🗣️ Rumor Spread — Ignorant/Spreader/Stifler Model

The model is a continuous-time Markov chain on the triple (X, Y, Z) representing the counts of Ignorants, Spreaders, and Stiflers in a well-mixed population of size N. Two types of events occur: an Ignorant-Spreader contact converts the Ignorant to a Spreader at rate proportional to X·Y, while a Spreader-Spreader or Spreader-Stifler contact converts the initiating Spreader to a Stifler at rate proportional to Y·(Y+Z-1). Daley and Kendall used a deterministic fluid limit of these equations to derive the asymptotic proportion of the population that remains ignorant as N grows large, a value that depends only on the ratio of the two contact rates, not on the population size itself.

Both models share a three-compartment structure, but the transition mechanism differs fundamentally. In SIR, an Infected individual recovers at a fixed rate independent of who they contact — recovery is an internal clock. In the Daley-Kendall rumor model, a Spreader only becomes a Stifler as a direct result of an unproductive social contact. This makes rumor spread inherently self-limiting through social friction rather than biological recovery, and it is why rumor models tend to leave a much larger untouched "ignorant" population than comparable disease models with similar contact rates.

Yes, significantly. On a well-connected network with short average path lengths, spreaders quickly bump into other informed nodes and stifle rapidly, but they also reach distant parts of the network faster before stifling. On a sparse or clustered network, rumors travel more slowly but spreaders may go longer stretches contacting only Ignorants, potentially reaching a similar or even larger final fraction. Real-world social networks exhibit "small-world" properties — high clustering with short average path length — which studies suggest produces rumor patterns closer to fully-mixed model predictions than a naive lattice model would.

The foundational stochastic treatment came from D. J. Daley and D. G. Kendall in their 1965 paper "Stochastic rumours" in the IMA Journal of Applied Mathematics, building on earlier deterministic rumor models proposed by researchers including Rapoport in the 1950s. The Maki-Thompson variant followed in 1973, refining the stifling rule to the version most commonly simulated today. Both papers drew explicit analogies to the mathematics of epidemic theory developed decades earlier by Kermack and McKendrick, whose SIR framework inspired the Ignorant/Spreader/Stifler naming convention.

Yes. The Daley-Kendall framework and its descendants are widely applied to model misinformation cascades on social media platforms, where "stifling" corresponds to a user losing interest in resharing a stale post. Marketing researchers use the same equations to forecast the reach of viral advertising campaigns, since word-of-mouth promotion follows an almost identical stifling dynamic once a recipient has already heard the pitch from multiple sources. Computer scientists have also adapted variants of the model to describe how patches and security alerts propagate — and stall — across large device networks.

This simulation, like the classical Daley-Kendall model, assumes every contact is equally likely to be made and every individual behaves identically regardless of personality, trust, or media type. Real rumor spread is also shaped by the credibility of the source, the emotional content of the message (highly emotional or surprising rumors spread faster and are stifled less), and asymmetric network structures such as broadcast media or influential hub nodes that can reach thousands simultaneously. More advanced variants of the model incorporate heterogeneous contact rates, weighted trust, and time-varying network structure to capture these effects.