Search for: All records

ABSTRACT Estimating parameters and their credible intervals for complex system dynamics models is challenging but critical to continuous model improvement and reliable communication with an increasing fraction of audiences. The purpose of this study is to integrate Amortized Bayesian Inference (ABI) methods with system dynamics. Utilizing Neural Posterior Estimation (NPE), we train neural networks using synthetic data (pairs of ground truth parameters and outcome time series) to estimate parameters of system dynamics models. We apply this method to two example models: a simple Random Walk model and a moderately complex SEIRb model. We show that the trained neural networks can output the posterior for parameters instantly given new unseen time series data. Our analysis highlights the potential of ABI to facilitate a principled, scalable, and likelihood‐free inference workflow that enhance the integration of models of complex systems with data. Accompanying code streamlines application to diverse system dynamics models. 

Abstract We discuss the emerging new opportunity for building feedback‐rich computational models of social systems using generative artificial intelligence. Referred to as generative agent‐based models (GABMs), such individual‐level models utilize large language models to represent human decision‐making in social settings. We provide a GABM case in which human behavior can be incorporated into simulation models by coupling a mechanistic model of human interactions with a pre‐trained large language model. This is achieved by introducing a simple GABM of social norm diffusion in an organization. For educational purposes, the model is intentionally kept simple. We examine a wide range of scenarios and the sensitivity of the results to several changes in the prompt. We hope the article and the model serve as a guide for building useful dynamic models of various social systems that include realistic human reasoning and decision‐making. © 2024 System Dynamics Society. 

The transmission dynamics of infectious diseases and human responses are intertwined, forming complex feedback loops. However, many epidemic models fail to endogenously represent human behavior change. In this study, we introduce a novel behavioral epidemic model that incorporates various behavioral phenomena into SEIR models, including risk-response dynamics, shifts in containment policies, adherence fatigue, and societal learning, alongside disease transmission dynamics. By testing our model against data from 8 countries, where extensive behavioral data were available, we simultaneously replicate death rates, mobility trends, fatigue levels, and policy changes, both in-sample and out-of-sample. Our model offers a comprehensive depiction of changes in multiple behavioral measures along with the spread of the disease. We assess the explanatory power of each model mechanism in capturing data variability. Our findings demonstrate that the comprehensive model that includes all mechanisms provides the most insightful perspective for understanding the influence of human behavior during pandemics. 

The recent pandemic emphasized the need to consider the role of human behavior in shaping epidemic dynamics. In particular, it is necessary to extend beyond the classical epidemiological structures to fully capture the interplay between the spread of disease and how people respond. Here, we focus on the challenge of incorporating change in human behavior in the form of “risk response” into compartmental epidemiological models, where humans adapt their actions in response to their perceived risk of becoming infected. The review examines 37 papers containing over 40 compartmental models, categorizing them into two fundamentally distinct classes: exogenous and endogenous approaches to modeling risk response. While in exogenous approaches, human behavior is often included using different fixed parameter values for certain time periods, endogenous approaches seek for a mechanism internal to the model to explain changes in human behavior as a function of the state of disease. We further discuss two different formulations within endogenous models as implicit versus explicit representation of information diffusion. This analysis provides insights for modelers in selecting an appropriate framework for epidemic modeling. 

Human behavior shapes epidemic trajectories, evolving as individuals reassess risks over time. Our study closes the loop between epidemic status, individual risk assessments, and interactions. We developed an agent-based model where the individuals can alter their decisions based on perceived risks. In our model, agents’ perceived risk is proxied by their full awareness of actual risks, such as the probability of infection or death. We conducted several simulations of COVID-19 spread for a large metropolitan city akin to New York City, covering the period from December 2020 to May 2021. Our model allows residents to decide daily on traveling to crowded city areas or stay in neighborhoods with relatively lower population density. Our base run simulations indicate that when individuals assess their own risk and understand how diseases spread, they adopt behaviors that slow the spread of virus, leading to fewer cumulative cases and deaths but extending the duration of the outbreak. This model was then simulated with various vaccination strategies such as random distribution, prioritizing older individuals, high-contact-rate individuals, or crowded area residents, all within a risk-response behavioral framework. Results show that, in the presence of agents’ behavioral response, there is only a marginal difference across different vaccination strategies. Specifically, vaccination in crowded areas slightly outperformed other vaccination strategies in reducing infections and prioritizing the elderly was slightly more effective in decreasing deaths. The lack of a universally superior vaccination strategy comes from the fact that lowering a risk leads to more risky behavior which partly compensates for vaccination effects. The comparable outcomes of random versus targeted vaccinations highlight the importance of equitable distribution as another key focus in pandemic responses. 

