Exploring voluntary vaccination with bounded rationality through reinforcement learning

Evidence shows that there exists a complex interaction between human vaccinating behaviors and disease prevalence during an epidemic. Usually, rational individuals make vaccinating decisions by strategically evaluating the cost of vaccination and the risk of infection. While in reality, individuals’ decisions can also be influenced by their social acquaintances. In this paper, we present a reinforcement learning-based mechanism to characterize human decision-making process with bounded rationality, which takes into consideration both individuals’ rational decisions and social influence from their neighbors. Specifically, we investigate human voluntary vaccinating behaviors in the face of flu-like seasonal diseases in locally-mixed social networks, where each individual together with his/her neighbors forms a well-mixed environment. Through carrying out simulations, we evaluate the performance of decision-making mechanisms with/without reinforcement learning in terms of vaccine coverage, final epidemic size, average payoff and vaccine effectiveness under different settings of relative cost of vaccination and infection. Simulation results show that reinforcement learning can improve the vaccine effectiveness through balancing individuals’ rationality and social influence. This emphasizes the importance of appropriately utilizing human bounded rationality in preventing disease epidemics through voluntary vaccination policies.