Rumor spread in social network¶

Rumor spread in a social network models propagation through stochastic individual interactions or aggregate population compartments.

Version: v1 · 2026-09-28 · History
Domain-specific #: 7749
Domain group: Interdisciplinary & Synthetic
Origin domain: Systems Thinking & Cybernetics
Subdomain: Social Network Modeling → Systems Thinking & Cybernetics

Core Idea¶

Rumor spread in a social network is the modeled propagation and cessation of an unverified message through contacts among socially connected individuals.^[1] The abstraction represents people as nodes, social contacts as edges, and message-related states or behaviors as variables whose transitions depend on encounters, influence, and network structure.^[2]

Macroscopic models compress the population into compartments. In the Daley–Kendall and Maki–Thompson tradition, individuals are ignorants who have not heard the rumor, spreaders who transmit it, or stiflers who know it but no longer transmit it.^[3] Contact between a spreader and an ignorant can create a new spreader, while contact with someone who already knows the rumor can turn a spreader into a stifler.^[4] The total population is conserved as agents move between states, and rate equations describe aggregate trajectories.^[5]

Network models retain who can contact whom. A node's state changes through pairwise encounters along edges, so degree, clustering, directionality, tie weights, and small-world shortcuts alter reach and extinction.^[6] Dense local clustering can exhaust spreaders into stiflers, whereas a small number of long-range ties can carry the rumor into distant regions.^[7] Microscopic approaches instead model “who influenced whom,” including independent-cascade rules, threshold rules, and behavior-sensitive transmission probabilities.^[8]

This is a family of formal propagation models, not a claim that rumors literally behave like pathogens or that every kind of information diffusion has the same transition rules.^[9] An analysis must declare the node states, contact process, transmission and cessation rules, network topology, seeding, and stopping condition.^[10] Without those components, an observed increase in discussion is a description of attention rather than an instantiated rumor-spread model.

Structural Signature¶

Sig role-phrases:

the social contact structure — the population or graph determining which individuals can encounter or influence one another
the rumor state partition — the declared statuses of individuals, such as ignorant, spreader, and stifler
the seeded message — the unverified content initially held by one or more transmitting individuals
the transmission transition — contact under which an ignorant becomes a spreader with the model's stated rate or probability
the cessation transition — encounter or condition under which a spreader becomes a stifler and stops retransmitting
the representation branch — compartment models aggregate state counts, network models preserve contact topology, and microscopic rules retain particular activation events
the topology response — changed reach or extinction produced by clustering, path length, degree structure, or seed placement under the chosen transitions
the terminal outcome — final reach, remaining ignorants, and extinction time relative to the declared stopping condition
the model boundary — an attention curve or epidemic analogy alone does not instantiate rumor spread without explicit states, contacts, transitions, seeds, and cessation rules

What It Is Not¶

Not a determination that the message is false or true. The model concerns circulation and cessation of an unverified message; epistemic accuracy requires separate evidence.^[11]
Not every instance of information diffusion. Innovations, verified news, advertising, and behavior adoption can use related propagation mathematics while carrying different states and transition meanings.
Not a claim that rumors literally behave like pathogens. Epidemic compartments are a modeling analogy whose rumor-specific stifling and retransmission rules must be stated rather than imported automatically.
Not an attention or popularity curve by itself. Rising discussion does not instantiate a rumor-spread model without declared agents or compartments, contacts, transitions, seeds, parameters, and a stopping rule.
Not one mandatory state system or level of representation. Ignorant–spreader–stifler compartments, explicit networks, independent cascades, thresholds, and behavioral models retain different information and cannot be exchanged without changing the model.
Not a reach or extinction outcome determined by message content alone. Network topology, contact process, seed placement, transmission and cessation rules, external media, and time scale jointly control the predicted trajectory.

Scope of Application¶

Rumor-spread modeling applies where an unverified message, a social contact structure, initial seeds, individual or aggregate states, transmission and cessation rules, and a stopping condition are all declared. Its literal scope covers propagation questions under those commitments; an attention curve or generic information-diffusion label without a rumor-specific transition mechanism does not enter the model.

