreement;
• value conflict;
• political/economic interests.

179. Not every public controversy is a scientific controversy

A topic can be publicly controversial even when the core science is relatively settled.

Conversely, scientists can disagree even when the public pays little attention.

180. Name the source of disagreement

Ask:

• Do scientists disagree about evidence?
• Magnitude?
• Mechanism?
• Policy values?

This prevents all disagreement from being labelled “the science is uncertain”.

181. Values should not masquerade as empirical disagreement

Two people can agree on:

“Policy A reduces emissions by X.”

and disagree on:

“Policy A is worth the cost.”

The second is partly a value/economic judgment.

182. Scientific communication should identify the interface

A useful public message can say:

The scientific evidence estimates these consequences; deciding which trade-off society prefers is a policy choice.

183. Controversy increases the need for provenance

In contested topics, link:

• source data;
• review;
• authoritative synthesis.

184. Amplification changes the information environment

Social media and advocacy networks can amplify:

• minority findings;
• outlier studies;
• dramatic anecdotes.

Communication should re-anchor the discussion in the full evidence base.

---

185. Correction architecture

Scientific communication needs a mechanism for updating messages.

Correction type 1 — Typographical

Minor wording/number formatting; scientific conclusion unchanged.

Correction type 2 — Numerical

A reported number changes; magnitude may or may not affect interpretation.

Correction type 3 — Inferential

The conclusion becomes weaker, stronger or more bounded.

Correction type 4 — Retraction

The underlying paper is no longer reliable as published.

186. Corrections should preserve version history

Do not silently replace the old message.

Record:

• what changed;
• when;
• why.

187. Update downstream communication

If a press release, blog, video and social post all repeat the old claim, correct them where feasible.

188. Corrections should be proportional

A minor typo does not need a dramatic apology.

A major conclusion change should not be buried.

189. Correction language

Useful structure:

Earlier we said X. A new analysis/error correction shows Y. This changes/does not change the main conclusion because Z.

190. Retraction communication should separate paper from field

Explain:

• what the retracted paper claimed;
• why it was retracted;
• what independent evidence remains.

---

191. Communicating data stories

A “data story” organises numbers into a sequence.

Good data stories preserve:

• denominator;
• time;
• comparison;
• uncertainty.

192. Start with the question, not the chart

Readers need to know what the visual is trying to answer.

193. Show comparison deliberately

A single number is hard to interpret.

Useful comparisons:

• before/after;
• treatment/control;
• current/historical;
• risk with/without factor.

194. Avoid cherry-picked endpoints

If a time series begins at an unusually low point, trend can look exaggerated.

Show a scientifically relevant window.

195. Cumulative counts versus rates

Cumulative counts almost always rise.

If the scientific question concerns changing risk, rate or incidence may be more informative.

196. Per-capita data

Counts across regions of different population size can mislead.

Use per-capita rates when appropriate.

197. Log scales

Logarithmic axes can clarify exponential change.

Label them clearly because many general audiences assume linear spacing.

198. Dual axes

Dual y-axes can create visually compelling but arbitrary correlations.

Use with extreme caution.

199. Smoothing can imply certainty

A smooth line can look more precise than noisy data warrant.

Show raw points or uncertainty where relevant.

200. Visual uncertainty should not dominate or disappear

Make it visible enough to interpret without making the chart unreadable.

---

201. Science Communication Glossary

Absolute risk

The probability or frequency of an event in a specified population and time period.

Audience analysis

Understanding what an audience knows, needs, values and can use before designing a message.

Baseline risk

The risk before or without the intervention/exposure being discussed.

Consensus

Broad convergence of relevant scientific experts/evidence around a conclusion, not necessarily unanimity.

Deficit model

The overly simple idea that public disagreement with science mainly results from missing scientific information.

Effect size

A quantitative description of how large a difference or relationship is.

Framing

The selection and emphasis of aspects of information in a message.

Natural frequency

A probability expressed as counts such as “10 out of 1,000”.

Plain language

Clear language organised so the intended audience can understand and use the message.

Preprint

A research manuscript publicly shared before formal journal peer review is complete.

Relative risk

The ratio of risk between two groups or conditions.

Risk communication

Communication about the probability, severity and uncertainty of possible outcomes.

Science communication

The process of conveying scientific information, methods, uncertainty and context to an audience for a defined purpose.

Scientific uncertainty

Limits in knowledge arising from sampling, measurement, models, future outcomes or incomplete evidence.

Source provenance

The traceable origin of a scientific claim or data source.

---

202. Advanced FAQ: Scientific Communication

1. What makes science communication scientific?

The message remains traceable to the evidence, preserves uncertainty and scope, and does not exaggerate the conclusion.

2. Is simplifying science dishonest?

No. Simplification is necessary. Distortion occurs when removed information changes the meaning.

3. Should scientists always communicate uncertainty?

Yes when it materially affects interpretation, but uncertainty should be proportional rather than ritualistically exaggerated.

4. Does mentioning uncertainty reduce trust?

Research is context-dependent; clear uncertainty communication can preserve credibility, while effects vary by message and audience. The key is calibrated, interpretable uncertainty rather than vague hedging.

5. Why are absolute risks important?

They show the practical size of a risk change and prevent relative percentages from floating without a baseline.

6. Should we use percentages or frequencies?

Both can work. Natural frequencies often improve comprehension for risk, especially when paired with percentages.

7. What is the best way to communicate a confidence interval?

Give the point estimate, the interval and a plain-language interpretation of the precision without turning it into a posterior probability.

8. Should scientists avoid jargon entirely?

No. Define useful terms and use them consistently.

9. Should every limitation go into a public summary?

No. Include limitations that materially change interpretation; provide access to deeper detail.

10. What is false balance?

Presenting positions as equally supported when the evidence is substantially asymmetric.

11. Should media give dissenters no voice?

Dissent can be relevant, but its evidential weight and representativeness should be clear.

12. How should a preprint be communicated?

Label it as not yet peer reviewed and avoid treating a single preliminary result as settled evidence.

13. How should a failed replication be communicated?

Compare estimates, uncertainty and total evidence rather than using a binary “debunked” label.

14. What is the difference between science communication and scientific communication?

The terms overlap. “Scientific communication” can also refer to communication among researchers; “science communication” often includes public communication.

15. Should scientists advocate policy?

They can, but should distinguish empirical findings from value or policy judgments.

16. Can storytelling mislead?

Yes if one anecdote becomes evidence for population prevalence or causation. Use stories to illustrate, not replace data.

17. What makes a graph misleading?

Scale, denominator, aggregation, visual encoding or omitted uncertainty can distort interpretation.

18. Should all graphs start at zero?

No. It depends on chart type and scientific question. Bar charts often need zero; line charts may reasonably zoom into variation.

19. What is the best scientific headline?

One that remains accurate for the study design, magnitude and scope even if read alone.

20. Why do press releases matter?

They often shape downstream media framing and can amplify or prevent overclaiming.

21. What should a policy maker know first?

The decision-relevant effect, uncertainty, affected population and major trade-offs.

22. What should a patient know first?

Expected benefit, harm, baseline risk and how the evidence applies to them.

23. What should a student know first?

The core concept and evidence chain, before technical detail.

24. How should corrections be handled?

Promptly, visibly, proportionally and with version history.

25. Does a retraction invalidate an entire field?

No. Evaluate how much the field depended on that paper and what independent evidence remains.

26. How do we communicate scientific consensus?

State what is broadly agreed, what evidence supports it and which details remain uncertain.

27. Is consensus proof?

No. It is useful information about the mature evidence state, not a substitute for evidence.

28. How should crisis communication change as evidence changes?

Timestamp messages, distinguish known/unknown and explain why guidance updates.

29. What is the best defence against misinformation?

Accurate, understandable information with strong provenance, plus correction of specific misconceptions when necessary.

30. What is the final science-communication question?

Will the intended audience leave with a mental model of the result that preserves its direction, magnitude, uncertainty, scope and evidential status?

---

203. Testing whether science communication works

A message is not effective merely because the communicator likes it.

Communication effectiveness should be evaluated against a goal.

Possible goals include:

• correct understanding;
• accurate risk perception;
• ability to explain the result;
• ability to make an informed decision;
• ability to distinguish evidence from inference.

204. Comprehension testing

After reading the message, ask:

• What did the study find?
• How large was the effect?
• How certain is it?
• Who does it apply to?
• Did the study prove causation?

If readers answer these incorrectly, the message needs revision.

205. Teach-back

Ask the audience to explain the message in their own words.

This reveals:

• what they understood;
• what they omitted;
• what they overinterpreted.

206. Recognition is weaker than understanding

A reader may recognise the words:

“confidence interval”

without understanding what the interval means.

Test application, not only vocabulary.

207. Recall testing

Immediate comprehension can look good while later memory retains only:

• dramatic number;
• headline;
• emotional image.

Delayed recall tests whether the right scientific elements survive.

208. Memory distortion

Messages can become stronger in memory.

Readers may remember:

“linked to”

as:

“causes”.

Use repeated boundary language where the distinction is critical.

209. Decision-quality testing

If the communication supports a decision, ask whether people can:

• compare options;
• understand benefits/harms;
• identify uncertainty.

210. Behaviour change is not always the right metric

If the goal is neutral informed decision-making, the desired outcome is not necessarily that everyone chooses the same option.

Communication can succeed when people understand the evidence and make different value-consistent decisions.

211. Knowledge gain is only one outcome

Other outcomes include:

• calibrated trust;
• risk comprehension;
• ability to detect misinformation;
• confidence in decision.

212. A/B testing science messages

Communicators can test:

• two headlines;
• two risk formats;
• two visual designs.

Measure comprehension, not only clicks.

213. Engagement can conflict with accuracy

A sensational headline may get more clicks and worse scientific understanding.

Optimising only:

• views;
• likes;
• shares;

can reward distortion.

