Demographic Dependence of Vaccine Adoption under Opinion Persuasion

Technical Summary: Demographic Dependence of Vaccine Adoption under Opinion Persuasion

Problem Statement
Vaccine hesitancy remains a critical barrier to reducing the burden of infectious diseases, driven not only by logistical factors but by heterogeneous beliefs and trust levels across demographic groups. While public health agencies increasingly deploy targeted communication campaigns, existing models often fail to capture how these interventions propagate through social networks and interact with epidemic dynamics. Previous work typically models behavioral responses as homogeneous functions of prevalence or media signals, overlooking the dependence on social structure, demographics, and the specific framing of policy messages. Furthermore, the interplay between misinformation, demographic homophily, and epidemic spread requires a framework that accounts for "epistemic injustice" and the social organization of credibility.

Methodology
The authors propose SIS-Vo, an epistemically-informed networked model that integrates susceptible-infected-susceptible (SIS) epidemic dynamics with signed opinion dynamics. The model operates on a multi-layer network structure where each node represents a subpopulation characterized by a demographic vector $d_i$.

1. Epidemic Level: The model extends the standard SIS-V framework. Each subpopulation $i$ is divided into susceptible ($s_i$), infected ($x_i$), and vaccinated ($v_i$) shares. Vaccination confers permanent immunity. Crucially, the rate of vaccination adoption is capped by a carrying capacity $\kappa_i$, which is a function of the subpopulation's opinion.
2. Epistemic Level: Opinion dynamics ($o_i$) evolve over a signed interaction network $G_i$. The influence between nodes is determined by demographic homophily, defined as the dot product of their demographic vectors ($\alpha_{ij} = d_i^\top d_j$). This structure allows demographic similarity to shape both the sign and strength of opinion influence.
3. Coupling Mechanisms:
   • Infection to Opinion: The current infection level $x_i$ influences the opinion dynamics, reflecting the rationale that the reality of the disease impacts vaccine attitudes.
   • Opinion to Vaccination: The opinion $o_i$ determines the carrying capacity $\kappa_i(o_i)$, which limits the maximum achievable vaccination rate. The function is defined such that negative opinions reduce capacity, while positive opinions increase it.
4. Control Input (Policy Intervention): A government intervention is modeled as a control input vector $u$. The impact of this message on a specific subpopulation is determined by the projection $u^\top d_i$. This allows for "microtargeting," where the same message can have heterogeneous effects (aligned or anti-aligned) depending on the demographic profile of the audience.

Key Contributions

• Model Formulation: The paper introduces the SIS-Vo framework, which explicitly links demographic vectors to both the topology of the opinion network (via homophily) and the efficacy of policy interventions.
• Fixed-Point Characterization: The authors derive closed-form characterizations for the healthy (disease-free) equilibrium and the endemic equilibrium. They show that in the disease-free state, the system stabilizes when vaccination levels reach $\min(1, \kappa_i^*)$.
• Stability Analysis: A sufficient condition for the local stability of the healthy state is derived. This condition relates the spectral properties of the contact network ($B$) and the recovery rates ($\gamma$) to the steady-state carrying capacities ($\kappa_i^*$). Specifically, the healthy state is stable if the carrying capacities are sufficiently high relative to the network's infection potential.
• Intervention Analysis: The paper formulates an optimization problem to minimize the total weighted infection in the endemic state by selecting an optimal intervention vector $u$. It distinguishes between aligned interventions (where $u$ resonates with the demographic vector, increasing $o_i$ and $\kappa_i$) and anti-aligned interventions (which decrease $o_i$ and $\kappa_i$).

Results
Numerical simulations on networks of varying sizes (5 to 200 nodes) illustrate the model's dynamics:

• Endemic Persistence: In scenarios with heterogeneous opinions, the disease often persists in subpopulations with negative vaccine opinions (low $\kappa_i$), while subpopulations with positive opinions achieve full vaccination and disease eradication.
• Validation: Simulations of the differential equations converge to the theoretical fixed points derived from the algebraic equations, validating the equilibrium analysis.
• Impact of Microtargeting:
  • Aligned Interventions: When policy messages are aligned with demographic vectors, they shift opinions positively, increase carrying capacities, and significantly reduce the steady-state infection proportion.
  • Anti-aligned Interventions: Messages that are misaligned with demographic profiles shift opinions negatively, leading to lower carrying capacities and higher endemic infection levels.
  • Random Interventions: Randomly chosen messages perform worse than aligned strategies but slightly better than anti-aligned ones, suggesting that even non-targeted communication can have some effect, though it is far from optimal.
• Stability Conditions: The simulations confirm that the derived stability condition (Proposition 2) is sufficient but not necessary; systems can remain stable even when the strict inequality is not met.

Significance and Claims
The paper claims that the SIS-Vo framework provides a natural basis for the control-theoretic analysis of vaccination policies. Its primary significance lies in demonstrating that:

1. Demographic Targeting Matters: Suitably targeted policy interventions, acting through opinion dynamics, can stabilize the epidemic process by shifting the system toward the healthy regime.
2. Robustness to Misinformation: The model accounts for how misinformation targeting specific subgroups can be countered by understanding the demographic dependence of opinion formation.
3. Epistemic Foundations: By incorporating epistemic dependence and homophily, the model offers a more realistic representation of how trust and credibility function in social networks compared to homogeneous behavioral models.

The authors conclude modestly, noting that the specific functional relationship chosen between opinion and carrying capacity is arbitrary and that future work should investigate the robustness of these results against different functional forms. They also highlight the potential for comparing different intervention strategies to minimize infection burdens in complex, heterogeneous populations.