A message arrives from someone trusted.
The claim is wrong.
The sender is not lying.
They read it somewhere else, remembered part of it, added a detail that seemed plausible and passed it on because the message felt useful.
The social consequence can still be real even when deception was absent.
That is one reason the study of misinformation cannot be reduced to bad actors.
Definition
In this Knowledge Library, misinformation refers to:
false, inaccurate or materially misleading information whose classification does not require evidence that the person creating or sharing it intended to deceive.
This wording is deliberately cautious.
Some literatures use misinformation as an umbrella for false or misleading information regardless of intent. Others reserve it more specifically for unintentional falsehood and contrast it with deliberate disinformation.
The Institute will preserve the source literature’s definition when discussing a specific study.
That prevents a common error: moving findings between incompatible taxonomies as if the words meant exactly the same thing everywhere.
What misinformation is not
Misinformation is not simply disagreement.
A contested claim cannot be classified as false merely because an authority, majority or institution rejects it.
It is not satire by default. Satire can contain false literal statements without functioning as deceptive factual assertion.
It is not disinformation unless intention is part of the definition and evidence supports an intentional deception claim.
And it is not the same as a mistake in private reasoning. Misinformation is generally studied as information content, communication or belief that can circulate between people and systems.
Truth classification is harder than the label suggests
Researchers often need to decide whether a stimulus is true or false before they can measure discernment.
That classification may be straightforward for a fabricated headline with a verified fact-check.
Other cases are more difficult.
Claims can be partly true, outdated, decontextualized, probabilistic or dependent on interpretation. A number may be accurate while the implied comparison is misleading. An authentic photograph may be paired with the wrong event. A genuine quotation may be edited so that its meaning changes.
Misinformation research therefore depends on careful operationalization of both accuracy and misleadingness.
People are not uniformly unable to tell truth from falsehood
Public discussion often assumes that ordinary people cannot distinguish accurate from inaccurate information.
Recent evidence is more nuanced.
Pfänder and Altay’s 2025 preregistered meta-analysis synthesized 303 effect sizes from 67 experimental articles, covering 194,438 participants in 40 countries across six continents.
On average, people rated true content as more accurate than false content, with a large truth-versus-falsehood discrimination effect. The analysis also found a skepticism bias: participants were better at identifying false content as false than true content as true.
This does not imply that misinformation is harmless.
It changes the baseline story. The average person is not simply credulous. Errors include both believing falsehoods and failing to trust truths.
Discernment is different from skepticism
Suppose an intervention makes people trust every headline less.
False-content belief declines.
So does true-content belief.
Has discernment improved?
Not necessarily.
Good misinformation research therefore distinguishes sensitivity to truth versus falsehood from a general response bias toward belief or disbelief.
This is one reason signal-detection approaches are useful. They can separate improvement in discrimination from a global shift in skepticism.
The distinction is central to the Institute’s framework because epistemic health requires appropriate trust, not maximum doubt.
Familiarity and repetition matter
Misinformation can become familiar through repeated exposure.
Familiarity is not the same as belief, and repetition does not mechanically force acceptance. But people often use processing fluency and familiarity as cues when making judgments under limited attention.
Corrections can reduce false belief, yet misinformation may continue to influence reasoning under some conditions.
The literature on the continued-influence effect is one reason debunking cannot be understood as simple deletion from memory.
A corrected narrative may leave a gap. If no coherent alternative explanation replaces it, parts of the original model can remain functionally active.
Educational interventions can help, with limits
A 2025 systematic review by Rau and Premo examines educational approaches to misinformation and the diversity of interventions designed to improve evaluation.
The broader intervention literature includes lateral-reading instruction, fact-checking strategies, inoculation or prebunking, source evaluation and media-literacy curricula.
Results are sufficiently positive to reject fatalism.
But no intervention creates universal immunity.
Effects depend on population, outcome, content domain, delivery, timing and whether the study measures immediate classification accuracy, durable belief change, sharing behaviour or something else.
Inoculation should improve discrimination, not generalized distrust
Prebunking and inoculation interventions expose people to misleading techniques or weakened forms of manipulation before later exposure.
A 2026 signal-detection meta-analysis by Simchon and colleagues synthesized 33 inoculation experiments involving 37,075 participants.
The analysis is important because it separates discrimination from response bias. It reports improved ability to distinguish misleading from reliable information without evidence that the benefit was merely generalized distrust.
That is a stronger outcome than simply teaching people to say “false” more often.