Daley–Kendall compartment models — populations move among ignorant, spreader, and stifler classes through pairwise contact and rumor-specific cessation rules.
Maki–Thompson variants — directed contact and asymmetric stifling rules alter which participant changes state after spreader–spreader encounters.
Aggregate trajectory analysis — rate equations track conserved population shares, spreader prevalence, final reach, and extinction under a well-mixed approximation.
Explicit social-network models — nodes and edges retain who can contact whom, so state transitions depend on adjacency as well as current rumor status.
Topology-effect studies — degree, clustering, path length, direction, weights, and rewiring are varied to assess reach, repeated-contact stifling, or extinction.
Discrete-time stochastic simulation — node pairs and probabilistic transitions are sampled step by step until no spreaders or qualifying activations remain.
Microscopic influence analysis — individual-level models address who influenced whom and preserve activation paths that compartment totals discard.
Independent-cascade formulations — newly active nodes receive declared probabilistic opportunities to activate inactive neighbors before the process terminates.^[12]
Linear-threshold formulations — a node activates when the accumulated weight of active neighbors reaches its assigned threshold.^[13]
Online social-network studies — rumor propagation is modeled under platform contacts, message exposure, user behavior, and measured or assumed retransmission probabilities.
Dynamic and multilayer networks — changing ties and multiple interaction layers are retained when a fixed single graph would omit important contact routes.
Rumor-mitigation analysis — interventions are evaluated by modifying seeds, network structure, contact rates, stifling rules, or behavior parameters within the declared model.

Clarity¶

Rumor-spread modeling turns a rise in discussion into an explicit propagation process. It requires declared states, contacts, transition rules, network structure, seeds, and a stopping condition; an observed popularity curve without those commitments is evidence about attention, not yet an instance of the model. The label also prevents “rumor behaves like disease” from substituting for the specific social rule that a spreader may stop transmitting after meeting someone who already knows the message.

Macroscopic and network accounts answer different questions. Compartment models track aggregate movement among ignorants, spreaders, and stiflers, while network models retain which pairs can interact and therefore expose effects of degree, clustering, direction, and long-range ties. The analyst’s sharper question is: who can influence whom, with what probability and cessation rule, and does the claimed reach follow from those transitions on the stated network?

Manages Complexity¶

Actual rumor circulation can involve large populations, repeated contacts, changing attention, heterogeneous relationships, and many possible chains of influence. A rumor-spread model makes that sprawl tractable by retaining a smaller state-transition system: who has not heard, is spreading, or has stopped spreading; who can contact whom; the transmission and stifling rates or probabilities; the initial seeds; and the stopping condition. At the compartment level, population conservation and a few rate equations replace individual histories, letting an analyst read off aggregate trajectories and final reach.

Network and microscopic formulations retain more structure when aggregation would erase the question of interest. They track node states and permitted contacts, so degree, clustering, direction, edge weight, and long-range ties can explain why a rumor remains local, crosses into a distant region, or extinguishes after spreaders repeatedly meet people who already know it. Independent-cascade, threshold, and behavior-sensitive variants form distinct branches because they reduce the interaction history with different activation or cessation rules.

The compression ends where those choices affect the result. A compartment model cannot recover a particular influence path, and a fixed-network model does not automatically capture changing ties, message credibility, individual judgment, or external media. Reach and extinction are therefore readable only relative to the declared states, contact process, topology, parameters, and time scale; they are not properties of the rumor alone.

Abstract Reasoning¶

A declared rumor model supports a transition-to-trajectory inference. From the current counts or node states, permitted contacts, and transmission and stifling rules to the next-state distribution, an analyst can derive how the population of ignorants, spreaders, and stiflers changes. Repeating that move yields predicted reach and extinction behavior; an observed attention curve without the transition mechanism cannot support the same inference.

The abstraction also permits topology intervention. From increasing local clustering to more repeated contacts among people who already know the rumor, a Daley–Kendall-like rule can predict faster creation of stiflers and earlier extinction. From adding long-range ties to shorter paths into previously distant regions, the same rules can predict wider reach. These are conditional predictions: changing from pairwise stifling to an independent-cascade or threshold rule can reverse which feature matters.

Model choice can be diagnosed from the question. From a need to predict aggregate prevalence to a compartment model, individual paths may be deliberately discarded; from a need to identify who influenced whom or which seed reaches a region to a network or microscopic model, adjacency and node state must be retained. Disagreement between an aggregate forecast and network data can therefore signal lost heterogeneity rather than a simple parameter error.

These inferences stop at the declared model. They do not establish a rumor's truth or credibility, and reach is not a property of the message alone. Changing the contact process, network, initial seeds, individual behavior, external media, or time scale changes the predicted outcome and requires a new analysis.

