RumourEngine

Field issue 01
Signals / memory / crowds

A social transmission laboratory — no login, no tracking

A message is not a parcel.
It is a relay.

Information moves through people, interfaces and institutions. At every hand-off it can be compressed, sharpened, sourced, doubted, corrected or made newly useful. Operate the network below to see one illustrative model of those pressures—then inspect where the model stops.

The Rumour Loom A coded artwork of a sentence passing through five colourful mouths and emerging fragmented, qualified and reconnected. DETAILFEELINGSOURCE?REPEATCHECK meaning is made in transit
Original code-drawn artwork: The Rumour Loom, 2026.
01

Information

A claim, observation, instruction or account that can be true, false, mixed, uncertain or not yet settled.

02

Misinformation

False or misleading information shared regardless of intent. Intentional deception is more specifically called disinformation.

03

Rumour

Unverified information circulating because it seems relevant under ambiguity. A rumour can later prove true, false or partly true.

Core exhibit / Rumour Mutation Network

Put one report through the whisper loom.

This is a deterministic teaching model, not a forecast. Change the pressures, send the report one relay at a time, and compare the visible result with its causal recipe.

1 / Seed report

78/240 characters. Do not enter private or identifying information; text stays in this browser. The seed locks after the first relay; undo to the source to edit it.

2 / Transmission pressures

Higher settings systematically remove descriptive words and time/detail markers.

Higher settings add urgency cues. It models message form, not anybody’s motive.

Local-only state: the experiment saves to localStorage on this device. Nothing is sent anywhere.

Live output

Transmission trace

Illustrative · unvalidated
  1. 00 / source

    The footbridge will close at 6 pm while inspectors examine a loose railing.

    Original fictional report

0 of 7 relays sent

Explainable readout

Seed detail retained
100%
Provenance visibility
70%
Added certainty / urgency
0%
Visible recipients
1one source node
Why this result?

No relay yet. The seed is shown unchanged; the 70% provenance baseline represents the selected “forwarded” label in this teaching model.

Read before interpreting

What this does not prove

The network cannot tell whether a real claim is true, predict whether it will spread, measure an individual’s susceptibility, infer why somebody shared something, or estimate an intervention’s effect. Its transformations are authored rules inspired by documented mechanisms. They are illustrations, not validated scores. Real transmission depends on audience, culture, platform design, incentives, timing, identity, evidence and many other conditions.

Field notes / evidence with seams showing

Rumours do social work. That does not make them reliable.

A useful account starts neither with “people believe anything” nor with “platforms made us irrational.” It asks what is uncertain, what a message offers, how the network rewards it, and what evidence can travel alongside it.

A / Mechanisms

Five pressures that can change a message

Reconstruction, not photocopying. Human remembering is selective. When people retell an account, they rebuild it from remembered elements and existing knowledge. Serial-reproduction research has long used chains of retellings to study this process. Compression can make a message easier to carry, but it can also remove dates, conditions, scope and uncertainty—the very details needed to evaluate it.

Repetition and familiarity. In controlled experiments, prior exposure can increase the perceived accuracy of plausible false headlines. Pennycook, Cannon and Rand found the effect after a single exposure in their particular materials and samples. That does not mean repetition guarantees belief, that every person responds alike, or that familiarity overwhelms strong contrary evidence. It means familiarity is one cue that can be misread as truth under some conditions.

Emotion and attention. Moral-emotional language has been associated with wider diffusion inside particular online networks, and novelty may help explain why false news travelled differently in a large Twitter dataset. These are not universal laws. An angry post may stall; a sober correction may travel. Content, community and platform affordances interact, while observational data cannot cleanly isolate every cause.

Social proof and identity. Visible shares, endorsements and group cues can signal that a claim matters or belongs. But identity-based explanations compete with inattention accounts: sometimes people protect a valued group position; sometimes they simply do not pause to consider accuracy. The same person may do both in different situations. “Bias” is not a diagnosis and should not become a label for people one disagrees with.

Correction and continued influence. A correction can reduce belief without erasing all downstream influence. Reviews recommend clear warnings, an explanation of what was wrong, credible evidence and—when possible—a coherent alternative account. Repeating a false claim unnecessarily can also increase familiarity, so corrections should lead with the accurate account and name the falsehood only as much as needed for clarity.

B / Competing perspectives

Why does questionable information move?

Lens 1: attention and reasoning

One research programme argues that many people do not primarily share false content because they prefer falsehood. Their attention is diverted from accuracy by novelty, social reward or the mechanics of the feed. Accuracy prompts can improve discernment in some experiments. This lens suggests friction, reflection and better interface cues.

Lens 2: motivation and belonging

Another tradition emphasizes goals: defending identity, expressing loyalty, sustaining a worldview or participating in a group. Under this lens, a factual correction may fail when it threatens the social function of the claim. Responses should consider trust, messengers and relationships—not only append more facts.

