A scientist explains the evidence carefully.
The audience hears the correct information.
Later, one person remembers the conclusion with appropriate caution. Another remembers it with absolute certainty. A third misremembers the conclusion but remains highly confident.
From the standpoint of information delivery, the first part of the job may look finished.
From the standpoint of judgment, it is not.
This is the problem at the center of Helen Fischer and Markus Huff's 2025 Nature Reviews Psychology Comment, Metacognition as a target of science communication.
Their argument is simple enough to fit into a short article and important enough to reorganize how communication success is imagined.
People do not act only on what they know.
They also act on how certain they think they are.
The paper
The Comment was published on 24 July 2025 in Nature Reviews Psychology, volume 4, pages 554-555.
Fischer is affiliated with the Leibniz-Institut fuer Wissensmedien in Tuebingen. Huff is affiliated with the same institute and the University of Tuebingen.
The paper's central proposition is that metacognitive accuracy, the alignment between confidence and actual correctness, should be a target of science communication.
That formulation matters because it separates two outcomes that are often treated as one.
A communication can improve knowledge without improving calibration.
A person can become more confident without becoming more accurate.
A person can even become more accurate while remaining unable to distinguish what they know from what they are guessing.
Knowledge is not the only variable that guides action
Suppose someone is deciding whether to repeat a medical claim, support a policy, challenge a colleague or ask for more evidence.
The factual content of their belief matters.
So does confidence.
Low confidence in a correct belief may lead to unnecessary deference or information search.
High confidence in an incorrect belief may lead to premature closure.
When certainty and correctness are aligned, people are better positioned to decide when to act and when to check.
This is why metacognitive calibration is not a decorative psychological variable.
It changes the behavior that follows knowledge.
The paper draws on a wider metacognition literature
Fischer and Huff point toward research showing that certainty can influence information seeking and belief updating.
One cited example is work by Schulz and colleagues showing that dogmatism was associated with reduced information search under uncertainty.
Another concerns misplaced versus well-placed certainty when people update false beliefs.
The Comment also cites work linking intellectual humility with metacognitive ability.
The synthesis is that communication aimed only at first-order knowledge may miss a second-order process: how people evaluate the reliability of their own knowledge.
The paper does not claim that confidence is the sole mediator of science communication.
It argues that confidence deserves a more explicit place in the target architecture.
First-order and second-order judgments should be separated
A first-order judgment answers a question such as:
Is this claim true?
A second-order judgment asks:
How sure am I that my answer is right?
The distinction is central to metacognition.
Someone can make a correct first-order judgment with poor second-order insight. They can also make an incorrect judgment while accurately recognizing their uncertainty.
Those two error patterns are not equivalent.
The second person may be more likely to seek help or defer action.
For communication, this means accuracy and calibration should sometimes be measured separately.
Misplaced certainty is different from ignorance
Ignorance is often easier to imagine than miscalibration.
If a person does not know something, the communication problem seems straightforward: provide information.
Misplaced certainty is harder.
The person may already feel that the question is settled.
Additional information can be ignored, selectively interpreted or treated as redundant.
This is why the paper's focus intersects with research on dogmatism, intellectual humility and information search.
The intervention target may not be simply "more facts."
It may be a better relationship between evidence and certainty.
Calibration should not become compulsory self-doubt
There is an obvious risk in turning metacognition into a cultural slogan.
If the message becomes "always doubt yourself," communication has failed in the other direction.
Well-calibrated confidence includes confidence.
When evidence is strong and understanding is good, appropriate certainty can support decisive action.
The aim is not low confidence.
It is confidence that tracks justification and performance as well as possible.
This distinction is especially important in public institutions. Excessive hedging can make reliable knowledge appear weaker than it is, while false certainty can make contested claims look settled.
Measuring calibration is not trivial
The paper's argument depends on the possibility of evaluating metacognition.
But metacognitive measurement has its own difficulties.
