Containing the spread of infectious diseases is a challenging problem for both central authorities as well as individuals, especially when the infected individuals are asymptomatic and the decisions need to be made under incomplete information and strategic uncertainty. In this paper, we consider the susceptible-infected-susceptible epidemic model, in which a large population of individuals decide whether to adopt partially effective protection without being aware of their exact infection status. Instead, they receive noisy signals from a strategic sender that convey incomplete information regarding their infection state in the form of action recommendations. These recommendation signals are strategically designed so that it is optimal for the receivers to follow their recommended actions, and the infected proportion at the next time step is minimized. When the sender has accurate information about the receivers’ states, we obtain analytical (closed-form) solutions of the optimal signals under suitable assumptions by exploiting the monotonicity of the cost function in the action recommendations. Finally, we consider the case when the sender is not omniscient and show that there may not exist any action recommendation that strategic followers are guaranteed to follow. We validate the proposed framework via numerical simulations and illustrate that disclosing true infection status to the receivers is not always the optimal choice, instead, strategic action recommendations can lead to a smaller infection prevalence. We also show that when the sender is not omniscient, the infected proportion and the strategies adopted by the receivers may not converge to a steady state. Funding: This research was not supported by any research grant.
This paper presents a comprehensive review of hybrid intelligence forecasting (HIF), an emerging paradigm that integrates human judgment and machine intelligence to create a better sociotechnical ensemble and enhance forecasting performance. HIF seeks to combine the complementary strengths of humans and machines through the principles of collectivity, superior results, and continuous learning in forecasting. This review evaluates the literature through these three pillars, identifying how forecasts are combined, the impact of trust and forecast correlation, conditions under which hybrid systems outperform individual components, and the challenges of fostering ongoing learning between human and machine agents to devise better sociotechnical ensembles. Whereas hybrid systems often yield improved predictive accuracy across diverse contexts and methodologies, significant gaps remain in trust calibration, correlation management, forecast assessment, and iterative colearning. The paper concludes by outlining research directions needed to advance HIF toward a robust, adaptive, and trustworthy forecasting paradigm, as well as directions for standardizing HIF studies to better inform public and private administrations about its value.