214. Communication metrics should match the scientific goal

If the goal is:

understanding uncertainty,

then measure:

understanding uncertainty.

Do not use page views as the sole success metric.

---

215. Numeracy

Audiences differ in comfort with:

• percentages;
• fractions;
• probabilities;
• graphs.

Messages should support readers with varied numeracy.

216. Use multiple representations

Example:

“Risk fell from 2% to 1%, or from 20 people per 1,000 to 10 people per 1,000.”

Two representations can reinforce understanding.

217. Avoid denominator neglect

“10 cases” has no scale without population size.

218. Percent versus percentage points

Change from 10% to 8% is:

• 2 percentage points;
• 20% relative reduction.

Do not confuse them.

219. Probability formats

Possible formats:

• 0.05;
• 5%;
• 1 in 20.

Choose one consistent format and provide an equivalent when useful.

220. Very small probabilities

For rare risks, “1 in 100,000” may be easier to grasp than “0.001%”.

221. Large denominators

Keep denominators consistent across comparisons.

Do not compare:

• 1 in 100;
• 50 in 100,000.

without converting.

222. Icon arrays

Icon arrays can show:

• events;
• non-events.

They can be helpful for risk.

Use enough icons to represent the baseline accurately.

223. Logarithmic risk

Risks ranging from 1 in 10 to 1 in 10 million may need logarithmic visualisation.

Explain the scale.

---

224. Cognitive load

Working memory is limited.

A communication can fail because it asks the audience to track:

• too many terms;
• too many groups;
• too many caveats at once.

225. Layer information

Structure:

1. core result;
2. key magnitude;
3. uncertainty;
4. detail on demand.

226. Progressive disclosure

Web pages can offer:

• summary;
• technical details;
• methods;
• data.

This serves audiences with different needs without hiding information.

227. Chunking

Group related information into meaningful units.

Example:

• Benefits;
• Harms;
• Uncertainty.

228. Signposting

Use headings such as:

• What we found;
• What this means;
• What remains uncertain.

This helps readers navigate the inferential structure.

229. Sentence load

A sentence carrying:

• method;
• estimate;
• confidence interval;
• two caveats;

may be grammatically correct and cognitively poor.

Split while preserving logical relationships.

230. Visual and text redundancy

A graph plus a sentence can reinforce the main pattern.

Do not simply duplicate every numeric label unnecessarily.

---

231. Audience mental models

People interpret new scientific information using existing mental models.

A message can fail if the audience maps a new term onto the wrong familiar concept.

232. Example: “theory”

Everyday meaning:

guess.

Scientific meaning:

broad explanatory framework supported by evidence.

Communicators should address the mismatch.

233. Example: “uncertainty”

Everyday interpretation:

scientists do not know.

Scientific meaning can be:

a quantified range around a well-supported estimate.

234. Example: “significant”

Everyday meaning:

important.

Statistical meaning:

related to a test threshold.

Prefer explaining the actual result over relying on the term.

235. Build the right mental model before adding detail

For vaccination risk:

start with:

• baseline risk;
• benefit;
• rare harm.

Then add technical evidence.

236. Misconceptions can block new information

If a reader believes:

“Any uncertainty means no knowledge,”

more uncertainty numbers alone may not help.

Explain how scientific uncertainty works.

---

237. Communicating methods

Methods communication matters because readers need to know:

• why the result deserves confidence;
• which limitations follow from design.

238. Method compression

A public summary may not need every laboratory step.

But it often needs the design label:

• randomized trial;
• observational cohort;
• survey;
• simulation.

239. Explain why design matters

Instead of:

“This was observational.”

add:

“Because researchers did not assign exposure, other differences between groups could contribute to the association.”

240. Sample size communication

Sample size is important but not a quality score.

A huge biased sample can mislead.

241. Randomization communication

Explain:

“Participants were assigned by chance, which reduces systematic baseline differences between groups.”

242. Blinding communication

Explain why blinding matters:

it reduces the chance that expectations influence treatment or outcome assessment.

243. Replication status

Tell audiences whether the finding is:

• first report;
• independently replicated;
• part of a large evidence base.

244. Reproducibility status

For computational studies, communicate whether:

• data;
• code;
• analysis;

are available for reconstruction where relevant.

---

245. Communicating conflicts of interest

Conflicts can include:

• financial;
• professional;
• institutional.

They do not prove the science is wrong.

They are relevant context for evaluating incentives and transparency.

246. Avoid conflict insinuation

Do not say:

“Industry-funded, therefore false.”

Evaluate:

• design;
• data;
• replication.

247. Avoid conflict concealment

Hidden interests can damage trust and prevent informed evaluation.

248. Communicating uncertainty about conflicts

If the role of a sponsor in analysis or publication is unknown, say so rather than speculate.

---

249. Communication feedback loop

Science communication should learn from audience response.

The loop is:

message → audience interpretation → comprehension test → revision.

250. Misunderstanding is data

If many readers misinterpret one sentence in the same way, revise the sentence.

Do not blame the audience reflexively.

251. Questions reveal missing context

Repeated audience questions can show that:

• baseline was missing;
• scope was unclear;
• causation boundary was weak.

252. Communication can be experimentally improved

Science communication itself can use:

• randomized message tests;
• comprehension measures;
• iterative design.

This aligns with the National Academies’ call for a research-based approach to science communication.

---

253. Communication Transformation Workbook

The following transformations show how to preserve scientific meaning while making a message clearer.

Transformation 1 — “X causes Y”

Study: observational association.

Rewrite: “Higher X was associated with higher Y in this study; the design cannot establish causation on its own.”

Transformation 2 — “Risk doubled”

Rewrite: “Risk increased from 1 in 10,000 to 2 in 10,000—a doubling in relative terms but an absolute increase of 1 in 10,000.”

Transformation 3 — “No effect”

Study: wide confidence interval.

Rewrite: “The study did not detect a clear effect and remains compatible with both modest benefit and modest harm.”

Transformation 4 — “Statistically significant improvement”

Rewrite: “Average score improved by 2.3 points; the interval around that estimate was 0.8 to 3.8 points.”

Transformation 5 — “Scientists prove…”

Rewrite: “The study provides evidence that…”

Transformation 6 — “Scientists disagree”

Rewrite: “Scientists broadly agree on X but continue to debate the size or mechanism of Y.”

Transformation 7 — “The result is uncertain”

Rewrite: “The estimated effect is about 10%, but values from roughly 3% to 17% remain compatible with the data.”

Transformation 8 — “The model predicts catastrophe”

Rewrite: “Under Scenario S, the model projects outcome O; other scenarios produce different results.”

Transformation 9 — “The treatment is safe”

Rewrite: “No common serious harms were detected in the trial, but rare harms may require larger or longer surveillance.”

Transformation 10 — “The chemical is toxic”

Rewrite: “The chemical can cause harm at sufficiently high exposure; actual risk depends on dose and route of exposure.”

Transformation 11 — “AI beats doctors”

Rewrite: “On this image benchmark, the model achieved higher accuracy than the clinician comparison group under the study protocol.”

Transformation 12 — “New teaching method works”

Rewrite: “Students using the method scored modestly higher on the tested outcome after the intervention.”

Transformation 13 — “The public does not understand science”

Rewrite: “Audience understanding varies by topic, prior knowledge, values and communication context; information deficit alone does not explain disagreement.”

Transformation 14 — “Experts say…”

Rewrite: identify the review, professional body or evidence synthesis.

Transformation 15 — “The paper was peer reviewed, so it is true”

Rewrite: “Peer review provides quality scrutiny before publication, but findings still require replication and cumulative evaluation.”

Transformation 16 — “The paper was retracted, so the theory is false”

Rewrite: “The paper is no longer reliable; whether the broader claim changes depends on independent evidence.”

Transformation 17 — “There is a 70% chance this happens”

Rewrite: add the time horizon, model and conditions behind the 70% estimate.

Transformation 18 — “This is the biggest study ever”

Rewrite: explain what the larger sample improves and which biases it does not remove.

Transformation 19 — “Scientists are baffled”

Rewrite: describe the actual unresolved question.

Transformation 20 — “Breakthrough”

Rewrite: state the concrete advance: first detection, improved effect, new mechanism, or early proof-of-concept.

---

254. Peer review in science communication

Peer review is a separate scientific process, but communication should represent it accurately.

A paper can be:

• preprint;
• peer reviewed;
• replicated;
• systematically reviewed.

These are different stages of evidence maturity.

255. Peer review is not certification of truth

Peer review can identify:

• method problems;
• missing literature;
• overclaiming.

It cannot guarantee:

• no error;
• replication;
• future validity.

256. Communicate peer-review status where it changes interpretation

Examples:

• preprint during outbreak;
• conference abstract;
• unreviewed company report.

257. Do not weaponise peer review

“Not peer reviewed” does not mean automatically false.

“Peer reviewed” does not mean automatically correct.

Evaluate evidence and maturity.

258. Peer-review boundary

A dedicated peer-review guide should cover:

• review process;
• reviewer selection;
• editorial decisions;
• review biases;
• open peer review.

This communication guide explains only how peer-review status should be conveyed to an audience.

---

259. Open science and communication

Open science can make communication more inspectable by linking to:

• data;
• code;
• protocol;
• pre-registration;
• preprint.

260. More links do not automatically create understanding

Public communication should still provide a coherent summary.

Open artifacts allow deeper inspection for readers who want it.

261. Layered transparency

A useful communication can provide:

1. plain-language summary;
2. technical summary;
3. paper;
4. data/code where available.

262. Data availability statements

Tell readers whether:

• data are public;
• controlled access;
• not shareable and why.

263. Code availability

For computational claims, code availability can strengthen auditability.

But a public repository that does not run is not equivalent to reproducibility.

264. Pre-registration communication

Explain whether:

• hypotheses;
• analysis;
• outcomes;

were specified before results.

