🧬 biology

Kinetic models of opinion-driven epidemic dynamics modulated by graphons

This paper introduces and analyzes kinetic models coupling epidemic dynamics with opinion formation on graphon-based social networks, proving convergence to equilibrium and demonstrating how opinion leaders and time-dependent reproduction numbers influence disease spread and generate epidemic waves.

Original authors: Andrea Bondesan, Jacopo Borsotti, Mattia Fontana

Published 2026-04-14

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Andrea Bondesan, Jacopo Borsotti, Mattia Fontana

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ⚕️ This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer

Imagine a massive, chaotic dance floor representing a society during an epidemic. On this floor, people are moving, dancing, and occasionally bumping into each other. Some people catch a "virus" (the disease), while others stay healthy or recover.

But here's the twist: Everyone on the dance floor is also arguing about whether dancing is safe.

Some people are convinced they must wear masks and stay back (protective behavior). Others think the virus is a hoax and dance wildly without care. The paper you're asking about is a sophisticated mathematical recipe that predicts how the virus spreads based on what people believe and who they talk to.

Here is the breakdown of their "recipe" using simple analogies:

1. The Dance Floor is a "Graphon" (The Social Map)

In real life, not everyone talks to everyone. You talk to your neighbors, your friends, and maybe a famous influencer you follow online, but you don't talk to the person in the next city.

The authors use a mathematical tool called a Graphon. Think of this as a super-detailed social map.

Normal People: They are like regular dancers. They talk to their neighbors and are easily influenced by the group.
Opinion Leaders: These are the "VIPs" or "Influencers" on the dance floor. They have thousands of people listening to them, but they rarely listen to anyone else. They can change the mood of the whole room, but their own mood stays the same.

The paper shows that if you have a few VIPs who say, "Hey, wear a mask!", the whole dance floor might calm down and stop the virus. But if those VIPs say, "The virus is fake!", the whole floor might go wild, and the virus spreads like wildfire.

2. The Two-Step Dance (Opinion vs. Virus)

The model treats the spread of the virus and the spread of ideas as two different dances happening at the same time:

The Virus Dance: This is fast. If you bump into an infected person, you might get sick. But how likely you are to get sick depends on your opinion. If you believe in masks, you are less likely to get infected.
The Opinion Dance: This is slower. People change their minds by talking to others. If you talk to someone who is very convinced (a "compromise"), you might move your opinion slightly toward theirs. If you are stubborn (self-thinking), you might stay put or even swing to the opposite extreme.

The authors found that how connected you are matters.

Social Butterflies: People who talk to lots of different people tend to reach a consensus (they agree on a middle ground).
Social Hermits: People who rarely talk to anyone tend to become extreme. They get stuck in their own echo chambers, believing the virus is either 100% deadly or 100% fake, with no middle ground.

3. The "Dangerometer" (Popularity of Ideas)

The paper also tracks how "popular" the idea of the virus being dangerous becomes. Imagine a popularity meter for news articles or videos.

If people think the virus is scary, they share scary news. The "popularity" of the danger goes up.
If people think it's a hoax, they share funny memes. The "popularity" of the danger goes down.

The authors discovered that this popularity follows a specific pattern (like a heavy-tailed curve). This means that while most ideas are average, sometimes a single idea goes viral and reaches a massive number of people, just like a blockbuster movie or a hit song.

4. The Surprise: Waves Without a Push

Usually, when we see waves of infection (like a second or third peak in a pandemic), we blame it on something external: a new holiday, a new variant, or a change in government rules.

But this paper found something fascinating: The waves can happen all by themselves.
Because people's opinions are constantly shifting back and forth (some days they are scared and wear masks, other days they get tired and ignore the rules), the "Effective Reproduction Number" (a fancy way of saying "how contagious the virus is right now") starts to oscillate.

It's like a pendulum. The virus doesn't need an external push to create a new wave; the internal arguing of the population creates the waves naturally.

5. The Computer Simulation (The Proof)

Since you can't actually run a giant experiment with millions of people, the authors wrote a computer program (code they made public) to simulate this dance floor.

They tested what happens if you have Opinion Leaders who support safety. Result: The virus dies out faster.
They tested what happens if the leaders support ignoring safety. Result: The virus spreads much worse.
They proved mathematically that eventually, the system settles down into a "steady state," but it can take a long time and look very messy before it gets there.

The Big Takeaway

The main message is that human behavior is just as important as biology.
You can't stop a pandemic just by understanding the virus; you have to understand the social network.

If you want to stop the virus, you need to encourage social interaction (so people don't become extreme) and empower responsible leaders to guide the conversation.
If people become isolated and extreme, or if the wrong leaders speak up, the virus will keep coming back in waves, even without any new external threats.

In short: To cure the epidemic, you have to cure the conversation.

1. Problem Statement

The paper addresses the critical limitation of classical compartmental epidemic models (like SIR), which often treat human behavior as static or exogenous. In reality, the spread of infectious diseases is heavily influenced by individual opinions regarding protective behaviors (e.g., mask-wearing, vaccination). These opinions evolve dynamically through social interactions, which are structured by complex networks.

The authors aim to develop a mathematical framework that:

Couples epidemic dynamics (SIR compartments) with opinion dynamics.
Accounts for the structure of social networks using graphons (continuous limits of dense graphs), allowing for the modeling of heterogeneous connectivity and the presence of "opinion leaders."
Analyzes the convergence to equilibrium of these coupled systems and quantifies the rates of relaxation.
Investigates how opinion polarization and consensus formation affect the effective reproduction number and the emergence of epidemic waves without external forcing.