Knowledge Transfer¶

Within social-network modeling, rumor-spread knowledge transfers literally across messages, populations, graph topologies, and microscopic or compartment models when states, contacts, transitions, seeds, and stopping rules are declared. The cargo that carries intact includes ignorant, spreader, and stifler roles where that convention is used, transmission and cessation probabilities, repeated-contact effects, network structure, and final reach. Interventions transfer by changing seeding, clustering, degree structure, or stifling rules and deriving the resulting trajectory.

Beyond rumors, the honest case is (B) shared stochastic propagation mechanism. Epidemics, innovations, and information cascades can use related contact processes, but their state meanings and transition rules are not interchangeable. The home-bound cargo is an unverified message, socially mediated retransmission, attention, credibility, and rumor-specific cessation. Calling every popularity curve “rumor spread” is analogy (A) without a model linking individual interactions to state changes. The stopping boundary is the declared social transition mechanism; superficial curve similarity cannot transfer causal or intervention conclusions.

Examples¶

Canonical¶

In a Maki–Thompson-style population model, begin with one spreader carrying an unverified message and all other people ignorant of it.^[14] When the spreader contacts an ignorant, the ignorant becomes a spreader; when a spreader initiates contact with another spreader or with a stifler, the initiating spreader becomes a stifler.^[15] Every person remains in exactly one of the three compartments, so the total number of ignorants, spreaders, and stiflers is conserved.^[16] Repeated stochastic contacts eventually leave no spreaders, at which point the rumor is extinct even though some people may never have heard it.^[17] The final number reached is therefore a consequence of the declared transmission and stifling rules, not simply a count of initial attention.

Mapped back: The population and its permitted encounters constitute the social contact structure, and ignorant, spreader, and stifler supply the rumor state partition. The initial carrier holds the seeded message; ignorant-to-spreader conversion is the transmission transition, while spreader-to-stifler conversion is the cessation transition. The no-spreader stopping state determines the terminal outcome, and the use of aggregate compartment counts selects the representation branch.

Applied / In Practice¶

A network researcher can run the same declared rumor transitions on two graphs with the same number of nodes and initial seed but different connectivity. In a highly clustered graph, newly created spreaders repeatedly encounter neighbors who already know the rumor, so the stifling rule can extinguish the process within a local region. Rewiring a small number of edges into long-range contacts can carry the message to previously distant regions before those local contacts exhaust it.^[18] The result is conditional: changing to an independent-cascade rule with one activation chance per edge would constitute a different propagation model, so its reach cannot be attributed to clustering under the Maki–Thompson rules.

Mapped back: The two graphs vary the social contact structure while holding the seeded message, the transmission transition, and the cessation transition fixed. The change in reach and extinction is the topology response, and retaining particular nodes and edges chooses the network form of the representation branch. Reporting final reach and extinction time instantiates the terminal outcome; the refusal to infer the result from a popularity curve or another model's rules enforces the model boundary.

Structural Tensions¶

T1: Aggregate tractability versus influence-path fidelity. Compartment models reduce a large population to ignorant, spreader, and stifler counts, making trajectories and final reach analyzable. That reduction discards who contacted whom and can hide degree, clustering, and seed-placement effects. Diagnostic: Does the question concern population totals that aggregation preserves, or individual transmission paths that require an explicit network or microscopic model?

T2: Transmission momentum versus contact-driven stifling. Each spreader–ignorant encounter can enlarge the transmitting population, while encounters with people who already know the rumor can remove spreaders under Daley–Kendall-like rules. The same rise in awareness therefore creates both fuel and inhibition. Diagnostic: Under the declared contact rule, which encounter types create spreaders and which end retransmission, and how do their rates shape extinction?

T3: Local clustering versus long-range reach. Dense neighborhoods create repeated exposures that may accelerate stifling and contain the rumor locally, whereas sparse long-range ties can deliver it into distant regions before local exhaustion. Connectivity can amplify or suppress spread depending on the transition rules. Diagnostic: Does the predicted topology effect follow from the stated transmission and cessation mechanism rather than from clustering or short paths alone?

T4: Model realism versus identifiability. Adding heterogeneous behavior, weighted ties, dynamic layers, hesitation, or memory can represent social interaction more faithfully, while each new state and parameter increases the evidence needed to distinguish mechanisms. A simple model is estimable but may be structurally wrong. Diagnostic: Which added feature changes a target prediction and is supported by observations capable of separating it from a rate adjustment in a simpler model?