Confidence ratings can be examined for absolute calibration, relative resolution or discrimination between correct and incorrect responses.
Different measures answer different questions.
A person may be overconfident on average yet still assign higher confidence to correct than incorrect answers.
Another may have an average confidence level that matches average accuracy but fail to distinguish individual successes from errors.
Recent work by Dobromir Rahnev evaluating 17 metacognitive measures under artificial data corruption is useful here. It shows that multiple measures can satisfy a particular validity criterion while differing in interpretation and statistical properties.
Science communication therefore should not simply "add a confidence question" and assume metacognition has been captured.
Asking for confidence can change cognition
Measurement can also be reactive.
When people are asked to report confidence, the act of reporting can alter attention, review or task performance.
This means confidence elicitation is not always a passive observation window.
In some contexts, that reactivity may be useful. Asking someone to state confidence can itself create a moment of reflection.
But then the prompt is partly an intervention.
The design should be honest about which role it is playing.
Science communication operates inside trust relationships
Metacognitive accuracy cannot be separated completely from source evaluation.
People rarely inspect raw evidence directly.
They rely on scientists, journalists, institutions, public agencies and expert communities.
A person may be uncertain about the content but highly confident in the source.
Or they may understand the evidence but distrust the institution communicating it.
Confidence can therefore attach to different objects:
- the claim;
- one's own understanding;
- the source;
- the evidence base;
- the institution;
- the decision that follows.
A communication strategy that asks only, "How confident are you?" may not know which object the answer refers to.
Communicators also need metacognition
The paper is written primarily about audiences, but the logic runs both ways.
Experts and institutions can also be miscalibrated.
A communicator can mistake disciplinary fluency for certainty outside their domain.
An institution can state a provisional estimate with a tone that sounds final.
A journalist can compress uncertainty until the public receives a binary conclusion.
A model-generated summary can erase the difference between strong consensus and one suggestive study.
Metacognitive communication therefore includes communicating the strength and limits of the evidence itself.
This is an Institute synthesis extending the paper's logic, not a claim tested by Fischer and Huff.
The argument matters more in an AI-mediated information environment
Generative AI can answer scientific questions in fluent, finished prose.
Fluency can make epistemic status difficult to see.
A system can produce the language of certainty without possessing human confidence, and users can infer confidence from style.
The problem becomes double-layered.
The user needs to assess the claim.
They also need to assess their confidence in their own assessment of an answer produced by a system whose uncertainty may not be legible.
This makes calibration, source access and verification central to AI literacy.
Again, that is a contemporary application of the paper's argument, not evidence contained in the Comment itself.
Science communication has traditionally optimized other outcomes
Communication research often asks whether audiences understood, remembered, trusted or accepted a message.
Those are legitimate outcomes.
Fischer and Huff's intervention is to add another question:
Do people know when they might be wrong?
That is a more demanding criterion.
A successful explainer may increase factual accuracy while also inflating confidence beyond what the evidence supports.
A more cautious explainer may reduce unjustified certainty while leaving people less satisfied.
The "best" communication therefore depends partly on the decision context.
High-stakes contexts need decision-linked calibration
In a low-stakes trivia question, poor confidence calibration may matter little.
In medicine, finance, public safety or policy, it can matter greatly.
The relevant outcome is not abstract metacognitive virtue.
It is whether people make better choices about checking, acting, deferring or revising.
This is one reason communication studies should connect confidence measures to behavior where possible.
Does low confidence trigger appropriate information search?
Does high confidence suppress checking even when the person is wrong?
Does communicating uncertainty help people calibrate or simply lower trust?
Those are empirical questions.
Intellectual humility is adjacent, not identical
The paper cites work connecting intellectual humility with metacognitive ability.
The constructs should still remain distinct.
Intellectual humility is often defined in terms of recognizing the limits or fallibility of one's beliefs or knowledge.
