Kinetic SIS opinion-driven models with asymmetric awareness feedback: macroscopic limit and polarization

Imagine a crowded room where everyone is trying to decide whether to wear a mask. Some people are sick, some are healthy, and everyone is constantly talking to each other, sharing news, and reacting to what's happening around them.

This paper is about building a mathematical model to understand how disease and public opinion dance together. It's not just about germs spreading; it's about how our fear (or lack of it) changes how the germs spread, and how the germs change our fear.

Here is the story of the paper, broken down into simple concepts:

1. The Two Main Characters

The model tracks two things for every person in the room:

The Health Status: Are you Healthy (Susceptible) or Sick (Infected)?
The Opinion Scale: Imagine a slider from -1 to +1.
- -1 means: "I don't care, I'll do whatever I want, no masks!" (Reckless).
- +1 means: "I am super cautious, I'll wear a mask and stay home!" (Compliant).

2. The "Asymmetric" Feedback Loop (The Twist)

Usually, we think that if you get sick, you become scared and start wearing a mask. If you stay healthy, you might get lazy and stop wearing one. This paper says: "Yes, but it's not fair."

The authors call this asymmetric awareness:

Getting Sick: When a healthy person gets infected, they get a shock. They immediately swing their opinion toward +1 (becoming very cautious). It's like a "wake-up call."
Not Getting Sick: When a healthy person meets a sick person but doesn't catch the disease, they get a false sense of security. They think, "Wow, I'm lucky! I don't need to worry." They swing their opinion toward -1 (becoming reckless).

The Metaphor: Imagine a game of Russian Roulette.

If you pull the trigger and get shot (get infected), you immediately start wearing a helmet and checking your gun every day.
If you pull the trigger and nothing happens, you think, "This gun is safe!" and you start taking the gun apart and playing with it carelessly.
The paper shows that this "false security" is actually more dangerous because it makes the whole group reckless faster than the "wake-up call" makes them cautious.

3. The "Fast Talk, Slow Sneeze" Rule

In the real world, people talk and change their minds all the time (fast), but catching a disease or recovering takes longer (slow).

The authors realized that because people talk so much faster than they get sick, we can simplify the math. Instead of tracking every single conversation, we can assume that opinions instantly settle into a pattern before the next person gets sick.

Without "Self-Thinking" (Noise): If everyone just listens to each other, they all eventually agree on one opinion (Consensus). The whole room becomes either all cautious or all reckless.
With "Self-Thinking" (Noise): People also have their own internal thoughts. Sometimes they ignore the crowd. This creates Polarization. The room splits: some people are extreme "mask-haters" and others are extreme "mask-lovers," with very few people in the middle.

4. The "Reduced" Model (The Cheat Sheet)

The original math is incredibly complex (like trying to track every single grain of sand on a beach). The authors did the hard work to create a simplified version (a "cheat sheet") that only tracks three numbers:

How many people are sick?
What is the average opinion of the healthy people?
What is the average opinion of the sick people?

They proved mathematically that this simple cheat sheet gives almost the exact same answer as the complex beach-of-sand model, provided people talk fast enough.

5. What the Simulations Showed

The authors ran computer simulations to see what happens in different scenarios:

The "Overconfident" Trap: If the "reckless" reaction (swinging to -1 when you don't get sick) is strong, the epidemic can actually get worse. People become so confident that they stop protecting themselves, leading to a massive outbreak.
The "Polarization" Effect: If people have strong internal opinions (noise), the group splits. The sick people might become very cautious, but the healthy people might become very reckless. This split makes it harder to control the disease because the two groups are moving in opposite directions.
The Sweet Spot: The model suggests that to stop an epidemic, we need to make sure that the "wake-up call" of getting sick is stronger than the "false confidence" of staying healthy.

The Big Takeaway

This paper teaches us that epidemics are not just biological; they are psychological.

If we only focus on the virus, we miss the most important part: how people react to it.

If people get scared only when they get sick, but get overconfident when they don't, the virus will spread faster.
To win the fight, public health messages need to counteract that "false confidence." We need to remind people that not getting sick today doesn't mean the virus is gone; it just means they got lucky.

The authors built a mathematical "lens" that lets us see these invisible social forces, helping leaders design better strategies to keep both the virus and the panic in check.

Here is a detailed technical summary of the paper "Kinetic SIS opinion-driven models with asymmetric awareness feedback: macroscopic limit and polarization" by Pinasco, Saintier, Tettamanti, and Zanella.

1. Problem Statement

The paper addresses the complex interplay between epidemic spreading and social opinion dynamics. Traditional compartmental models (e.g., SIS, SIR) treat population behavior as static or discrete, failing to capture the continuous evolution of individual attitudes toward non-pharmaceutical interventions (NPIs).

The core problem is to model a bidirectional feedback loop:

Opinion $\to$ Epidemic: An individual's compliance with protective measures (modeled as a continuous opinion variable) influences their probability of transmitting or contracting the disease.
Epidemic $\to$ Opinion: The outcome of physical interactions (infection vs. non-infection) asymmetrically reshapes individual opinions. Specifically:
- Infection: Induces caution, shifting opinions toward protective behavior.
- Failed Transmission (No infection): Induces recklessness, shifting opinions toward extreme non-compliance (polarization).

The authors aim to derive a rigorous macroscopic description of this system from a microscopic kinetic framework, specifically investigating the regime where social interactions occur much faster than epidemic transitions.

2. Methodology

A. Microscopic Kinetic Framework

The authors introduce a multi-agent system where each agent is characterized by:

Epidemiological State: Susceptible ( $S$ ) or Infected ( $I$ ).
Opinion Variable: A continuous variable $w \in [-1, 1]$ , where $-1$ represents total rejection of protection and $+1$ represents full compliance.