This helps readers distinguish confirmatory from exploratory findings.

---

265. Communicating replication status

A result should be described differently depending on whether it is:

• first report;
• same-lab repeat;
• independent direct replication;
• multi-lab replication;
• meta-analytic synthesis.

266. First-study language

Use:

• initial;
• preliminary;
• requires confirmation.

when appropriate.

267. Replicated-study language

Use:

“The finding has been observed in multiple independent studies.”

Then communicate:

• effect-size variation;
• known boundaries.

268. Failed replication language

Avoid:

“The original study is false.”

Prefer:

“The replication estimated a smaller or no clear effect, weakening confidence in the originally reported magnitude.”

269. Replication disagreement is a communication opportunity

Explain why results can differ:

• sampling;
• population;
• method;
• original inflation.

270. Meta-analysis communication

Report:

• pooled effect;
• uncertainty;
• heterogeneity;
• number and quality of studies.

Do not call a pooled result “settled” if heterogeneity is large and unexplained.

---

271. Communication Governance

Institutions that communicate science repeatedly need standards, not ad hoc judgment every time.

Governance can define:

• who approves public claims;
• how uncertainty is represented;
• how corrections are issued;
• how conflicts are disclosed;
• how evidence status is updated.

272. Communication review

High-stakes messages can be reviewed by:

• subject expert;
• statistics/methods expert;
• communication specialist;
• legal/privacy reviewer where necessary.

273. Scientific sign-off

Someone with domain responsibility should confirm:

• core claim;
• magnitude;
• uncertainty;
• scope.

274. Communication sign-off

A communication specialist should check:

• clarity;
• audience fit;
• visual integrity;
• comprehension risks.

275. Legal review should not weaken scientific accuracy

Legal wording can be necessary.

It should not force:

• vague claims;
• buried corrections;
• hidden uncertainty.

276. Release checklist

Before publication:

• claim checked;
• numbers checked;
• links checked;
• visuals checked;
• conflicts checked;
• update owner named.

277. Communication ownership

Who is responsible after publication?

Someone should monitor:

• major new evidence;
• corrections;
• misinterpretation.

278. Communication maintenance

Long-lived pages can become outdated.

Maintain:

• evidence date;
• current consensus;
• links;
• known corrections.

279. Version timestamps

Fast-moving topics should show:

• last reviewed;
• evidence cutoff.

280. Archive old versions

Preserving old messages helps explain:

• how guidance changed;
• what evidence was available then.

---

281. Misinformation Resilience

Good science communication should not only correct one false claim.

It should teach the audience how to resist similar errors.

282. Teach the manipulation pattern

Examples:

• relative risk without baseline;
• cherry-picked study;
• association presented as causation;
• graph with misleading axis.

283. Prebunking

Prebunking introduces common misinformation techniques before people encounter them.

The goal is to build recognition of:

• false dichotomy;
• fake expert;
• conspiracy framing;
• statistical manipulation.

284. Prebunking should not create paranoia

Teach evaluation:

• source;
• evidence;
• reasoning.

Do not teach that every institution or claim is suspicious by default.

285. Source checking

Ask:

• Is the source identifiable?
• Does it link the underlying evidence?
• Is the evidence current?
• Does the headline match the source?

286. Lateral reading

Instead of staying on one page, check:

• what independent sources say;
• whether the institution is legitimate;
• whether the claim appears in primary evidence.

287. Screenshot skepticism

A screenshot can remove:

• date;
• caption;
• source;
• correction.

Find the original.

288. Quote verification

A sentence can be quoted accurately but stripped of:

• qualification;
• population;
• uncertainty.

Read surrounding context for important scientific claims.

289. Misleading truth

A claim can use true numbers and still mislead through:

• denominator choice;
• time window;
• omitted comparison.

Scientific literacy includes contextualising true statements.

290. Correction fatigue

Repeating every false claim can overwhelm audiences.

Prioritise:

• high-impact;
• widely circulating;
• decision-relevant misinformation.

---

291. Communicating with uncertainty during disagreement

When studies disagree, avoid announcing:

“Science is confused.”

Explain:

• which results differ;
• how designs differ;
• which study is more precise;
• whether heterogeneity is expected.

292. Disagreement can narrow the question

Example:

Effect exists in Population A but not B.

Communication should shift from:

“Does X work?”

to:

“For whom and under which conditions does X work?”

293. Conflicting results should be weighted

Do not present:

• one small study;
• one large replication;

as equivalent merely because they point in opposite directions.

294. Science updates can be communicated as refinement

Old:

“Large effect.”

New:

“Smaller but more stable effect.”

This is progress, not contradiction.

295. Boundary discovery is communication-worthy

Public messages often want universal conclusions.

But discovering:

“works only under Condition C”

can be the scientifically important result.

---

296. Communicating scientific models

Models are simplified representations.

A model output should be communicated with:

• inputs;
• assumptions;
• validation;
• uncertainty.

297. “The model says” is incomplete

Models do not speak independently of:

• data;
• parameters;
• scenario choices.

298. Scenario models

Use:

“If X, Y and Z occur, the model projects…”

rather than:

“The model predicts exactly…”

299. Model ensembles

Multiple models can produce a range.

Communicate whether the range represents:

• model variation;
• scenario variation;
• statistical uncertainty.

300. Model validation

Tell readers:

• what observations the model has reproduced;
• where it performs poorly.

301. Model disagreement

Different models can agree on direction while differ in magnitude.

Communicate the common conclusion and remaining spread.

---

302. Scientific uncertainty language ladder

Communicators can match language to evidence.

Very limited evidence

“A preliminary result suggests…”

Moderate evidence

“Several studies support…”

Strong evidence

“Multiple independent studies consistently show…”

Very strong mature evidence

“The evidence strongly establishes the core relationship within these boundaries…”

These phrases are illustrative, not universal formal grades.

303. Do not use hedges as decoration

Words such as:

• may;
• could;
• possibly;

should reflect actual uncertainty, not automatic academic style.

304. Do not remove justified confidence

If evidence is mature, excessive hedging can create false doubt.

305. Uncertainty should answer a question

Tell the audience:

• what could vary;
• by how much;
• whether it changes the decision.

---

306. Communicating sample and population

Always distinguish:

• who was studied;
• who the conclusion is meant to apply to.

307. “Adults” is not one population

Relevant differences can include:

• age;
• country;
• health status;
• socioeconomic context.

308. Generalisation should be earned

Do not write:

“Humans prefer X”

from one university student sample.

309. Communicate subgroup evidence carefully

Subgroup findings can be exploratory and unstable.

State whether the subgroup was:

• pre-specified;
• replicated.

310. Average effect can hide heterogeneity

If an intervention helps one group and not another, communicate both the average and the variation.

---

311. Science communication by field: medicine

Medical communication must preserve:

• baseline risk;
• benefit;
• harm;
• time horizon;
• uncertainty;
• patient applicability.

Patients often need decision-relevant quantities more than abstract statistical significance.

312. Medicine: number formats

Useful formats include:

• absolute risk;
• relative risk;
• natural frequencies;
• number needed to treat or harm when appropriate.

313. Medicine: diagnosis communication

A test result should not be presented as certainty without considering:

• prevalence;
• sensitivity;
• specificity;
• follow-up testing.

314. Medicine: informed consent

Communication should enable understanding of:

• benefits;
• harms;
• alternatives;
• uncertainty.

The goal is informed choice rather than mere compliance.

---

315. Science communication by field: public health

Public-health communication must often address populations under uncertainty.

Messages may need to balance:

• individual risk;
• population benefit;
• rapidly changing evidence.

316. Public health: incidence versus prevalence

Communicators should distinguish:

• new cases over time;
• total existing cases.

317. Public health: rate versus count

A rise in counts can reflect population growth or testing changes.

Rates and denominators matter.

318. Public health: surveillance signals

A signal should be communicated as something requiring investigation, not automatically as a confirmed causal effect.

---

319. Science communication by field: climate

Climate communication often involves:

• long-term trends;
• model projections;
• scenario dependence;
• regional uncertainty.

320. Climate: weather versus climate

One weather event is not the same as climate trend.

Communication should distinguish event, long-term distribution and attribution.

321. Climate: scenario language

Use:

“Under higher-emissions Scenario A…”

not:

“This exact future will happen.”

322. Climate: uncertainty does not imply no direction

Models can strongly agree on direction and broad mechanism while differing on magnitude or regional detail.

---

323. Science communication by field: ecology

Ecological communication should preserve:

• geographic scope;
• taxonomic scope;
• sampling effort;
• trend duration.

324. Ecology: abundance versus detection

Fewer sightings can mean fewer animals, lower detectability or less survey effort.

Communicate the distinction.

325. Ecology: biodiversity headlines

A decline in monitored populations should not automatically be communicated as:

“X% of all animals disappeared.”

Metric definitions matter.

---

326. Science communication by field: physics and astronomy

Communicators often face extreme scales, probabilistic detections and indirect evidence.

327. Detection significance

A “5-sigma” threshold has a technical meaning in a statistical context.

It should not be translated simply as:

“100% certain.”

328. Candidate versus confirmed detection

Use these terms accurately for exoplanets, particles and astronomical events.

329. Scale analogies

Analogies can help explain distance, energy and mass.

Do not imply the analogy is a literal mechanism.

---

330. Science communication by field: engineering

Engineering communication often needs failure probability, safety margin, operating conditions and uncertainty.

331. Reliability language

“99.9% reliable” is incomplete without time, load and failure definition.

332. Safety communication

Separate hazard, probability and mitigation.

333. Root-cause communication

An engineering incident report should distinguish observed failure, proximate cause, contributing factors and root cause.

---

334. Science communication by field: education

Education communication should distinguish test performance, learning, retention and transfer.