COVID-19 highlighted the importance of considering human behavior change when modeling disease dynamics. This led to developing various models that incorporate human behavior. Our objective is to contribute to an in-depth, mathematical examination of such models. Here, we consider a simple deterministic compartmental model with endogenous incorporation of human behavior (i.e., behavioral feedback) through transmission in a classic Susceptible–Exposed–Infectious–Recovered (SEIR) structure. Despite its simplicity, the SEIR structure with behavior (SEIRb) was shown to perform well in forecasting, especially compared to more complicated models. We contrast this model with an SEIR model that excludes endogenous incorporation of behavior. Both models assume permanent immunity to COVID-19, so we also consider a modification of the models which include waning immunity (SEIRS and SEIRSb). We perform equilibria, sensitivity, and identifiability analyses on all models and examine the fidelity of the models to replicate COVID-19 data across the United States. Endogenous incorporation of behavior significantly improves a model’s ability to produce realistic outbreaks. While the two endogenous models are similar with respect to identifiability and sensitivity, the SEIRSb model, with the more accurate assumption of the waning immunity, strengthens the initial SEIRb model by allowing for the existence of an endemic equilibrium, a realistic feature of COVID-19 dynamics. When fitting the model to data, we further consider the addition of simple seasonality affecting disease transmission to highlight the explanatory power of the models. 

Behavioral epidemic models incorporating endogenous societal risk-response, where changes in risk perceptions prompt adjustments in contact rates, are crucial for predicting pandemic trajectories. Accurate parameter estimation in these models is vital for validation and precise projections. However, few studies have examined the problem of identifiability in models where disease and behavior parameters must be jointly estimated. To address this gap, we conduct simulation experiments to assess the effect on parameter estimation accuracy of a) delayed risk response, b) neglecting behavioral response in model structure, and c) integrating disease and public behavior data. Our findings reveal systematic biases in estimating behavior parameters even with comprehensive and accurate disease data and a well-structured simulation model when data are limited to the first wave. This is due to the significant delay between evolving risks and societal reactions, corresponding to the duration of a pandemic wave. Moreover, we demonstrate that conventional SEIR models, which disregard behavioral changes, may fit well in the early stages of a pandemic but exhibit significant errors after the initial peak. Furthermore, early on, relatively small data samples of public behavior, such as mobility, can significantly improve estimation accuracy. However, the marginal benefits decline as the pandemic progresses. These results highlight the challenges associated with the joint estimation of disease and behavior parameters in a behavioral epidemic model. 

In the first two years of the COVID-19 pandemic, per capita mortality varied by more than a hundredfold across countries, despite most implementing similar nonpharmaceutical interventions. Factors such as policy stringency, gross domestic product, and age distribution explain only a small fraction of mortality variation. To address this puzzle, we built on a previously validated pandemic model in which perceived risk altered societal responses affecting SARS-CoV-2 transmission. Using data from more than 100 countries, we found that a key factor explaining heterogeneous death rates was not the policy responses themselves but rather variation in responsiveness. Responsiveness measures how sensitive communities are to evolving mortality risks and how readily they adopt nonpharmaceutical interventions in response, to curb transmission. We further found that responsiveness correlated with two cultural constructs across countries: uncertainty avoidance and power distance. Our findings show that more responsive adoption of similar policies saves many lives, with important implications for the design and implementation of responses to future outbreaks.