These lenses are not mutually exclusive. Evidence varies by topic, population and method. A third structural lens asks what systems amplify: recommender design, advertising incentives, elite cues and unequal access to reliable local information can shape exposure before an individual ever makes a judgment. Good analysis keeps person, message and system in the same frame.

C / Case files

Two events, two kinds of amplification

Case 01 · 1938 · broadcast + newspapers

The War of the Worlds: the panic and the story about the panic

On 30 October 1938, CBS aired Orson Welles’s radio adaptation of H. G. Wells’s novel, shaped in part like breaking-news bulletins. Archival letters and contemporary records show that some listeners were frightened and that institutions received anxious calls. But the later image of a nation in mass flight rests on weak audience estimates, a small and selected interview sample, and sensational newspaper coverage.

Hadley Cantril’s 1940 study interviewed 135 people, including 100 known to have been upset. It asked a valuable question—why did some listeners treat the drama as news?—but could not establish a national prevalence from that sample. Later scholarship by Jefferson Pooley and Michael Socolow challenged the inflated mass-panic narrative and examined how early audience research and press-radio competition helped stabilize it.

Transmission lesson: the event generated at least two messages: the fictional invasion and the enduring claim that a whole nation panicked. Evidence supports real confusion among some listeners; the scale and meaning of “mass panic” remain contested. The correction is not that “nothing happened,” but that vivid anecdotes outran representative evidence.

Case 02 · 2013 · crisis + networked crowds

Boston Marathon bombing: collaborative attention becomes misidentification

After the 15 April 2013 bombing, online participants examined photographs, assembled clues and circulated candidate identities while official information was scarce. Some activity helped surface information; some produced false accusations and harmful attention toward innocent people. The FBI’s 18 April briefing explicitly told the public that its released images were the only ones that should be treated as credible, warning that other photos could divert attention and create unnecessary work.

Maddock, Starbird and colleagues later analysed four Twitter rumours from the crisis. Their descriptive study showed different temporal, lexical and source patterns rather than one universal “rumour curve.” That matters: denial, speculation, conspiracy and misidentification can have different signatures and require different responses.

Transmission lesson: uncertainty plus urgency can turn collective investigation into a confidence cascade. Provenance, calibrated language and a route for submitting evidence privately are safety mechanisms—not cosmetic disclaimers. This case does not show that crowds are always wrong or that official sources are infallible; it shows the cost of publishing an unverified identification.

D / Evidence ledger

Keep finding, interpretation and myth apart

Established in bounded studies

Repetition can increase perceived truth.

Laboratory results support an illusory-truth effect across many statement types. Size and persistence depend on materials and context; the effect is not mind control.

Large observational finding

False news diffused differently in one major Twitter dataset.

Vosoughi, Roy and Aral studied verified true and false stories posted from 2006–2017. False stories reached farther and faster in that setting. This is not a timeless rate for all platforms or topics.

Contested interpretation

Motivated reasoning explains susceptibility.

Identity-protective cognition explains some results; analytic-attention accounts explain others. Measures, topics and samples affect which account fits.

Myth to retire

Only gullible people share misinformation.

Sharing can reflect attention, humour, outrage, signalling, uncertainty, trust or interface friction. A share does not reveal one motive, and intelligence is not an immunity badge.

Plausible hypothesis

Adding friction will improve every information environment.

Prompts and pauses help in some studies, but poorly placed friction can delay urgent, accurate information or burden people unevenly. Test the intervention in context.

Fictional simulation

The scores in the Rumour Mutation Network.

They are transparent comparisons generated by this site’s rules. They do not estimate human behaviour, truth, virality or risk.

E / Field protocol

The TRACE pause: sixty useful seconds before forwarding

  1. T — Type the claim.Reduce the item to one checkable sentence. Separate the factual claim from opinion, prediction and emotional framing.
  2. R — Recover the source.Open the original, not a screenshot of it. Check author, date, full context and whether the source is in a position to know.
  3. A — Ask what is missing.Look for scope, denominators, comparison points, uncertainty and what evidence would change the conclusion.
  4. C — Check laterally.Leave the page. See whether independent sources and relevant primary records support the same specific claim.
  5. E — Express proportionally.Share with calibrated language: known, reported, plausible, disputed or unknown. If the identification could harm someone, do not publish it.

Guided reflection: reconstruct one recent relay

Choose a non-sensitive message you recently received. Write the earliest version you can locate and the version you remember. Circle what was lost; underline what was added; box every source cue. Then answer: What action did the message invite? Which uncertainty mattered? Who could verify it? What wording would preserve urgency without overstating certainty? The aim is not to catch a liar. It is to see the system around the sentence.

F / Ethical limits

Verification should reduce harm, not become surveillance.

Do not diagnose a person from what they share, infer a hidden motive, or turn media literacy into a test of worth. People may be responding to incomplete local information, language barriers, historical distrust or a real need for safety. Correct the claim while preserving dignity. When a rumour names a private person, pause publication, preserve evidence, use an appropriate private reporting route and consider the asymmetric harm of being wrong.