It also remains bounded by the specific intervention designs, populations and materials in the evidence base.
Source cues help, but source heuristics can fail
People often use source reputation, professional appearance, social endorsement and familiarity to judge claims.
These cues can be adaptive. No one can independently verify everything.
The problem arises when cues are copied more easily than the underlying reliability.
A misleading site can imitate a news layout. An AI system can generate references that look scholarly. A familiar person can confidently relay a false claim.
This makes source criticism important but insufficient unless source judgments connect back to evidence and provenance.
Generative AI changes scale, not the definition by itself
Generative systems can produce inaccurate content rapidly and at low cost.
They can also help check claims, summarize evidence and identify inconsistencies.
An AI-generated error is not automatically disinformation. If no deceptive intent is established, the safer classification may be misinformation, model error, unsupported generation or another category appropriate to the context.
The temptation to infer motive from output should be resisted.
For institutions, the practical issue is often provenance: what system generated the claim, from what inputs, under whose responsibility and with what verification process?
Measurement choices can manufacture conclusions
A misinformation study may ask participants to rate accuracy on a scale, choose true/false, decide whether to share, recall claims later or select which sources they would consult.
These tasks measure different things.
Sharing intention is not belief.
Belief is not memory.
Accuracy judgment is not behaviour under social pressure.
And performance on a curated set of headlines may not generalize to a person’s everyday media environment.
The measurement claim should therefore be no broader than the task supports.
European competence frameworks increasingly include misinformation evaluation
DigComp 3.0 integrates misinformation and disinformation awareness into digital competence, including source evaluation, credibility assessment and attention to how digital content is generated.
This is especially relevant in AI-mediated environments where origin and authenticity can be opaque.
The framework is not causal evidence that European citizens possess those skills.
It is a policy and competence architecture identifying them as necessary.
The correction problem is partly a memory problem
Correcting misinformation is not equivalent to erasing it.
People build mental models. When a false explanation fills a causal gap, removing the false piece can leave the model incomplete. Under some conditions the original information continues to influence inference even after a correction is remembered.
This is one reason effective corrections often do more than say “false.” They provide an alternative explanation, repeat the accurate information clearly and avoid unnecessary repetition of the myth.
The literature is nuanced, and correction effects vary. The important principle is that belief revision is constructive, not simply subtractive.
Identity and social context change uptake
Information arrives through relationships.
A correction from a trusted peer may land differently from the same correction delivered by a distrusted institution. A claim can signal group membership in addition to conveying factual content. Sharing can serve social purposes other than expressing belief.
This makes it risky to infer belief from sharing alone.
People sometimes share because content is funny, identity-confirming, outrageous or socially useful. Conversely, people can believe something without sharing it.
Misinformation research increasingly benefits from separating these outcomes.
Accuracy prompts reveal a useful dissociation
Some experiments show that drawing attention to accuracy before sharing can improve the quality of what people choose to share.
The interpretation should remain specific.
Such findings suggest that part of low-quality sharing can reflect attention allocation rather than stable endorsement of falsehood.
They do not imply that every misinformation problem is solved by a reminder to “think before you share.” Social incentives, ideology, platform design and motivated reasoning can still matter.
The broader lesson is that the same person may display different epistemic performance depending on what the environment makes salient.
Prebunking and debunking answer different temporal questions
Debunking responds after exposure.
Prebunking or inoculation tries to prepare people before later exposure, often by teaching manipulation techniques or warning about recurring patterns.
Neither should be treated as universally superior.
The relevant choice depends on whether a known false claim is already circulating, whether a technique generalizes, how much time is available and what outcomes matter.
The best evidence architecture may combine both: strengthen general discrimination while maintaining capacity for claim-specific correction.
Misinformation can be generated by institutions without deception
Institutional scale does not imply malicious intent.
Poor data pipelines, outdated guidance, ambiguous communication and automated systems can propagate inaccurate information through legitimate organizations.
This matters because a purely adversarial model of misinformation can miss systemic error.
An organization that treats every false output as an attack may fail to improve its own correction processes.
Discernment requires enough humility to ask whether the error came from manipulation, ordinary uncertainty, bad incentives, weak review or simple mistake.
Measuring success should include truth acceptance
An intervention that lowers belief in false headlines but also lowers belief in true headlines may look successful if researchers report only the first outcome.
A stronger evaluation asks whether discernment improves.
This is why the Pfänder and Altay meta-analysis matters. It treats true and false judgments together, revealing both sensitivity and skepticism bias.