T5: Intervention leverage versus rule dependence. Changing seeds, contact structure, or cessation probabilities can reduce reach in one formalization, yet an independent-cascade, threshold, or compartment model may respond differently to the same intervention. Optimization inside a misspecified rule system produces precise but fragile advice. Diagnostic: Does the proposed intervention remain effective across plausible transition models, or is it conditional on one unverified activation or stifling rule?

T6: Propagation evidence versus truth assessment. Reach, velocity, and influence paths reveal how an unverified message circulates, while none of those quantities determines whether its content is true. Separating circulation from verification preserves model clarity but leaves an important social judgment outside the model. Diagnostic: Is a claim about message accuracy supported by independent evidence rather than inferred from popularity, persistence, or network position?

T7: Rumor Spread autonomy versus reduction to Contagion (Contagion). The parent Prime carries the portable propagation of an affected state through contacts among susceptible carriers, with transmission, cessation, and burn-out or persistence. Every covered Rumor Spread model is a strict kind of Contagion because spreaders reproduce the rumor state in contacted ignorants and eventually become stiflers, but the child additionally requires an unverified message, social-network topology, rumor-specific state semantics, and reach outcomes. Reduction loses those social-semantic roles; total autonomy hides the general contagion structure. Diagnostic: Does the account preserve the message and rumor-state semantics as differentia of this Contagion, rather than imposing a Fickian Diffusion model?

Structural–Framed Character¶

Rumor Spread in a Social Network is mixed-structural: its propagation dynamics instantiate Contagion, while its identity also requires a socially transmitted, unverified message and rumor-specific cessation semantics. Susceptible contacts, affected transmitters, reproduction through links, removal from transmission, topology, and extinction supply the smallest reviewed Prime skeleton. The cross-domain reach belongs to that Prime. Ignorant, spreader, and stifler states, message credibility, social ties, and the distinction among compartment, network, and microscopic models remain domain-bound.

Its evaluative_weight is medium-low because propagation can be modeled neutrally, although labeling content a rumor carries an epistemic qualification and mitigation questions introduce practical stakes. Its human_practice_bound is high because the spreading state consists of communication and retransmission by social actors. Its institutional_origin is low: platforms and organizations shape contact structures, but no institution constitutes the general model family. Its vocab_travels is medium: node, edge, transition, seed, reach, and extinction move across propagation models, while ignorant, spreader, stifler, rumor, and credibility preserve the social-network frame. Its import_vs_recognize judgment is mixed because explicit state transitions and topology can be recognized in a declared model, whereas choosing rumor states and cessation rules imports a modeling frame rather than reading them directly from an attention curve.

Its character: Contagion owns the portable contact-mediated reproduction-and-cessation skeleton, while the child fixes what spreads, what social states mean, and which transition semantics qualify as rumor circulation. Removing those social commitments leaves general contagion; removing the contagion structure leaves discourse about attention or communication without the candidate's formal propagation identity.

Structural Core vs. Domain Accent¶

Rumor Spread in a Social Network is a domain-specific specialization of the Prime Contagion: an affected state crosses contact links, reproduces in new hosts, and may either become self-sustaining or die out. The specialization fixes the carrier to social actors and the state to circulation of an unverified message.

What is skeletal (could lift toward a cross-domain prime). Contagion supplies susceptible, transmitting, and removed or nontransmitting states; a contact topology; a per-contact transition; reproduction of the affected state in newly reached hosts; cessation or removal; and an outbreak-versus-extinction regime governed by rates and connectivity. That complete signature recurs in at least three unrelated domains—for example, pathogen transmission through interpersonal contact, a computer worm reproducing through network links, and financial distress propagating through counterparty exposures. Rumor models fill those same roles with people and social ties, so the full contact-mediated reproduction-and-cessation skeleton remains after the rumor vocabulary is stripped.

What is domain-bound. The social-network accent supplies a seeded unverified message, people represented as nodes, communication or influence ties as edges, and the ignorant–spreader–stifler partition. It fixes transmission as hearing and retransmitting the message, cessation as becoming informed but no longer spreading it, and the model branches among aggregate compartments, explicit networks, and microscopic activation rules. Credibility, social behavior, degree, clustering, seed placement, and the chosen stopping condition determine how the formal states are interpreted. Remove these commitments and the remainder is general contagion, not a rumor-spread model.