335. Avoid “months of learning” without method context

Some effect-size translations depend on assumptions about test scaling and comparison groups.

Explain the basis.

336. School comparisons

Average scores reflect students, selection, resources and instruction.

Do not communicate rank as pure school quality automatically.

337. Tuition communication integrity

For educational services, avoid guaranteed outcomes, unverified rankings and anecdotes presented as representative evidence.

Describe teaching model, observed progress and boundaries accurately.

---

338. Science communication by field: AI and machine learning

AI communication is especially vulnerable to anthropomorphic language, benchmark overgeneralisation, rapid model change and demo selection.

339. Benchmark versus capability

A benchmark score supports a claim about performance on that benchmark under that protocol.

Broader capability requires more evidence.

340. “Understands” and “thinks”

Anthropomorphic verbs can be useful shorthand.

They can also imply mental mechanisms not established by the evaluation.

341. Demo selection

One impressive output or one failure is an anecdote.

Communicate distributions across tasks, runs and prompts.

342. Model version

Hosted AI systems change.

Communicate model/version, evaluation date and prompt protocol.

343. Error rates need definitions

“Hallucination rate” can depend on what counts as an error, the unit of analysis and benchmark.

Define the metric.

---

344. Science communication lifecycle

Stage 1 — Scientific result

Identify the actual conclusion from the study or synthesis.

Stage 2 — Goal

Decide what the communication should accomplish.

Stage 3 — Audience

Identify prior knowledge, needs and context.

Stage 4 — Core scientific message

Preserve direction, magnitude, uncertainty and scope.

Stage 5 — Format

Choose article, visual, talk, press release, video or another channel.

Stage 6 — Integrity review

Check against the source evidence.

Stage 7 — Comprehension test

Ask what the audience understood.

Stage 8 — Release

Publish with provenance and date.

Stage 9 — Monitor interpretation

Look for systematic misunderstanding or misuse.

Stage 10 — Correct/update

Revise when the science or message changes.

345. Communication is a maintained scientific artifact

A public explanation can become outdated just as software, guidelines and models can.

Maintenance is part of scientific integrity.

---

346. Communication review protocol

Review a scientific message in seven passes.

Pass 1 — Source

Does every important claim trace to evidence?

Pass 2 — Design

Does language match experimental or observational status?

Pass 3 — Magnitude

Is the size of the effect visible?

Pass 4 — Uncertainty

Is interpretation-changing uncertainty preserved?

Pass 5 — Scope

Who or what does the finding apply to?

Pass 6 — Audience

Can intended readers understand and use it?

Pass 7 — Distortion test

Could a reasonable reader leave with a stronger or different claim than the science?

347. Distortion test examples

If a reader leaves believing causation when only association exists, certainty when evidence is preliminary, or large benefit when absolute effect is tiny, the communication has failed scientifically.

---

348. Scientific communication versus marketing

Science communication and marketing can use similar tools:

• headlines;
• visuals;
• stories;
• audience segmentation.

The scientific obligation is different.

Marketing often optimises for:

• attention;
• conversion;
• preference.

Scientific communication should optimise for:

• accurate understanding;
• calibrated confidence;
• informed decision.

349. Institutional promotion creates pressure

Universities, journals and companies benefit from:

• coverage;
• citations;
• reputation.

This can encourage:

• “first ever” language;
• breakthrough framing;
• overgeneralised significance.

350. Promotion should not outrun evidence

An institution can celebrate:

• technical advance;
• publication;
• collaboration.

without inflating the scientific conclusion.

351. Commercial science claims

If communication supports a product, distinguish:

• published evidence;
• company interpretation;
• marketing claim.

352. Testimonials are not controlled evidence

A positive user story can illustrate experience.

It should not be used to estimate:

• average benefit;
• success rate;
• causal effect.

353. “Clinically tested” is vague

Useful communication specifies:

• what study;
• what population;
• what outcome;
• what result.

---

354. Cross-language science communication

Translating science is not merely replacing words.

Scientific concepts can lack direct equivalents across languages.

355. Preserve technical meaning across languages

Terms such as:

• risk;
• significance;
• theory;
• uncertainty;

may carry different everyday meanings.

Translate the concept, not only the term.

356. Back-translation

For high-stakes messages, translate into the target language and independently translate back.

Differences can reveal meaning drift.

357. Cultural examples

Examples should be familiar enough to help but not introduce a false analogy.

358. Numeracy differs across cultures and educational contexts

Test whether:

• percentages;
• fractions;
• icon arrays;

are understood similarly.

359. Translation should preserve uncertainty

A source phrase such as:

“evidence suggests”

should not become:

“science proves”

in another language.

---

360. Accessibility

Science communication should be usable by people with different:

• vision;
• hearing;
• reading ability;
• cognitive needs.

361. Alt text for scientific images

Alt text should describe the scientific message of a graph or image, not only its appearance.

Example:

“Line graph showing treatment group consistently lower than control from weeks 2–8, with overlapping uncertainty at week 1.”

362. Captions for audio/video

Captions improve accessibility and can increase comprehension for everyone.

363. Colour-blind accessibility

Do not encode groups by colour alone.

Use:

• labels;
• line styles;
• symbols.

364. Screen-reader-friendly tables

Use logical headers and avoid complex merged cells where possible.

365. Reading level

Shorter sentences and defined terms improve accessibility without reducing scientific sophistication.

366. Cognitive accessibility

Use:

• clear hierarchy;
• chunking;
• summary before detail.

---

367. Communication Assurance Standard

A high-stakes science message should pass six assurance checks.

Assurance 1 — Scientific fidelity

Does the message match the source conclusion?

Assurance 2 — Magnitude fidelity

Can the audience tell how large the effect is?

Assurance 3 — Uncertainty fidelity

Can the audience tell how uncertain the estimate is?

Assurance 4 — Scope fidelity

Can the audience tell who/what the finding applies to?

Assurance 5 — Status fidelity

Is the evidence stage clear: preliminary, peer reviewed, replicated, synthesised?

Assurance 6 — Audience comprehension

Does testing show the intended audience understood the message?

368. Communication Assurance Record

Field	Record
Source	paper/report/guideline
Evidence status	preprint/peer reviewed/replicated
Core claim	one sentence
Magnitude	absolute/relative effect
Uncertainty	interval/range/qualifier
Scope	population/context
Audience	target group
Goal	inform/explain/decision support
Review	scientific + communication sign-off
Update owner	responsible team/person

369. Communication assurance is not censorship

The goal is not to make every message sound identical.

The goal is to ensure:

• creative form;
• accurate scientific core.

---

370. Science Communication Audit: 30 questions

1. What is the source evidence?
2. What is the communication goal?
3. Who is the intended audience?
4. What is the one-sentence core claim?
5. Does it match the study design?
6. Is the magnitude visible?
7. Is baseline risk visible where relevant?
8. Is uncertainty visible?
9. Is the population/scope visible?
10. Is preliminary status visible?
11. Is replication status visible?
12. Are serious limitations near the claim?
13. Does the headline overclaim?
14. Do quotes overclaim?
15. Are visuals honest?
16. Are axes/scales clear?
17. Are denominators consistent?
18. Are conflicts disclosed?
19. Are science and policy judgment separated?
20. Are values disguised as evidence?
21. Is jargon necessary or defined?
22. Is the message accessible?
23. Can the audience explain it back accurately?
24. Does the message invite a predictable misconception?
25. Is the source linked or identifiable?
26. Is the evidence current?
27. Is the correction pathway clear?
28. Who owns future updates?
29. What would change the message?
30. Would a reasonable reader leave with the same scientific confidence as the evidence supports?

371. Audit result categories

High-fidelity

Claim, magnitude, uncertainty, scope and evidence status are all preserved.

Accurate but incomplete

No false statements, but missing context may distort interpretation.

Potentially misleading

Framing, omissions or visuals push readers toward a stronger/different conclusion.

Scientifically inaccurate

The communicated claim conflicts with the evidence.

Outdated

The message was once reasonable but no longer reflects current evidence.

---

372. Twenty-five myths about science communication

Myth 1 — If people know more facts, they will automatically agree

False. People interpret science through goals, values, trust and social context as well as knowledge.

Myth 2 — Simplification always damages accuracy

False. Good simplification removes unnecessary complexity while preserving scientific meaning.

Myth 3 — Technical language is always more precise

False. Jargon can hide ambiguity when the audience does not share the definition.

Myth 4 — Plain language means childish language

False. Plain language can communicate sophisticated science clearly.

Myth 5 — Uncertainty makes people distrust science automatically

False. Effects depend on context and message; calibrated uncertainty can support durable credibility.

Myth 6 — Hiding uncertainty makes messages clearer

False. It can create false certainty and make later updates look like reversals.

Myth 7 — More caveats always make communication more honest

False. Excessive caveats can obscure the main evidence and create false doubt.

Myth 8 — A statistically significant result is automatically newsworthy

False. Magnitude, novelty, design and relevance matter.

Myth 9 — A large relative risk means a large practical effect

False. Baseline risk determines absolute magnitude.

Myth 10 — A graph is objective because it uses data

False. Axis, denominator, aggregation and colour can shape interpretation.

Myth 11 — A press release can be stronger than the paper because it is promotional

False. Promotion does not change evidential status.

Myth 12 — Peer review means the result is true

False. Peer review is one quality-control stage.

Myth 13 — A preprint is unreliable by definition

False. It is simply earlier in the review process and should be communicated as such.

Myth 14 — A failed replication automatically debunks the original

False. Compare effect estimates and total evidence.

Myth 15 — Consensus means unanimity

False. Consensus can exist alongside bounded disagreement.

Myth 16 — Dissent deserves equal airtime

Not necessarily. Representation should reflect evidence weight and relevance.

Myth 17 — Scientists should never advocate

Scientists can advocate, but should distinguish evidence from values and policy judgments.

