Search

The COVID-19 pandemic highlighted the critical role of human behavior in influencing infectious disease transmission and the need for models capturing this complex dynamic. We present an agent-based model integrating an epidemiological simulation of disease spread with a cognitive architecture driving individual mask-wearing decisions. Agents decide whether to mask based on a utility function weighting factors like peer conformity, personal risk tolerance, and mask-wearing discomfort. By conducting experiments systematically varying behavioral model parameters and social network structures, we demonstrate how adaptive decision-making interacts with network connectivity patterns to impact population-level infection outcomes. The model provides a flexible computational framework for gaining insights into how behavioral interventions like mask mandates may differentially influence disease spread across communities with diverse social structures. Findings highlight the importance of integrating realistic human decision processes in epidemiological models to inform policy decisions during public health crises., Comment: 12 pages, 6 figures

We use mobile device data to construct empirical interpersonal physical contact networks in the city of Portland, Oregon, both before and after social distancing measures were enacted during the COVID-19 pandemic. These networks reveal how social distancing measures and the public's reaction to the incipient pandemic affected the connectivity patterns within the city. We find that as the pandemic developed there was a substantial decrease in the number of individuals with many contacts. We further study the impact of these different network topologies on the spread of COVID-19 by simulating an SEIR epidemic model over these networks, and find that the reduced connectivity greatly suppressed the epidemic. We then investigate how the epidemic responds when part of the population is vaccinated, and we compare two vaccination distribution strategies, both with and without social distancing. Our main result is that the heavy-tailed degree distribution of the contact networks causes a targeted vaccination strategy that prioritizes high-contact individuals to reduce the number of cases far more effectively than a strategy that vaccinates individuals at random. Combining both targeted vaccination and social distancing leads to the greatest reduction in cases, and we also find that the marginal benefit of a targeted strategy as compared to a random strategy exceeds the marginal benefit of social distancing for reducing the number of cases. These results have important implications for ongoing vaccine distribution efforts worldwide., Comment: 15 pages, 9 figures. This article is a RAND Working Report

Network data is increasingly being used in quantitative, data-driven public policy research. These are typically very rich datasets that contain complex correlations and inter-dependencies. This richness both promises to be quite useful for policy research, while at the same time posing a challenge for the useful extraction of information from these datasets - a challenge which calls for new data analysis methods. In this report, we formulate a research agenda of key methodological problems whose solutions would enable new advances across many areas of policy research. We then review recent advances in applying deep learning to network data, and show how these methods may be used to address many of the methodological problems we identified. We particularly emphasize deep generative methods, which can be used to generate realistic synthetic networks useful for microsimulation and agent-based models capable of informing key public policy questions. We extend these recent advances by developing a new generative framework which applies to large social contact networks commonly used in epidemiological modeling. For context, we also compare and contrast these recent neural network-based approaches with the more traditional Exponential Random Graph Models. Lastly, we discuss some open problems where more progress is needed., Comment: 77 pages, 38 figures. This article is a RAND Working Report. v2: corrected typo in metadata

Agent-based Models (ABMs) are valuable tools for policy analysis. ABMs help analysts explore the emergent consequences of policy interventions in multi-agent decision-making settings. But the validity of inferences drawn from ABM explorations depends on the quality of the ABM agents' behavioral models. Standard specifications of agent behavioral models rely either on heuristic decision-making rules or on regressions trained on past data. Both prior specification modes have limitations. This paper examines the value of reinforcement learning (RL) models as adaptive, high-performing, and behaviorally-valid models of agent decision-making in ABMs. We test the hypothesis that RL agents are effective as utility-maximizing agents in policy ABMs. We also address the problem of adapting RL algorithms to handle multi-agency in games by adapting and extending methods from recent literature. We evaluate the performance of such RL-based ABM agents via experiments on two policy-relevant ABMs: a minority game ABM, and an ABM of Influenza Transmission. We run some analytic experiments on our AI-equipped ABMs e.g. explorations of the effects of behavioral heterogeneity in a population and the emergence of synchronization in a population. The experiments show that RL behavioral models are effective at producing reward-seeking or reward-maximizing behaviors in ABM agents. Furthermore, RL behavioral models can learn to outperform the default adaptive behavioral models in the two ABMs examined., Comment: This is a more detailed version of a paper ("Modeling Agent Behaviors for Policy Analysis via Reinforcement Learning") accepted to appear in IEEE ICMLA 2020. This also corrects an error in Fig. 7 of the original arXiv submission. Fig. 7 now specifies the right ABM architecture ("flu" instead of "tax")

Inspired by Minority Games, we constructed a novel individual-level game of adaptive decision-making based on the dilemma of deciding whether to participate in voluntary influenza vaccination programs. The proportion of the population vaccinated (i.e., the vaccination coverage) determines epidemic severity. Above a critical vaccination coverage, epidemics are prevented; hence individuals find it unnecessary to vaccinate. The adaptive dynamics of the decisions directly affect influenza epidemiology and, conversely, influenza epidemiology strongly influences decision-making. This feedback mechanism creates a unique self-organized state where epidemics are prevented. This state is attracting, but unstable; thus epidemics are rarely prevented. This result implies that vaccination will have to be mandatory if the public health objective is to prevent influenza epidemics. We investigated how collective behavior changes when public health programs are implemented. Surprisingly, programs requiring advance payment for several years of vaccination prevents severe epidemics, even with voluntary vaccination. Prevention is determined by the individuals' adaptability, memory, and number of pre-paid vaccinations. Notably, vaccinating families exacerbates and increases the frequency of severe epidemics.

We define and analyze an inductive reasoning game of voluntary yearly vaccination in order to establish whether or not a population of individuals acting in their own self-interest would be able to prevent influenza epidemics. We find that epidemics are rarely prevented. We also find that severe epidemics may occur without the introduction of pandemic strains. We further address the situation where market incentives are introduced to help ameliorating epidemics. Surprisingly, we find that vaccinating families exacerbates epidemics. However, a public health program requesting prepayment of vaccinations may significantly ameliorate influenza epidemics., Comment: 20 pages, 7 figures

Background: Approximately 40% of adults in Botswana are HIV-infected. The Botswana antiretroviral program began in 2002 and currently treats 34,000 patients with a goal of treating 85,000 patients (~30% of HIV-infected adults) by 2009. We predict the evolution of drug-resistant strains of HIV that may emerge as a consequence of this treatment program. We discuss the implications of our results for the World Health Organization's (WHO's) proposed surveillance system for detecting drug-resistant strains of HIV in Africa. Methods: We use a mathematical model of the emergence of drug resistance. We incorporate demographic and treatment data to make specific predictions as to when the WHO surveillance threshold is likely to be exceeded. Results: Our results show - even if rates of acquired resistance are high, but the drug-resistant strains that evolve are only half as transmissible as wild-type strains - that transmission of drug-resistant strains will remain low (< 5% by 2009) and are unlikely to exceed the WHO's surveillance threshold. However,our results show that transmission of drug-resistant strains in Botswana could increase to ~15% by 2009 if resistant strains are as transmissible as wild-type strains. Conclusion: The WHO's surveillance system is designed to detect transmitted resistance that exceeds a threshold level of 5%. Whether this system will detect drug-resistant strains in Botswana by 2009 will depend upon the transmissibility of the strains that emerge. Our results imply that it could be many years before the WHO detects transmitted resistance in other sub-Saharan African countries with less ambitious treatment programs than Botswana.