Institutional communicators also carry duties: say what is unknown, time-stamp updates, keep corrections attached to the original record, and explain why advice changed. Transparency is not the same as dumping every uncertain detail. In emergencies, concise and actionable information can be more ethical than exhaustive commentary. When certainty changes, show the revision trail.

Six questions worth keeping open

FAQ

Is a rumour always false?

No. “Rumour” describes the information’s unverified status and social circulation, not its final truth value. A rumour may later be confirmed, disproved or found to mix accurate and inaccurate elements. Treating every rumour as false can be as careless as treating it as fact.

What is the difference between misinformation and disinformation?

Misinformation is false or misleading information shared regardless of intent. Disinformation is deliberately deceptive. The distinction matters for accountability, but intent is often difficult to establish from a post alone. Focus first on the claim, evidence and likely harm; avoid mind-reading.

Does repeating a false claim—even to correct it—make things worse?

Repetition can increase familiarity, but a clear correction can still improve accuracy. Lead with the fact, explain the error, provide a coherent alternative, use credible evidence, and avoid gratuitous repetition. The blanket advice “never repeat the myth” is too simple for every context.

Do corrections backfire?

They can provoke resistance in some situations, but large later studies suggest attitude backfire is not the normal outcome. Corrections vary in messenger, tone, timing and evidence. Design for comprehension and dignity, then test whether the intended audience understood the accurate account.

Are social platforms the whole problem?

No. Rumours predate digital media. Platforms can alter speed, visibility, persistence and incentives, while journalism, political elites, private groups, institutions and face-to-face conversation also shape transmission. Blaming one channel hides the wider information ecology.

Can this exhibit tell me whether a post is true?

No. It is not a fact-checker, risk score or diagnostic tool. It makes a few transmission pressures visible through authored rules. To evaluate a real claim, trace primary evidence, compare independent expertise, examine context and keep uncertainty explicit.

Further reading / URLs checked 28 October 2026

Follow the evidence upstream.

These sources were selected for relevance and methodological range. A link is not an endorsement of every inference made in a paper; read methods, populations and limitations.

  1. 01

    Vosoughi, S., Roy, D. & Aral, S. (2018). “The spread of true and false news online.” Science, 359(6380), 1146–1151. Large observational analysis of Twitter rumour cascades; platform- and period-specific.

  2. 02

    Pennycook, G., Cannon, T. D. & Rand, D. G. (2018). “Prior exposure increases perceived accuracy of fake news.” Journal of Experimental Psychology: General, 147(12), 1865–1880. Experiments on familiarity and perceived accuracy, with stated boundary conditions.

  3. 03

    Ecker, U. K. H. et al. (2022). “The psychological drivers of misinformation belief and its resistance to correction.” Nature Reviews Psychology, 1, 13–29. Review of cognitive, social and affective drivers plus correction evidence.

  4. 04

    Brady, W. J. et al. (2017). “Emotion shapes the diffusion of moralized content in social networks.” Proceedings of the National Academy of Sciences, 114(28), 7313–7318. Observational and experimental work on moral-emotional language in bounded networks.

  5. 05

    Guess, A., Nagler, J. & Tucker, J. (2019). “Less than you think: Prevalence and predictors of fake news dissemination on Facebook.” Science Advances, 5(1), eaau4586. A reminder that exposure and sharing were concentrated, not universal, in its U.S. election sample.

  6. 06

    Wood, T. & Porter, E. (2019). “The elusive backfire effect: Mass attitudes’ steadfast factual adherence.” Political Behavior, 41, 135–163. Large set of experiments testing whether factual corrections routinely backfire.

  7. 07

    Roozenbeek, J. & van der Linden, S. (2019). “Fake news game confers psychological resistance against online misinformation.” Palgrave Communications, 5, 65. Large pre/post intervention study with convenience-sample and design limitations.

  8. 08

    Maddock, J. et al. (2015). “Characterizing online rumoring behavior using multi-dimensional signatures.” Proceedings of CSCW ’15, 228–241. Descriptive mixed-method analysis of four Boston Marathon rumours.

  9. 09

    Potter, L. A. (2003). “‘Jitterbugs’ and ‘Crack-pots’: Letters to the FCC about the ‘War of the Worlds’ broadcast.” Prologue, U.S. National Archives. Archival overview including the design limits of Cantril’s interviews.

  10. 10

    Pooley, J. D. & Socolow, M. J. (2013). “Checking up on The Invasion from Mars: Hadley Cantril, Paul Lazarsfeld, and the making of a misremembered classic.” International Journal of Communication, 7, 1920–1948. Historical reappraisal of research and myth-making around the 1938 broadcast.

  11. 11

    FBI Boston (18 April 2013). Remarks on the Boston Marathon bombing investigation. Primary institutional record asking the public to rely on the released suspect images rather than other circulated photographs.

Carry the method, not the certainty

Trace before you transmit.