Metacognitive calibration concerns correspondence between confidence and performance.
A person can endorse humility as a value without being well calibrated on a task.
Another person can be statistically well calibrated without showing humility in interpersonal disagreement.
The overlap is important.
The distinction protects measurement.
What the paper does not establish
The 2025 Comment does not present a new randomized intervention.
It does not show that confidence prompts improve science understanding.
It does not establish which metacognitive measure communicators should use.
It does not show that calibration training transfers across domains.
It does not show that lower confidence is better.
It does not establish that metacognitive accuracy alone solves misinformation, polarization or mistrust.
And because it is a brief Comment, it should not be represented as a comprehensive systematic review of science communication.
Calibration is more informative when accuracy is difficult enough to vary
Metacognitive accuracy can only be studied meaningfully when the first-order task has room for both correct and incorrect judgments.
If every science claim in a study is obvious, participants may be nearly perfect and confidence will have little opportunity to discriminate good from poor knowledge. If every item is impossibly obscure, low accuracy may reflect missing knowledge rather than a metacognitive problem.
Good measurement therefore needs an appropriate range of difficulty.
It also needs enough observations per person to estimate whether confidence tracks performance rather than fluctuating randomly. A single confident mistake is not evidence of globally poor metacognition, just as one lucky low-confidence answer is not evidence of excellent calibration.
This is one reason Rahnev's 2025 comparison of metacognitive measures is relevant. Different metrics answer different questions and have different statistical properties. There is no reason to assume that one number captures every meaningful aspect of metacognition.
Confidence language can be designed without pretending uncertainty is simple
Science communicators often use verbal categories such as “high confidence,” “likely,” “uncertain” or “preliminary.”
These labels can help readers, but only when their meaning is clear.
Different institutions use the same words differently. Readers also map verbal expressions onto different implicit probabilities. “Likely” may mean 60 percent to one person and 90 percent to another.
Where quantitative uncertainty is available and understandable, numerical ranges can help. Where it is not, communicators can still state what is known, what is inferred, what could change the conclusion and which uncertainties matter for the decision.
The goal is not to put a probability on every sentence.
It is to stop rhetorical certainty from outrunning evidential certainty.
Corrections need metacognitive design too
Science communication does not end when an error is identified.
A correction has to compete with the memory, familiarity and emotional salience of the original claim.
If the correction simply says “false” without providing a coherent alternative account, people may remember the original narrative more readily than the update. If the correction is overly absolute, later revisions can damage trust.
A metacognitive approach asks not only whether the factual update is present but whether the audience can tell how much confidence the corrected claim deserves.
That may require explaining why the original evidence looked persuasive, what new information changed the assessment, and whether the revision is small or fundamental.
Transparent correction can model science as an updating process rather than as a performance of infallibility.
Communicator confidence and audience confidence can diverge
A scientist can be appropriately cautious while an audience becomes overconfident.
A headline can remove caveats. A visual can make a noisy estimate look precise. A press release can describe a correlational result with causal verbs. A social-media summary can turn “may” into “does.”
Metacognitive accuracy therefore has a transmission problem.
It is not enough for the original researcher to understand uncertainty if every later layer amplifies confidence.
A useful provenance chain should preserve not only source identity but epistemic qualifiers: study design, uncertainty, population, limitations and whether a claim is replication, preliminary evidence or synthesis.
This is especially important for AI-generated summaries, which can compress a nuanced paper into smooth declarative prose.
Trust should be earned through calibrated institutions, not only calibrated individuals
The paper's focus on metacognition can be extended carefully from people to institutions.
Institutions also communicate confidence.
They publish forecasts, risk levels, evidence grades and recommendations. Their credibility depends partly on whether those expressions of certainty track later outcomes and whether errors are corrected visibly.
Institutional calibration cannot be measured exactly like individual confidence, but the analogy is useful.