The dynamics are governed by four mechanisms:

Physical Interactions (Contagion): The transmission rate $\kappa(w, w^*)$ $κ (w, w^{*})$ depends on the opinions of the interacting agents. If an $S$ $S$ agent interacts with an $I$ $I$ agent:
- Infection: The agent moves $S \to I$ and shifts opinion toward $+1$ (caution).
- No Infection: The agent stays $S$ but shifts opinion toward $-1$ (recklessness).
Social Interactions: Binary exchanges between agents (modeled via Boltzmann-type operators) drive opinion consensus or polarization. These include a "compromise" term (aggregation) and a "self-thinking" term (stochastic diffusion).
Memory Fading: Agents naturally drift toward recklessness ( $w \to -1$ ) over time due to habituation.
Recovery: Infected agents recovering ( $I \to S$ ) may perceive reduced vulnerability and shift toward recklessness.

The system is described by coupled kinetic equations for the opinion distributions $f^S_t(w)$ and $f^I_t(w)$ .

B. Macroscopic Limit Derivation

The authors analyze the fast social-interaction regime, where the rate of social interactions ($1/\varepsilon $) is much higher than the rates of epidemic transitions ($ \nu, \gamma, 1/\tau$).

Zero-Diffusion Case (Consensus): In the absence of self-thinking noise, the opinion distribution collapses to a Dirac mass at the mean opinion $m_t$ . The system reduces to a set of Ordinary Differential Equations (ODEs) for the infected fraction $\rho^I_t$ and the mean opinion $m_t$ .
Non-Zero Diffusion Case (Polarization): When self-thinking (noise) is present, the opinion distribution converges to a Beta-type distribution centered at the mean. The macroscopic system is derived using a closure relation based on this equilibrium distribution.

C. Rigorous Convergence Analysis

A key methodological contribution is the rigorous proof of convergence from the kinetic system to the reduced macroscopic ODE system.

The authors utilize a modified Wasserstein distance ( $\tilde{W}_2$ ) to quantify the distance between the microscopic opinion distributions and the macroscopic Dirac/Beta approximations.
They prove that under general assumptions on the interaction kernels, the error between the kinetic solution and the reduced ODE solution is of order $O(\sqrt{\varepsilon})$ , providing a solid analytical foundation for the model reduction.

3. Key Contributions

Asymmetric Awareness Feedback: The model introduces a novel, non-symmetric update rule where infection promotes caution, but lack of infection promotes extremism. This captures the psychological phenomenon of "false security" leading to riskier behavior.
Macroscopic Reduction with Error Bounds: The paper derives reduced ODE systems for both consensus (noiseless) and polarization (noisy) regimes and provides rigorous convergence rates using optimal transport metrics (Wasserstein distance).
Beta-Distribution Closure: In the presence of noise, the authors show that the opinion distribution converges to a Beta distribution, allowing for a tractable macroscopic description that accounts for heterogeneity and polarization, rather than assuming a single mean opinion.
Numerical Validation: The authors employ Direct Simulation Monte Carlo (DSMC) with a time-splitting strategy to validate the theoretical derivations, demonstrating consistency between the full kinetic model and the reduced ODEs.

4. Key Results

Emergence of Polarization: The numerical simulations reveal that the interplay between infection-induced caution and non-infection-induced recklessness can lead to bimodal opinion distributions (polarization), particularly when the "self-thinking" noise parameter is high.
Impact on Epidemic Trajectories:
- Low Reaction Parameter: If individuals do not drastically change behavior after a non-infection event, the system tends toward consensus, and the epidemic dies out.
- High Reaction Parameter: If individuals become significantly more reckless after avoiding infection, the mean opinion shifts toward non-compliance. This increases the effective transmission rate, potentially causing a secondary epidemic outbreak or sustaining a higher endemic level, even if the initial infection rate was low.
Consistency of Reduced Models: The reduced ODE models accurately predict the long-time behavior of the full kinetic system, provided the social interaction rate is sufficiently high compared to epidemic rates.
Role of Self-Thinking: The ratio between the tendency to compromise and self-thinking forces ( $\kappa = \sigma^2/\xi$ $κ = σ^{2} / ξ$ ) dictates the equilibrium state:
- $\kappa < 1$ : Consensus formation.
- $\kappa > 1$ : Polarization (opinions concentrate at extremes).

5. Significance

Theoretical Advancement: The work bridges the gap between microscopic agent-based models and macroscopic epidemiological models, offering a mathematically rigorous framework for coupled socio-epidemic systems. The use of Wasserstein distances for convergence proofs in this context is a significant analytical contribution.
Public Health Implications: The model highlights a critical risk: asymptomatic or non-infected individuals may become the primary drivers of an epidemic due to behavioral overconfidence. This suggests that public health strategies must address not just the fear of infection, but also the complacency that arises from not getting sick.
Policy Design: The reduced ODE models offer a computationally efficient tool for sensitivity analysis and optimal control. They allow policymakers to explore how different communication strategies (influencing the opinion update rules) could steer the population toward protective behaviors and minimize epidemic peaks.
Future Directions: The framework is designed to be extensible to more complex compartmental structures (SEIR, SIRS) and can serve as a basis for studying optimal control of opinion to mitigate disease transmission.

In summary, this paper provides a robust mathematical framework demonstrating that behavioral feedback loops driven by epidemic outcomes are a primary determinant of epidemic trajectories, capable of generating complex phenomena like polarization and secondary outbreaks that standard models fail to predict.