Myth 18 — Storytelling is unscientific

False. Stories can aid understanding if they illustrate rather than substitute for evidence.

Myth 19 — Visuals are only decoration

False. Visual encoding can change understanding of magnitude and uncertainty.

Myth 20 — Social media cannot communicate serious science

False. It can, but severe compression requires careful provenance and boundaries.

Myth 21 — Corrections reduce trust

Not necessarily. Transparent correction can demonstrate integrity.

Myth 22 — The public cannot understand uncertainty

False. Uncertainty can be communicated clearly with appropriate formats and explanation.

Myth 23 — One communication format works for everyone

False. Goals, prior knowledge and context differ.

Myth 24 — Attention equals communication success

False. Viral content can be scientifically misunderstood.

Myth 25 — Science communication is downstream from science and cannot affect it

False. Communication can expose unclear inference, missing context and new research questions.

---

373. Science Communication Quality Matrix

Dimension	Weak	Strong
Goal	hidden	explicit
Audience	generic	analysed
Claim	overstated	source-faithful
Magnitude	omitted	interpretable
Uncertainty	hidden/exaggerated	calibrated
Scope	universalised	bounded
Visuals	decorative/distorting	evidence-aligned
Provenance	“scientists say”	traceable source
Accessibility	barriers ignored	multiple access needs considered
Maintenance	static	versioned/update owner

The matrix is diagnostic, not a universal score.

374. Communication bottleneck principle

The message is only as scientifically strong as its weakest essential component.

A perfect graph cannot rescue:

• wrong causal language;
• missing baseline;
• outdated evidence.

---

375. Communication Maintenance Standard

Long-lived scientific communication should be reviewed when:

• major new study appears;
• replication changes effect magnitude;
• guideline changes;
• source is corrected or retracted.

376. Review frequency should match evidence speed

Stable physics concepts may need little updating.

Fast-moving outbreak or AI pages may need frequent review.

377. Evidence cutoff

State:

“Evidence reviewed through September 2026.”

This makes staleness visible.

378. Update owner

Assign responsibility to:

• person;
• team;
• institution.

379. Update triggers

Examples:

• meta-analysis changes direction;
• major safety alert;
• consensus statement changes;
• study retracted.

380. Update the main message, not only a footnote

If evidence changes the conclusion, the visible headline and summary should change too.

381. Preserve version history

Public readers may ask:

“Why did the advice change?”

Version history provides the answer.

---

382. Communication Incident Response

If a scientific message is found to be materially misleading:

1. freeze/archive the original version;
2. identify the error;
3. assess scientific impact;
4. correct the message;
5. update downstream channels;
6. record the change.

383. Communication incident categories

Minor

Typographical or small numeric issue; interpretation unchanged.

Moderate

Important context omitted; likely misunderstanding.

Major

Causal, magnitude or uncertainty claim materially wrong.

Critical

Message could drive harmful decisions or fundamentally misrepresent evidence.

384. Root-cause analysis for communication failure

Ask whether failure came from:

• source misunderstanding;
• headline editing;
• visual design;
• institutional promotion;
• audience testing failure;
• evidence update lag.

385. Preventive action

Examples:

• new review checklist;
• mandatory baseline risk;
• headline approval;
• scheduled evidence review.

---

386. Science Communication Handoff Standard

A mature science message should be handed to future communicators with:

• source evidence;
• core claim;
• magnitude;
• uncertainty;
• scope;
• audience;
• goal;
• visual assets;
• version history;
• update triggers.

387. Why handoff matters

Without this structure, later communicators can inherit:

• headline;
• graphic;

without knowing why the wording was bounded.

388. Communication lineage

The ideal route is:

study → scientific conclusion → institutional summary → media/public message.

Each transformation should preserve claim strength.

389. Claim drift

At each handoff, claims can become:

• shorter;
• stronger;
• broader.

Audit the lineage.

---

390. Communication scope and boundaries

This guide focuses on how scientific results, uncertainty and context are conveyed without distorting scientific meaning.

Related scientific-method topics are treated separately:

• Observation — what was recorded.
• Measurement — quantified values and uncertainty.
• Evidence — how observations support claims.
• Inference — reasoning to a conclusion.
• Peer review — formal expert review of scientific work.
• Consensus — how communities converge on conclusions.
• Replication — new-data re-testing.
• Reproducibility — computational reconstruction.

The durable communication question is:

How can we help this audience understand the scientific result without changing its magnitude, uncertainty, scope or evidential status?

391. Where this page sits in the eduKateYishun scientific-method system

• How Scientific Observation Works — source observations.
• How Scientific Measurement Works — quantities and uncertainty.
• How Scientific Evidence Works — evidential support.
• How Scientific Inference Works — conclusions.
• How Scientific Uncertainty Works — uncertainty architecture.
• How Scientific Replication Works — new-data confirmation.
• How Scientific Reproducibility Works — computational reconstruction.
• When Science Results Disagree — conflicting evidence and next tests.

392. Authoritative further reading

• National Academies — Communicating Science Effectively: A Research Agenda
• National Academies — The Complexities of Communicating Science
• National Academies — Building the Knowledge Base for Effective Science Communication
• Open research review — Communicating uncertainty about facts, numbers and science

393. Science Communication Contract

A serious scientific message makes a contract with its audience:

I will preserve the result’s direction, magnitude, uncertainty, scope and evidence status. I will adapt language and format to help you understand, but I will not simplify by turning association into causation, uncertainty into certainty, or a local result into a universal claim.

---

394. Science Communication Decision Manual

Decision 1 — What is the communication goal?

Inform, explain, support a decision, correct a misconception or advocate?

Decision 2 — What is the scientific core?

Write the strongest one-sentence conclusion the evidence supports before adapting it for any audience.

Decision 3 — What is the evidence stage?

Preprint, peer reviewed, replicated, meta-analysed or consensus-level?

Decision 4 — What magnitude matters?

Choose the quantity that lets the audience understand practical size.

Decision 5 — Which uncertainty changes interpretation?

Do not include every technical uncertainty equally. Prioritise what affects the conclusion or decision.

Decision 6 — What is the scope?

Name the population, setting, time horizon or tested range.

Decision 7 — Does the design support causation?

If not, remove causal verbs.

Decision 8 — Does relative risk need an absolute baseline?

Usually yes for public risk communication.

Decision 9 — Which format fits the task?

Graph for pattern, table for exact values, prose for explanation, diagram for mechanism.

Decision 10 — Does the audience need the technical term?

If yes, define it. If not, use plain language.

Decision 11 — Is an analogy helpful?

Use only if it clarifies more than it distorts.

Decision 12 — Is the headline independently accurate?

Test it without the body text.

Decision 13 — Does the visual preserve scale?

Check axes, denominator, aggregation and uncertainty.

Decision 14 — Is the study being placed in context?

One new paper rarely stands alone.

Decision 15 — Is disagreement scientific or value-based?

Communicate the distinction.

Decision 16 — Is the message current?

Check whether important evidence has changed since source publication.

Decision 17 — What misconception is predictable?

Pre-empt it near the main claim.

Decision 18 — Can the audience act on this message?

If the goal is decision support, include the quantities and trade-offs needed.

Decision 19 — How will comprehension be tested?

Use teach-back, survey questions or message testing.

Decision 20 — Who will update the message?

Assign ownership before release.

---

395. Communication Integrity Rules

Rule 1 — Never strengthen the claim for the headline

Rule 2 — Never hide baseline risk when relative risk is prominent

Rule 3 — Never turn uncertainty into ignorance

Rule 4 — Never turn consensus into certainty

Rule 5 — Never treat one study as the whole field

Rule 6 — Never use an anecdote as a prevalence estimate

Rule 7 — Never use a chart design that changes perceived magnitude unnecessarily

Rule 8 — Never let promotional goals change the evidential status

Rule 9 — Never bury a correction that changes the main conclusion

Rule 10 — Never treat the audience as a passive container for facts

These rules preserve the scientific core across channels.

---

396. AI-assisted science communication

AI can help:

• summarise papers;
• translate technical language;
• draft captions;
• generate audience variants;
• test reading level.

AI also creates new distortion risks.

397. AI summarisation can compress away caveats

A model may reduce:

“association under these conditions with substantial uncertainty”

to:

“X causes Y.”

Human scientific review remains necessary.

398. AI can invent provenance

Generated summaries can fabricate:

• citations;
• sample sizes;
• effect estimates.

Every scientific detail should be checked against the source.

399. AI can amplify confident tone

Language models often produce fluent declarative prose.

Communicators should deliberately restore:

• uncertainty;
• scope;
• evidence status.

400. AI translation needs scientific review

Automated translation can alter:

• probability language;
• technical terms;
• negation.

High-stakes translations should be reviewed by a scientifically competent speaker.

401. AI-generated visuals need provenance

Illustrative images can help explain concepts.

They should not be presented as:

• microscope images;
• satellite observations;
• scientific data.

unless they actually are.

402. AI should not generate scientific numbers from prose memory

For effect sizes, confidence intervals or sample counts, pull from the verified source rather than allowing a model to guess.

403. AI-assisted audience testing

AI can simulate possible misunderstandings.

But real audience testing remains stronger because actual readers bring real context and beliefs.

404. AI communication logs

For important outputs, preserve:

• source document;
• prompt/instruction;
• model/version;
• human edits.

This supports audit.

---

405. Scientific Communication Verification Standard

A message can be verified by reconstructing its path from source evidence to public wording.

Verification 1 — Source identity

Can the underlying paper, review or dataset be identified?

Verification 2 — Claim identity

Does the core public claim match the source conclusion?

Verification 3 — Magnitude identity

Are effect size and baseline preserved?

Verification 4 — Uncertainty identity

Is uncertainty represented proportionally?

Verification 5 — Scope identity

Is the population or context preserved?

Verification 6 — Status identity

Is preprint/peer-review/replication status accurate?

