Epidemic Models¶
Epidemic models from epidemiology have been adapted to model information diffusion on social media and social networks. The core idea is that information spreads through a population of users similarly to how infectious diseases spread: exposure to information creates a probability of adoption, which in turn exposes other users to the same information.
Standard models:
SIR (Susceptible-Infected-Recovered) — The classic compartmental model dividing the population into three states: S (susceptible to infection/unaware of information), I (infected/active adopters sharing information), and R (recovered/no longer actively spreading). The model captures the typical rise and plateau pattern of spreading.
SEIR (Susceptible-Exposed-Infected-Recovered) — An extension that includes an exposed state for users who have seen the information but not yet started sharing.
EXP (Exponential growth) — A simpler model capturing only early-stage exponential growth without the saturation effects of SIR.
Key parameter — R₀ (basic reproduction number) — Represents the expected number of secondary adopters (users who see and share information) generated by one initially infected user in a fully susceptible population. If R₀ > 1, information exhibits epidemic behavior and spreads broadly; if R₀ < 1, information naturally dies out.
Application to misinformation — Researchers estimate R₀ for information about specific topics (e.g., COVID-19) or from specific sources (e.g., questionable news sources) to characterize how "infectious" different pieces of information are on each platform.
Key papers¶
- The structure and function of complex networks — Mathematical foundations of epidemic processes spreading on networks
- Cinelli et al. (2020) — The COVID-19 Social Media Infodemic — Applies both EXP and SIR models to COVID-19 information diffusion across platforms, establishing R₀ benchmarks for different platforms and source types
Related topics¶
- Information diffusion in social networks — the phenomenon being modeled
- Infodemic — application domain (health crises)
- Network analysis of misinformation — underlying network structure affects spreading