For Inner Technology, this becomes a general measurement principle: whenever an intervention aims to protect people from error, also ask whether it damages appropriate trust.
Institutional correction is a capacity
Organizations often focus on preventing false information while underinvesting in what happens after an error escapes.
A correction-capable institution can identify the original claim, publish the corrected information in the same channels, explain what changed, preserve an audit trail and update downstream materials.
This is harder than deleting a post.
Copies may persist. Search engines may cache old versions. AI systems may continue reproducing the earlier statement. Partners may have incorporated it into their own documents.
Correction therefore needs propagation, not only publication.
The Institute treats this as an institutional analogue of metacognitive revision: a knowledge system should be able to notice error, change state and leave evidence of why it changed.
Uncertainty should not be mistaken for misinformation
Fast-moving science and emergencies produce provisional claims.
A forecast can be wrong without having been misinformation when issued. An estimate can change because new data arrived. A scientific hypothesis can fail without becoming a deceptive statement retroactively.
This temporal distinction matters for public trust.
If every revised claim is labelled misinformation, normal learning begins to look like incompetence or deceit. If uncertainty is hidden to avoid that appearance, institutions become less honest.
Good communication makes confidence and update conditions visible from the start.
The denominator problem
Misinformation prevalence matters when interpreting risk.
A feed containing a small number of false items can still create serious harm, but base rates affect what detection performance means. A classifier or human judge that labels almost everything false may appear accurate in one environment and fail in another.
This is another reason precision, recall, discrimination and response bias should not be collapsed into one “accuracy” number.
For public communication, base-rate awareness also protects against sensationalism. Evidence that some misinformation spreads widely does not establish that most information is false. The ecology has to be measured rather than imagined.
The same claim can change category over time
Information classification is temporal.
A statement based on the best available evidence can later become outdated. A rumor can later be verified. A tentative scientific interpretation can harden into consensus or be overturned.
Researchers should therefore record when a truth judgment was made and against which evidence state. Otherwise a later reader may mistake revision for proof that the earlier communicator acted deceptively.
This temporal discipline is especially important for archives and AI-generated summaries that detach claims from the date on which they were evaluated.
What the evidence does not show
The evidence does not show that people believe most misinformation they encounter.
It does not show that skepticism is always protective.
It does not show that one media-literacy, prebunking or fact-checking intervention transfers to every topic and population.
It does not show that all false AI output is deliberate disinformation.
It does not show that intention can reliably be inferred from content alone.
And a laboratory accuracy judgment does not by itself reveal what someone will share, remember or act on in a real social setting.
What this changes for Inner Technology
Misinformation turns discernment into a discrimination problem rather than a distrust problem.
The Institute therefore emphasizes a sequence:
Separate claim from messenger. Establish what would count as accurate. Trace provenance. Seek independent support. Distinguish uncertainty from falsity. Notice whether the judgment is becoming globally skeptical. Correct with an alternative explanation where possible. Revisit the belief when evidence changes.
That sequence is Institute synthesis, not a validated misinformation scale.
Its purpose is to preserve a human capacity that becomes more important as information becomes cheaper to produce.
The trusted message
The original sender may never have intended harm.
That does not make the claim true.
Nor does discovering the error make the sender permanently untrustworthy.
Discernment has to hold both facts at once.
Misinformation is difficult partly because truth moves through relationships.
The mature response is not to stop trusting.
It is to become better at giving trust the right evidential shape.
Research references
- Wardle C, Derakhshan H. Information Disorder: Toward an interdisciplinary framework for research and policy making. Council of Europe. 2017. https://rm.coe.int/information-disorder-toward-an-interdisciplinary-framework-for-researc/168076277c
- Pfänder J, Altay S. Spotting false news and doubting true news: a systematic review and meta-analysis of news judgements. Nature Human Behaviour. 2025. https://doi.org/10.1038/s41562-024-02086-1
- Rau MA, Premo AE. A Systematic Review of Educational Approaches to Misinformation. Educational Psychology Review. 2025. https://doi.org/10.1007/s10648-025-10012-8
- Simchon A, Zipori T, Teitelbaum L, Lewandowsky S, van der Linden S. A signal detection theory meta-analysis of psychological inoculation against misinformation. Current Opinion in Psychology. 2026;67:102194. https://doi.org/10.1016/j.copsyc.2025.102194
- Cosgrove J, Cachia R. DigComp 3.0. European Commission Joint Research Centre. 2025. https://doi.org/10.2760/0001149