Verification 7 — Visual identity

Do visuals represent the same quantities without distortion?

Verification 8 — Audience interpretation

Do readers understand the intended scientific model?

406. Verification is stronger than proofreading

Proofreading catches:

• grammar;
• typos.

Scientific verification catches:

• causal drift;
• magnitude drift;
• scope drift;
• uncertainty drift.

407. Independent message review

For high-stakes communication, ask a scientist who did not draft the message:

If you read only this communication, would you infer the same scientific conclusion as from the source evidence?

---

408. Communication Certification Levels

Level 1 — Source-faithful

The message does not contradict the underlying science.

Level 2 — Context-complete

Magnitude, uncertainty and scope are preserved.

Level 3 — Audience-tested

The intended audience demonstrably understands the core message.

Level 4 — Decision-ready

The message supports the audience’s real decision with comparable benefits, harms and uncertainty.

Level 5 — Maintained

The message has version history, review date and update ownership.

These levels are descriptive, not a formal universal certification scheme.

409. Communication certification record

A record can include:

• source;
• evidence cutoff;
• audience;
• goal;
• reviewer;
• comprehension test;
• version;
• next review date.

---

410. Science Communication Stop Rule

A message is ready to release when:

• core claim is accurate;
• magnitude is interpretable;
• uncertainty is proportional;
• scope is explicit;
• audience testing finds no major predictable misinterpretation;
• source and update path are clear.

411. Science Communication Reopen Rule

Reopen the message when:

• new evidence changes the effect;
• guidelines change;
• source is corrected/retracted;
• audience testing reveals systematic misunderstanding.

412. Communication never becomes permanently finished

Long-lived messages are part of scientific infrastructure and should evolve with the evidence.

---

413. Communicating systematic reviews and evidence syntheses

Systematic reviews and meta-analyses are often treated as the highest form of evidence communication.

They can be powerful because they organise many studies.

They still require careful communication.

414. State what was synthesised

Readers should know:

• how many studies;
• which populations;
• which outcomes;
• which study designs.

415. Pooled effect is not the only result

Also communicate:

• heterogeneity;
• risk of bias;
• study quality;
• publication bias.

416. “Meta-analysis proves” is too strong

Meta-analysis estimates a combined effect under a model.

It does not erase flaws in the studies being combined.

417. Systematic review without meta-analysis

Sometimes studies are too different to pool meaningfully.

A narrative or structured synthesis can be more scientifically honest than one forced number.

418. Communicate heterogeneity as knowledge

Instead of:

“Studies are inconsistent,”

say:

“Effects are larger in Population A and smaller in B, suggesting a possible moderator.”

419. Certainty of evidence

Some evidence frameworks classify certainty or confidence.

If used, explain what the category means and which factors lowered or raised confidence.

---

420. Citation and source communication

Scientific communication should make it possible to find the underlying source.

421. Primary versus secondary source

Primary paper:

original study.

Secondary source:

review, guideline, news article.

For a new result, link the primary source when possible.

422. Reviews can be better sources for mature questions

If the question is:

“What does the field know overall?”

a systematic review or authoritative synthesis can be more useful than one paper.

423. Cite the source that supports the claim

Do not cite a general page for a precise numeric result if the number comes from one study.

424. Avoid citation laundering

A blog cites a news story, which cites a press release, which paraphrases a paper.

Trace important claims back to the scientific source.

425. Citation count is not scientific strength

A famous paper can be:

• wrong;
• outdated;
• superseded.

426. Date matters

For fast-moving science, old sources can represent an outdated evidence state.

Communicate:

• publication date;
• review date.

---

427. Communicating null and negative findings

Null findings deserve careful language.

428. “No evidence of effect”

This can mean:

• no clear signal;
• insufficient precision.

429. “Evidence of no meaningful effect”

This stronger statement requires a study capable of excluding effects large enough to matter.

430. Communicate precision with nulls

Example:

“The estimate was near zero, and the interval excludes improvements larger than 2 points.”

This is much more informative than:

“No significant difference.”

431. Negative findings are still findings

Communication systems that reward only positive results contribute to publication bias and public misunderstanding.

432. Failed hypothesis communication

A failed prediction can be framed as:

“The study did not support the proposed mechanism under these conditions.”

That is scientifically valuable.

---

433. Communicating effect heterogeneity

Average effects can hide variation.

A public message should mention heterogeneity when it changes decisions.

434. Subgroup effect caution

Ask:

• Was subgroup pre-specified?
• Was interaction statistically supported?
• Was it replicated?

435. Avoid deterministic subgroup labels

“Group A responds”

can hide:

• overlap;
• uncertainty;
• individual variation.

436. Personalized risk communication

When a risk estimate is individualized, state:

• which variables drive it;
• uncertainty;
• validation population.

437. Fairness and subgroup communication

Group-level performance differences should be communicated carefully to avoid:

• stereotyping;
• individual determinism.

Describe the data and system, not intrinsic worth.

---

438. Communicating mechanisms

Mechanisms are powerful because they make results understandable.

They are also tempting places to overclaim.

439. Established versus proposed mechanism

Use language such as:

• established pathway;
• candidate mechanism;
• one possible explanation.

440. Diagrams can imply certainty

A clean arrow:

A → B → C

can look proven even when the mechanism is hypothetical.

Use:

• dashed lines;
• labels;
• uncertainty notes.

441. Animation can imply causation

An animated molecule binding and “causing” an outcome may represent a model rather than direct observation.

Label it as an illustration or model.

442. Mechanistic stories should remain tied to evidence

Humans like coherent stories.

A plausible story can feel truer than the data justify.

443. Explain what part is known and what part is inferred

This preserves the observation–inference–explanation structure.

---

444. Communicating scientific disagreement to students

Students should learn that disagreement can arise from:

• different data;
• different methods;
• different interpretations.

445. Avoid “scientists keep changing their minds” framing

Better:

“Scientific conclusions update as evidence improves.”

446. Use version history as teaching

Show:

• earlier hypothesis;
• new evidence;
• revised model.

This teaches science as self-correcting.

447. Communicate consensus and uncertainty together

Example:

“Scientists strongly agree on the core mechanism, while estimates of exact magnitude still vary.”

This is more accurate than either:

• “everyone agrees on everything”;
• “scientists are divided”.

---

448. Communication Memory Standard

Important scientific messages should preserve:

• what was said;
• what evidence existed;
• what uncertainty was known.

449. Communication archives

Archives help researchers and the public understand:

• why guidance changed;
• whether earlier communication was reasonable at the time.

450. Do not rewrite history silently

If a page changes materially, preserve the update note.

451. Communication memory supports trust

Visible updates demonstrate:

• evidence changed;
• message changed accordingly.

452. Scientific memory should preserve errors too

Old incorrect messages can teach:

• how overclaiming occurred;
• how review can improve.

---

453. Communication Handoff to Peer Review

Peer review evaluates scientific work before or around publication.

Communication can benefit from peer-review findings but should not absorb the review process.

A dedicated peer-review guide should answer:

• how reviewers evaluate;
• how editors decide;
• how review fails;
• how open review works.

454. Communication Handoff to Consensus

A dedicated consensus guide should answer:

• how scientific communities converge;
• how dissent is evaluated;
• how consensus shifts.

This communication guide addresses how that consensus should be represented accurately to an audience.

---

455. Communication Validation Appendix

A science message can be tested just as a scientific instrument can be tested. The goal is to determine whether the message actually produces the intended scientific understanding.

Validation 1 — Claim comprehension

Ask readers to state the main conclusion.

A message fails if readers consistently convert:

• association into causation;
• small effect into large effect;
• uncertainty into certainty.

Validation 2 — Magnitude comprehension

Ask:

“How large was the effect?”

If readers remember only “significant”, the message failed to communicate magnitude.

Validation 3 — Uncertainty comprehension

Ask:

• What is known?
• What remains uncertain?
• Does the uncertainty change the decision?

Validation 4 — Scope comprehension

Ask who or what the result applies to.

If a study of 60-year-old patients is remembered as applying to everyone, scope failed.

Validation 5 — Status comprehension

Ask whether readers know if the finding is:

• preliminary;
• peer reviewed;
• replicated;
• consensus-level.

Validation 6 — Decision comprehension

For decision-support communication, ask whether people can compare:

• benefits;
• harms;
• uncertainty.

456. Communication Validation Matrix

Target	Test	Failure signal
Claim	teach-back	causal/magnitude drift
Magnitude	numeric recall	only significance remembered
Uncertainty	range interpretation	certainty or ignorance distortion
Scope	population question	overgeneralisation
Status	evidence-stage question	preliminary treated as settled
Decision	option comparison	benefit/harm imbalance

457. Validation should use the real audience

Scientists and communication professionals can predict misunderstanding incorrectly.

Test with:

• patients;
• parents;
• students;
• policy staff;

depending on the target audience.

458. Expert comprehension is not public comprehension

A sentence obvious to specialists can be ambiguous to non-specialists.

459. Audience misunderstanding can reveal scientific ambiguity

Sometimes a message is hard to explain because the underlying scientific claim is itself poorly defined.

Communication testing can therefore improve the science.

---

460. Scientific Decision Aids

Decision aids organise evidence so people can compare options.

They are useful in:

• medicine;
• environmental risk;
• technology choices;
• education pathways.

461. Decision aid core

For each option, present:

• expected benefit;
• expected harm;
• uncertainty;
• time horizon;
• who the evidence applies to.

462. Use comparable formats

If Option A benefit is shown as percentage and Option B harm as “1 in 1,000”, readers must mentally convert.

Use consistent denominators.

463. Values clarification

Decision aids can ask:

• Which outcome matters most to you?
• How do you value short-term harm versus long-term benefit?

This separates scientific evidence from personal values.

464. Decision aids should not hide the default

Interface design can nudge people strongly.