A public-health body that distinguishes strong evidence from provisional evidence supports better decision-making. A newsroom that labels developing stories differently from verified investigations does the same. A research institute that records evidence grades and open questions makes its own uncertainty inspectable.
This is not merely style.
It is epistemic infrastructure.
The target is not humility in the abstract
A communication program can accidentally reward chronic hedging.
That would be a mistake.
When evidence is strong, excessive uncertainty can mislead as much as overconfidence. Communicating “scientists are not completely certain” about a mature finding can create a false impression of substantive controversy.
Calibration demands movement in both directions.
The communicator should become more confident when evidence converges and less confident when it weakens.
The audience should learn that uncertainty is not a defect to be removed from science but a property to be estimated.
That is a more demanding educational goal than teaching people simply to “question everything.”
What the evidence does not show more broadly
Research on confidence and metacognition does not show that people have transparent access to all the reasons behind their certainty.
It does not justify treating confidence as a stable global trait.
It does not show that one calibration score travels across science, politics, medicine and everyday judgment.
It does not eliminate the role of domain knowledge, source trust, motivation, identity or social context.
Metacognition adds a necessary layer.
It does not replace the rest of epistemic life.
Why this matters in Northern Europe
The authors' institutional base in Tuebingen gives the paper direct European relevance, but the stronger reason to include it is conceptual.
European digital-literacy and AI-literacy frameworks increasingly ask learners to evaluate information and AI outputs critically.
Calibration is one way to make that demand more precise.
A learner who can say "I do not know enough to trust this answer yet" is demonstrating a different capacity from someone who merely knows a checklist of reliable sources.
For Nordic education systems, the research opportunity is not to add another self-report scale.
It is to test whether confidence-sensitive instruction improves information search, verification and revision in actual learning environments.
What this changes for Inner Technology
Inner Technology is concerned with capacities that govern action from within.
Metacognition belongs near the center because people act from their model of their own knowing, not from objective knowledge alone.
For the Institute, Fischer and Huff's argument can be translated into a simple design principle:
Do not ask only, "Did the person learn the answer?"
Also ask:
Can they tell when the answer is uncertain?
Can they distinguish memory from inference?
Can they notice when confidence outruns evidence?
Can they seek help before certainty hardens into action?
Those questions turn science communication from information transfer into capacity design.
The capacity claim is broader than the paper.
The paper provides one important reason to take it seriously.
The audience leaves
The scientist finishes the explanation.
The listeners leave with facts, impressions and different levels of certainty.
The communication has not ended at the edge of the room.
It continues in the next decision: what to repeat, what to verify, what to doubt and what to act on.
The useful question is not whether everyone became less certain.
It is whether certainty became better attached to reality.
Research references
- Fischer H, Huff M. Metacognition as a target of science communication. Nature Reviews Psychology. 2025;4:554-555. https://doi.org/10.1038/s44159-025-00478-2
- Schulz L, Rollwage M, Dolan RJ, Fleming SM. Dogmatism manifests in lowered information search under uncertainty. Proceedings of the National Academy of Sciences. 2020;117:31527-31534. https://doi.org/10.1073/pnas.2009641117
- Olcaysoy Okten I, Huang T, Oettingen G. Updating false beliefs: the role of misplaced vs. well-placed certainty. Psychonomic Bulletin & Review. 2023;30:712-721. https://doi.org/10.3758/s13423-022-02196-9
- Fischer H, Kause A, Huff M. Intellectual humility links to metacognitive ability. Personality and Individual Differences. 2025;238:113028. https://doi.org/10.1016/j.paid.2024.113028
- Rahnev D. A comprehensive assessment of current methods for measuring metacognition. Nature Communications. 2025;16. https://doi.org/10.1038/s41467-025-56117-0
- Double KS, Birney DP, Walker SA. A meta-analysis and systematic review of reactivity to judgments of learning. Memory. 2018;26(6):741-750. https://doi.org/10.1080/09658211.2017.1404111