Why this does not clear the prime bar. Stripping the social and epistemic accent leaves a portable Prime already established independently across disease, software, and finance; it does not preserve what counts as a rumor, a spreader, or a stifler. Conversely, keep an attention curve, a message, and a social graph but remove contact-mediated state reproduction and cessation, and the case may describe discussion or information exposure without instantiating this model family. The two removal directions therefore justify strict subsumption: Contagion is the autonomous cross-domain structure, whereas Rumor Spread in a Social Network is its socially interpreted, transition-specific specialization.

This entry is a kind of Contagion.

Instantiates — Contagion (Contagion). A rumor state crosses a social contact link from a spreader to an ignorant, reproduces in the newly informed person so that onward transmission becomes possible, and ceases when a spreader becomes a stifler. Contact topology and transition rates determine whether the seeded process expands or burns out. The ignorant–spreader–stifler partition is the rumor-specific specialization of Contagion's susceptible–affected–removed roles; removing the unverified-message semantics leaves the full contact-mediated, self-reproducing state-spread signature.

Related to — Propagation (Propagation). Rumor models also track a source, paths through a network, reach, and stopping conditions. Propagation is broader, however, because it does not require each newly affected host to reproduce the state through contact.

Related to — Cascade (Cascade). Independent-cascade and threshold variants can satisfy Cascade's local-trigger, re-emission, and damping structure. That relation is model-dependent rather than universal across all compartment and stifling formulations covered by this entry.

Decline — Diffusion (Diffusion). The rumor model does not require down-gradient flux, stochastic particle microdynamics, a permeability coefficient, or square-root distance–time scaling. Its conserved population changes state through contact; it is not redistribution of a fixed quantity along a concentration gradient.

Relationships to Other Abstractions¶

Current abstraction Rumor spread in social network Domain-specific

Parents (1) — more general patterns this builds on

Rumor spread in social network is a kind of Contagion Prime

A rumor state crosses a social contact link from a spreader to an ignorant, reproduces in the newly informed person so that onward transmission becomes possible, and ceases when a spreader becomes a stifler.

Hierarchy path (1) — routes to 1 parentless root

Rumor spread in social network → Contagion → Associative Property Transfer

Neighborhood in Abstraction Space¶

Rumor spread in social network sits in a sparse region of the domain-specific corpus (83^rd percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (2551 abstractions)

Nearest neighbors

Computed from structural-signature embeddings · 2026-10-08

Not to Be Confused With¶

Epidemic spread. Epidemic models track transmission of a biological infection, whereas rumor models reinterpret contact states and add rumor-specific cessation when spreaders meet people who already know the message. Tell: ask whether the affected state is infection or willingness to retransmit unverified content.
Information diffusion. Information diffusion covers circulation of verified news, innovations, advertising, and other content, while rumor spread requires an unverified message and declared rumor-state transitions. Tell: ask whether truth status and rumor-specific spreading and stifling rules are constitutive or incidental.
An independent-cascade model. Independent cascade gives each newly active node a specified opportunity to activate neighbors; it is one possible microscopic propagation rule, not the whole family of rumor models. Tell: ask whether activation follows that one-chance rule or an ignorant–spreader–stifler contact process.
A linear-threshold model. A threshold model activates a node when weighted active-neighbor influence reaches its threshold, rather than through the pairwise transmission and cessation rules of a classical rumor model. Tell: ask whether state change is caused by aggregate neighbor weight or a specified contact event.
Opinion dynamics. Opinion dynamics models changes among attitudes such as support, doubt, neutrality, or denial; rumor spread models who is ignorant, actively transmitting, or no longer transmitting. Tell: ask whether the state variable is belief position or retransmission status.
Misinformation detection. Detection or fact-checking evaluates a message's veracity or flags content, whereas a rumor-spread model predicts its circulation and extinction without settling truth. Tell: ask whether the output is an epistemic classification or a propagation trajectory.
An attention curve. A time series of mentions or popularity can record aggregate interest without specifying agents, contacts, seeds, transmission, cessation, or topology. Tell: ask whether a state-transition mechanism generates the curve or the curve is merely observed.

References¶

[1] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[2] David Kempe, Jon Kleinberg, and Éva Tardos, “Maximizing the Spread of Influence through a Social Network” (KDD 2003) (source). registry ↩ Show verification details

[3] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[4] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[5] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[6] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[7] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[8] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[9] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[10] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[11] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[12] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[13] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[14] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[15] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[16] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[17] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩

[18] Unverified encyclopedia synthesis; no authoritative source located for the claim as written. ↩