If one option is preselected, that is part of communication.

465. Balanced option ordering

Where neutrality matters, randomise or use a principled ordering rather than placing the preferred option first.

466. Decision support is not decision substitution

The tool should help people understand the evidence.

It should not pretend the science determines personal values automatically.

---

467. Communicating scientific confidence over time

Scientific confidence can increase or decrease.

Communication should reflect that evolution.

Early stage

“Initial evidence suggests…”

Replication stage

“Independent studies report similar results…”

Synthesis stage

“Across multiple studies, the average effect is…”

Mature stage

“The core relationship is strongly supported, with uncertainty remaining about…”

468. Confidence can decrease

If replications shrink the effect:

say so.

If a mechanism fails:

update the explanation.

469. Avoid confidence inertia

A public statement should not remain “breakthrough” years after the evidence became mixed.

470. Confidence version history

Preserve:

• date;
• status;
• evidence base.

---

471. Communication Under Deep Uncertainty

Some scientific problems have uncertainty that cannot be summarised by one narrow interval.

Examples:

• new technologies;
• long-term climate risks;
• rare catastrophic events.

472. Use scenarios when probabilities are not defensible

Instead of inventing precise probability, describe:

• plausible scenarios;
• conditions;
• consequences.

473. Separate “unknown probability” from “zero probability”

Lack of a credible numerical estimate does not mean an event cannot happen.

474. Deep uncertainty communication should support robustness

Decision makers may ask:

Which option performs acceptably across many plausible futures?

This is different from choosing the option with the highest point estimate.

475. Avoid false precision under deep uncertainty

Do not convert expert discomfort into:

“37.2% probability”

without a defensible elicitation/model.

---

476. Communication and Scientific Disagreement Matrix

Disagreement	Communication
data quality	explain which measurements differ
effect magnitude	show estimate ranges
mechanism	state shared result and rival pathways
generalisation	state populations/contexts
policy values	separate science from normative trade-off

477. Agreement deserves equal structure

Communicators often focus on disagreement because it is dramatic.

Also state:

• what all serious interpretations agree on;
• what evidence is not disputed.

478. Do not manufacture controversy

A disagreement among a tiny unrepresentative minority should not automatically be framed as:

“scientists are divided”.

479. Do not erase legitimate uncertainty

The opposite error is claiming complete consensus when meaningful scientific uncertainty remains.

---

480. Science Communication Integrity Record

A mature communication artifact can record:

1. source evidence;
2. evidence date;
3. core scientific claim;
4. magnitude;
5. uncertainty;
6. scope;
7. audience;
8. goal;
9. reviewers;
10. message version;
11. comprehension results;
12. update trigger.

481. Why preserve an integrity record?

It helps future reviewers distinguish:

• communication error;
• science update;
• audience misunderstanding.

482. Integrity record for visuals

Also record:

• source data;
• chart code or construction method;
• axis decisions;
• uncertainty encoding.

483. Integrity record for social posts

Store:

• published wording;
• link target;
• date;
• correction if any.

---

484. Communication Revision Standard

A revised message should answer:

• What changed?
• Why?
• Does it change the conclusion?
• What should readers do differently?

485. Minor revision

Example:

confidence interval typo corrected; conclusion unchanged.

486. Major revision

Example:

new replication reduces effect estimate substantially; headline and decision guidance updated.

487. Revisions should travel downstream

Update:

• website;
• PDF;
• social post;
• press release;

where feasible.

488. Scientific communication should be correctable by design

Use channels and versioning practices that allow updates.

---

489. Final Communication Audit Checklist

• Source is identifiable.
• Evidence is current.
• Goal is explicit.
• Audience is defined.
• Core claim matches the source.
• Study design is represented correctly.
• Magnitude is visible.
• Baseline is visible when relevant.
• Uncertainty is proportional.
• Scope is bounded.
• Evidence stage is visible.
• Replication status is visible where relevant.
• Important alternatives are not hidden.
• Policy/value judgment is separated from empirical evidence.
• Headline passes integrity test.
• Quotes pass integrity test.
• Visuals preserve magnitude.
• Denominators are comparable.
• Accessibility is considered.
• Conflicts are disclosed.
• Comprehension was tested where stakes justify it.
• Correction pathway exists.
• Update owner exists.

490. Final communication acceptance standard

A science message is ready when a reasonable member of the intended audience can understand:

what the science found, how large the result is, how uncertain it is, who or what it applies to, how mature the evidence is, and what the result does not establish.

---

491. Science Communication Stewardship

Scientific communication needs stewardship because public messages can outlive the evidence state that created them. A headline, infographic or explainer may continue circulating long after the underlying science has changed.

Stewardship responsibility 1 — Monitor evidence

Track:

• major replications;
• systematic reviews;
• guideline changes;
• retractions;
• important corrections.

Stewardship responsibility 2 — Monitor interpretation

Look for recurring public misunderstandings such as:

• association interpreted as causation;
• relative risk interpreted as absolute risk;
• scenario interpreted as forecast certainty.

Stewardship responsibility 3 — Update the visible layer

If the science changes, update:

• headline;
• summary;
• visual;
• FAQ.

Do not hide the change in an obscure note.

Stewardship responsibility 4 — Preserve history

Archive old versions so future readers can understand why the message changed.

492. Communication expiry risk

Some science messages age quickly.

High-expiry domains include:

• outbreaks;
• AI capability evaluations;
• new drug safety;
• fast-moving technology.

Low-expiry domains may include:

• well-established physical constants;
• basic mature mechanisms.

Review frequency should reflect expiry risk.

493. Evidence freshness label

Useful labels include:

• evidence reviewed through [date];
• last updated [date];
• preliminary evidence;
• living guidance.

494. Stale communication is a scientific error

A statement can have been accurate when written and become misleading later.

Scientific integrity includes maintenance.

---

495. Communicating uncertainty about the science versus uncertainty about the decision

These are different.

Scientific uncertainty:

How large is the effect?

Decision uncertainty:

Given the effect, cost, values and trade-offs, which action is best?

496. A precise estimate can still leave a difficult decision

Example:

Science estimates benefit and harm precisely, but people value the outcomes differently.

Communication should not claim that science eliminates the value trade-off.

497. A clear decision can exist under scientific uncertainty

If one option is:

• low cost;
• low harm;
• potentially high benefit;

action may be reasonable even while effect magnitude remains uncertain.

The message should explain both the evidence and the decision logic separately.

498. Decision recommendations should expose assumptions

Example:

“This recommendation assumes preventing Outcome Y is valued more than the small increase in Side Effect Z.”

This respects audience agency.

---

499. Communication Under Evidence Revision

When evidence changes, avoid the phrase:

“Science was wrong.”

Ask what changed:

• new sample?
• better measurement?
• larger study?
• new mechanism?

500. Revision categories

Magnitude revision

Effect remains, but size changes.

Boundary revision

Effect applies to fewer or more contexts.

Mechanism revision

Effect remains but explanation changes.

Direction revision

New evidence changes whether effect is beneficial, harmful or null.

Evidence-status revision

Preliminary finding becomes replicated or unsupported.

501. Communicating revision should teach self-correction

Useful structure:

Earlier evidence supported X. New evidence shows Y. The updated conclusion is Z because…

This makes revision intelligible rather than embarrassing.

---

502. Communication and Scientific Consensus Handoff

A dedicated scientific-consensus guide should explain:

• how communities converge;
• how expert agreement is measured;
• how consensus shifts;
• how dissent is evaluated.

This communication guide should preserve one limited job:

When consensus exists, represent it accurately without turning it into unanimity or certainty.

503. Communication and Peer Review Handoff

A dedicated peer-review guide should explain:

• reviewer selection;
• review reports;
• editorial decisions;
• review bias;
• open review;
• post-publication review.

This communication guide focuses only on:

how peer-review status is communicated to audiences without treating it as proof.

---

504. Scientific Communication Provenance Standard

Every important public claim should be traceable through:

public wording → institutional summary → scientific conclusion → underlying source evidence.

505. Provenance should survive sharing

Graphics and social posts can include:

• short source citation;
• date;
• QR/link;
• version.

506. Detached claims are high-risk

A number without source can be reused in contexts the original authors never intended.

507. Communication source hierarchy

For a mature scientific question, prefer:

• systematic synthesis;
• authoritative scientific report;
• current high-quality primary evidence.

The exact order depends on the question.

508. Source disagreement should be resolved scientifically

Compare:

• date;
• design;
• population;
• methods;
• evidence base.

Do not choose the source with the most convenient conclusion.

---

509. Final Communication Verification Appendix

Before a scientific message becomes public, another competent reviewer should be able to reconstruct its logic.

Verification step 1 — Read the source

Recover the actual scientific conclusion.

Verification step 2 — Read only the headline

Does it preserve the same conclusion?

Verification step 3 — Read the first paragraph

Are magnitude, uncertainty and scope still correct?

Verification step 4 — Inspect the visuals

Do they imply the same effect size and uncertainty?

Verification step 5 — Inspect the caveats

Are interpretation-changing limits visible enough?

Verification step 6 — Test the audience

Can readers teach the finding back accurately?

Verification step 7 — Check provenance

Can readers find the underlying evidence?

Verification step 8 — Check maintenance

Who updates the message if evidence changes?

510. Communication failure escalation

Minor

Unclear wording; core interpretation intact.

Action: revise copy.

Moderate

Missing baseline, denominator or status creates likely misunderstanding.

Action: correct visible message.

Major

Association communicated as causation or effect magnitude seriously exaggerated.

Action: prominent correction across channels.

Critical

Communication could cause harmful medical, safety or public decisions.

Action: urgent correction and incident review.

511. Communication incident report

Record:

• original message;
• error;
• audience reach;
• scientific consequence;
• correction;
• preventive change.

This turns a communication error into organisational learning.

