Abstract
Nonpharmaceutical interventions (NPIs) such as mask wearing can be effective in mitigating the spread of infectious diseases. Therefore, understanding the behavioral dynamics of NPIs is critical for characterizing the dynamics of disease spread. Nevertheless, standard infection models tend to focus only on disease states, overlooking the dynamics of "beneficial contagions," e.g., compliance with NPIs. In this work, we investigate the concurrent spread of disease and mask-wearing behavior over multiplex networks. Our proposed framework captures both the competing and complementary relationships between the dueling contagion processes. Further, the model accounts for various behavioral mechanisms that influence mask wearing, such as peer pressure and fear of infection. Our results reveal that under the coupled disease-behavior dynamics, the attack rate of a disease—as a function of transition probability—exhibits a critical transition. Specifically, as the transmission probability exceeds a critical threshold, the attack rate decreases abruptly due to sustained mask-wearing responses. Our results highlight that without proper enforcement of NPIs, reductions in the disease transmission probability via other interventions may not be sufficient to reduce the final epidemic size.
TL;DR
This research reveals a counterintuitive "control paradox": a more infectious disease can result in a smaller total epidemic because it triggers a massive, sustained social response (like mask-wearing). By modeling disease and behavior as "dueling contagions" on a multiplex network, the authors demonstrate that reducing a virus's transmissibility without maintaining mask mandates can actually increase the final number of infections.
What Problem Does This Paper Address?
Standard epidemiological models, such as the classic SIR model, often treat human behavior as a static background variable. However, during a real-world pandemic, behavior is dynamic and reactive. People start wearing masks when they see others doing so or when they become afraid of rising case counts.
The authors address the gap in modeling these "beneficial contagions"—prosocial behaviors that spread through a population—alongside biological contagion. Specifically, they investigate how the coevolution of mask-wearing and disease spread leads to non-linear outcomes that traditional models miss, such as why infection waves oscillate and why some "less infectious" diseases end up infecting more people in the long run.
Methods
The researchers developed a multicontagion framework using a two-layer multiplex network.
- Layer 1 (Social): Governs the spread of mask-wearing behavior.
- Layer 2 (Biological): Governs the spread of the disease.
The same population exists on both layers, but the connections (edges) can differ. The disease follows an SIR (Susceptible-Infectious-Recovered) process, while mask-wearing is modeled as a reversible complex contagion. An individual decides to wear a mask based on three factors: peer pressure (local threshold), fear (global threshold based on total prevalence), and prosociality (fixed internal trait).
Unlike previous models where behavioral adoption was permanent, this model allows individuals to stop wearing masks if the social or fear-based pressure subsides, allowing for the simulation of multiple infection waves.
Key Insights and Ideas
- The Non-Monotonic Attack Rate: In a standard model, higher transmissibility ($p$) always leads to a higher final "attack rate" (total infected). Here, the curve is non-monotonic. At a certain "tipping point," increasing the infectiousness of the disease actually decreases the final attack rate because it triggers a "saturated" behavioral response that crushes the epidemic early.
- The Persistence-Prevalence Trade-off: "First-regime" diseases (lower transmissibility) are dangerous because they are "stealthy." They don't infect enough people to trigger widespread fear or peer pressure, allowing the disease to persist for a long time through multiple small waves.
- Regime Shifts: The authors identify two distinct regimes. In Regime 1, behavior and disease oscillate. In Regime 2, the disease is so infectious that it triggers a permanent shift in social norms (sustained masking), leading to a single, shorter wave.
Key Results
- Critical Transitions: The authors identified a sharp discontinuity in the attack rate. As the baseline transmission probability ($p$) increases, the attack rate eventually drops abruptly due to the sudden "ignition" of the mask-wearing contagion.
- The Paradox of Intervention: If a policy (like a vaccine or a partially effective treatment) reduces the transmission probability $p$ but doesn't reach herd immunity, it might inadvertently move the population from Regime 2 back into Regime 1. This could lead to a higher final attack rate because the "beneficial contagion" of mask-wearing is no longer sustained.
- Duration vs. Peak: Higher $p$ leads to higher peaks but significantly shorter durations. Conversely, lower $p$ leads to lower peaks but much longer-lasting epidemics that eventually infect a larger cumulative percentage of the population.
Limitations and Future Work
- Permanent Immunity: The model assumes an SIR structure. Future work could explore SIS or SIRS models where waning immunity might interact even more complexly with behavior.
- Asymptomatic Spread: Incorporating asymptomatic spread—where the disease spreads without triggering the "fear" threshold—could change the tipping points.
- Homogeneous Compliance: Future research could look at "habit formation," where the cost of mask-wearing decreases over time, potentially stabilizing the beneficial contagion further.
Summary updated