2. Methodology

A. Mathematical Modeling

The authors propose a multi-species kinetic model where the state of an agent is defined by:

Compartment: Susceptible ( $S$ ), Infected ( $I$ ), or Recovered ( $R$ ).
Opinion ( $w$ ): A continuous variable in $[-1, 1]$ , where $w=1$ represents strong support for protective behaviors and $w=-1$ represents rejection.
Graphon Position ( $x$ ): A static position in $\Omega=[0,1]$ representing the agent's location in a social network.

The system is governed by the evolution of distribution functions $f_J(t, x, w)$ for each compartment $J \in \{S, I, R\}$ :
$\partial_t f = E(f, f) + \frac{1}{\tau} Q(f, f)$

Operator $E$ : Models epidemiological transitions (infection and recovery). The transmission rate $\beta_T$ depends on the opinions of interacting agents (e.g., protective agents have lower transmission risk).
Operator $Q$ : A Fokker-Planck type operator modeling opinion dynamics. It includes:
- Compromise dynamics: Agents adjust opinions based on interactions with others, modulated by a graphon $B(x,y)$ (network connectivity) and an interaction function $P(x,y)$ (influence strength).
- Self-thinking (diffusion): Random fluctuations in opinion.
Timescales: The parameter $\tau$ separates the fast timescale of opinion exchange from the slower epidemic spread.

B. Graphon and Propensity to Interact

The paper introduces a propensity to interact $p(x)$ , derived from the graphon $B$ and interaction function $P$ . This quantity unifies the effects of network connectivity and influenceability.

Opinion Leaders: Modeled by specific choices of $B$ and $P$ where highly connected agents ( $x \approx 0$ ) have a low propensity to change their own opinion ( $p(x) \approx 0$ ) but high influence on others.
Simplification: A simplified model is derived where the complex graphon structure is replaced by the scalar propensity $p(x)$ , allowing for explicit calculation of equilibria.

C. Numerical Scheme

The authors design a structure-preserving numerical scheme based on operator splitting:

Opinion Step: Solved using a semi-implicit Chang-Cooper scheme (a finite volume method for Fokker-Planck equations). This ensures mass conservation, non-negativity of solutions, and correct long-time asymptotic behavior.
Epidemic Step: Solved using a classical fourth-order Runge-Kutta method.
Coupling: The two steps are coupled via operator splitting over time intervals.

3. Key Contributions

Unified Kinetic Framework: The introduction of a kinetic SIR model on graphons that simultaneously tracks epidemic states and opinion distributions, generalizing classical ODE-based SIR models.
Analytical Convergence Proofs:
- $L^1$ Convergence: Proved that solutions to the simplified kinetic model converge to equilibrium in the strong $L^1$ norm using relative entropy methods (Csiszár–Kullback–Pinsker inequality). The rate of convergence is shown to be $O(1/\sqrt{t})$ .
- Sobolev Convergence: For a model tracking the "popularity" of disease-related information (a secondary Fokker-Planck equation), convergence is proved in homogeneous Sobolev spaces $\dot{H}^{-s}$ for $s \in (1/2, 1)$ using Fourier-based techniques.
Characterization of Equilibria:
- Derived explicit Beta distributions for opinion equilibria in the simplified model.
- Demonstrated that consensus formation occurs for agents with high propensity to interact, while opinion polarization (extreme opinions) occurs for agents with low propensity to interact.
- Showed that Inverse Gamma distributions arise as equilibria for the popularity of disease-related information.
Numerical Innovation: Development of a novel, publicly available MATLAB code implementing a structure-preserving scheme capable of handling coupled kinetic and non-kinetic operators with multiple distribution functions.

4. Results and Findings

Opinion Polarization vs. Consensus: The simulations confirm that social isolation (low propensity to interact) leads to opinion polarization, whereas high social activity promotes consensus. Polarized populations are more susceptible to extreme anti-protective views.
Role of Opinion Leaders:
- Without Leaders: If highly connected individuals are not opinion leaders, the population tends to form a global consensus against protective behaviors (if the initial majority is skeptical), leading to severe epidemic outbreaks.
- With Leaders: If highly connected individuals act as opinion leaders supporting protective behaviors, they can steer the entire population toward a protective consensus, significantly reducing the infection peak and total cases.
Epidemic Waves without External Forcing: The model generates oscillating effective reproduction numbers ( $R_{eff}$ ) and multiple epidemic waves purely from the internal dynamics of opinion-epidemic coupling. This occurs due to the competition between polarized groups (pro- and anti-protective) within the population, eliminating the need for exogenous time-dependent forcing (like seasonal changes or policy shifts) to explain waves.
Popularity of Information: The "popularity" of the idea that the disease is dangerous follows a fat-tailed distribution (Inverse Gamma), consistent with empirical data on viral content and scientific citations.

5. Significance

Theoretical Advancement: The paper bridges the gap between kinetic theory, graphon theory, and mathematical epidemiology. It provides rigorous proofs of convergence for these complex, non-linear coupled systems, a area where such results are scarce.
Policy Implications: The results highlight the critical importance of opinion leaders in pandemic management. It suggests that targeting highly connected individuals with protective messaging is more effective than broad, undifferentiated campaigns, as it can prevent polarization and drive global consensus.
Mechanism for Waves: The finding that epidemic waves can arise naturally from opinion dynamics offers a new explanation for recurrent outbreaks that does not rely on external factors, suggesting that managing the social dynamics is as crucial as managing the biological transmission.
Open Science: The authors provide the full MATLAB code, enabling reproducibility and further exploration of kinetic epidemic models by the scientific community.

In summary, this work provides a robust mathematical and numerical framework to understand how social network structures and opinion dynamics drive epidemic outcomes, offering deep insights into the mechanisms of consensus, polarization, and the emergence of epidemic waves.