---

512. Final Science Communication Integrity Standard

A science message deserves confidence when:

• its source is identifiable;
• its claim matches the evidence;
• its magnitude is visible;
• its uncertainty is proportional;
• its scope is bounded;
• its evidence status is visible;
• its visuals are honest;
• its audience can understand it;
• its correction/update path exists.

513. The final communication test

Ask:

If a reasonable member of the intended audience remembers only this message, will they remember the science more accurately—or more dramatically—than the evidence deserves?

If the answer is “more dramatically”, revise the message.

514. Communication continuity

Future updates should deepen:

• audience analysis;
• uncertainty communication;
• risk communication;
• visual integrity;
• misinformation correction;
• communication testing;
• maintenance.

They should not absorb peer review or consensus as processes.

515. How Scientific Communication Works: final synthesis

Scientific communication is successful when the message becomes easier to understand without becoming easier to misunderstand.

The durable sequence is:

start from the actual result → define the goal → understand the audience → preserve magnitude, uncertainty and scope → choose an honest format → test comprehension → publish with provenance → update when the science changes.

Science communication is therefore not decoration around evidence.

It is the final quality-control layer between scientific knowledge and the mental model another person carries away.

The strongest message is not the most dramatic.

It is the one whose clarity survives without sacrificing the boundaries that make the science true.

---

516. Communication Fidelity Appendix

Scientific communication fidelity means that the public or technical message preserves the essential structure of the source science even after compression. Fidelity does not require identical wording. It requires semantic continuity.

Fidelity 1 — Direction

If the source finding is positive, negative or null, the communication should preserve that direction.

Fidelity 2 — Magnitude

A small effect should remain small in the communication. A large effect should not be hidden.

Fidelity 3 — Uncertainty

A precise result should not be described as wildly uncertain. A wide interval should not be described as settled.

Fidelity 4 — Scope

A finding from one population, time period or context should not silently become universal.

Fidelity 5 — Design

Observational evidence should remain observational. Randomized causal evidence can be described causally within its domain.

Fidelity 6 — Evidence maturity

A preliminary paper, replicated finding and systematic synthesis should not be communicated as equivalent evidence stages.

517. Fidelity Drift

Scientific claims often drift as they move through communication layers.

Example:

1. Paper: “X was associated with a 5% lower risk in this cohort.”
2. Press release: “X linked to lower risk.”
3. News story: “X may protect against disease.”
4. Social post: “X prevents disease.”

Each step may look small.

The final claim is scientifically different.

518. Preventing claim drift

Attach a core claim to the communication package:

This sentence may be simplified, but its scientific meaning must not be strengthened.

Then review downstream versions against it.

519. Fidelity across visuals

A chart can drift even when text does not.

Check:

• axis range;
• annotation language;
• colour;
• highlighting.

The visual should not make a small difference look like a crisis or a noisy trend look deterministic.

520. Fidelity across translation

Cross-language communication should preserve:

• causal status;
• probability language;
• technical boundaries.

Independent scientific review can catch translation drift.

---

521. Communicating What Science Does Not Know

A complete scientific message sometimes needs a direct statement of unknowns.

Examples:

• long-term effects unknown;
• rare harms not yet well measured;
• mechanism uncertain;
• effect in children not tested.

522. Unknown does not mean everything is unknown

Good structure:

Known: Treatment A lowers Outcome Y over six months in adults.

Unknown: durability beyond one year and effect in adolescents.

This preserves both confidence and boundary.

523. Unknowns should be prioritised

Do not overwhelm readers with every conceivable uncertainty.

Highlight unknowns that affect:

• decision;
• scope;
• safety;
• causal interpretation.

524. Unknowns can guide future research

Communication can end with:

“The next important question is…”

This turns uncertainty into a research agenda rather than a rhetorical weakness.

---

525. Science Communication and Institutional Memory

Institutions should preserve why a message was written the way it was.

Useful records include:

• source study;
• approved core claim;
• review notes;
• audience-testing results;
• correction history.

526. Communication memory reduces repeated mistakes

If one institution repeatedly overstates:

• relative risk;
• early findings;
• causal language;

incident reviews can reveal the pattern.

527. Training should use past communication failures

Real examples teach:

• how distortion happened;
• how review could have caught it.

528. Communication owners should be named

Every durable public scientific page should have:

• content owner;
• scientific reviewer;
• review date.

---

529. Final Communication Certification

Certification A — Scientifically accurate

No material conflict with the source evidence.

Certification B — Interpretively complete

Magnitude, uncertainty and scope are visible enough to prevent predictable overinterpretation.

Certification C — Audience-validated

Target readers understand the message as intended.

Certification D — Decision-capable

The message supports the relevant decision with balanced benefits, harms and uncertainty.

Certification E — Maintained

Version, evidence date and update ownership are preserved.

These are descriptive levels, not a universal formal standard.

530. Certification should remain revocable

If evidence changes or the message is shown to mislead, the certification status should change.

531. Communication quality is empirical

The communicator’s intention is not enough.

If readers systematically misunderstand the claim, the message should be redesigned.

---

532. Final Source Integrity Note

The central research foundation for this owner includes the National Academies report Communicating Science Effectively: A Research Agenda, which emphasises that effective science communication depends on purpose, audience, context, uncertainty and trust, and that the simple information-deficit model is inadequate.

The owner also follows the broader scientific principle that uncertainty and evidence maturity should remain visible as findings move from technical publication into public communication.

Authoritative references retained in this article include:

• National Academies — Communicating Science Effectively: A Research Agenda
• National Academies — The Complexities of Communicating Science
• National Academies — Building the Knowledge Base for Effective Science Communication
• Communicating uncertainty about facts, numbers and science

533. Scope and future extensions

This guide focuses on the scientific problem of moving a result into another person’s mind without changing what the result scientifically means.

Future updates should deepen:

• risk communication;
• uncertainty communication;
• visual integrity;
• audience testing;
• corrections;
• misinformation resilience;
• communication maintenance.

Peer review and scientific consensus are related but distinct processes and deserve separate treatment.

534. Final Communication Integrity Close

The most important communication skill in science is not simplification by itself.

It is fidelity under simplification.

The final rule is:

Change the language, the examples, the format and the level of detail as needed—but do not change the direction, magnitude, uncertainty, scope or evidential status of the science.

When that rule survives every headline, graph, press release, classroom explanation and policy brief, science communication has done its job.

---

535. Final Communication Verification Appendix

A science message should be verifiable as a transformation rather than judged only by whether it sounds clear. Another competent reviewer should be able to move backward from the public message to the underlying scientific result and confirm that no scientifically important meaning was gained or lost during compression.

Verification step 1 — Recover the source result

Identify the paper, review, dataset or guideline that supports the message.

Verification step 2 — Recover the source conclusion

Write the strongest conclusion the original evidence actually supports.

Verification step 3 — Compare the public claim

Check whether:

• association became causation;
• small effect became large;
• one population became universal;
• uncertainty disappeared;
• preliminary evidence became settled fact.

Verification step 4 — Compare the numbers

Verify:

• baseline;
• absolute effect;
• relative effect;
• time horizon;
• denominator.

Verification step 5 — Compare the visual

Ensure axes, scales, labels and uncertainty bands do not create a stronger impression than the numbers support.

Verification step 6 — Compare the audience interpretation

Ask intended readers to explain the finding back. If they consistently leave with a stronger causal, larger or more certain mental model, the message is not high-fidelity even if every sentence is technically defensible.

Verification step 7 — Compare the update path

Identify who will revise the communication if new evidence changes the scientific conclusion.

536. The communication preservation rule

The final public artifact should preserve enough context that future readers can still tell:

• what evidence existed at the time;
• what level of confidence was justified;
• which uncertainties remained.

This matters because science communication can persist online long after the original study, news cycle or decision context has passed.

537. Final acceptance sentence

A science message is ready when simplifying it has made the result easier to understand without making it scientifically stronger, broader, more certain or more dramatic than the evidence allows.

That sentence is the final publication gate for this owner.

---

538. Final Scientific Communication Integrity Note

Scientific communication should remain useful even when the audience, channel and technology change. New platforms will appear, new visual formats will emerge, and AI may automate more of the drafting process. The durable obligation is not tied to any one medium.

A high-integrity scientific message should always preserve five elements:

1. The claim. What does the evidence actually support?
2. The magnitude. How large is the effect, risk, difference or trend?
3. The uncertainty. How precise is the result, and which important unknowns remain?
4. The scope. Who, what, where and when does the conclusion apply to?
5. The evidence status. Is this preliminary, peer reviewed, replicated, synthesised or broadly established?

If any communication format drops one of these elements, the communicator should ask whether the missing information changes the mental model a reasonable reader will carry away.

This is especially important when compression is extreme. A six-word headline, a 15-second video and a one-panel graphic can all communicate science responsibly—but only if they avoid silently upgrading the scientific claim.

The final communication rule is therefore:

Scientific communication is complete only when the audience can understand the finding without inheriting a stronger conclusion than the evidence itself earns.

That is the standard that should survive every future update to this guide.

539. Final Verification Close

A scientific message should remain auditable after the original communicator is gone. A future reviewer should be able to recover the source evidence, the intended audience, the communication goal, the effect magnitude, the uncertainty, the population boundary and the evidence status without guessing why the message was written the way it was.

That persistence is the final proof that the communication belongs to the scientific record rather than to one news cycle. It allows later researchers to distinguish three possibilities clearly: the science changed, the communication was wrong, or the audience interpreted an accurate message differently than intended.

Integrity: simplify the language as much as needed, but preserve the scientific claim exactly enough that the audience can still tell what is known, how strongly it is known, and what remains outside the evidence.

Final standard: accurate science communication preserves evidence, magnitude, uncertainty, scope, status and revision history closely enough that simplification never becomes distortion.

Scientific clarity must remain faithful to scientific limits.