Risk Analysis最新文献

Recent artificial intelligence (AI) risk management frameworks and regulations place stringent quality constraints on AI systems to be deployed in an increasingly competitive environment. Thus, from a software engineering point of view, a major issue is deciding when to release an AI system to the market. This problem is complex due to, among other features, the uncertainty surrounding the AI system's reliability and safety as reflected through its faults, the various cost items involved, and the presence of competitors. A novel general adversarial risk analysis framework with multiple agents of two types (producers and buyers) is proposed to support an AI system developer in deciding when to release a product. The implementation of the proposed framework is illustrated with an example and extensions to cases with multiple producers and multiple buyers are discussed.

We organized adversarial collaborations between subject-matter experts and expert forecasters with opposing views on whether recent advances in Artificial Intelligence (AI) pose an existential threat to humanity in the 21st century. Two studies incentivized participants to engage in respectful perspective-taking, to share their strongest arguments, and to propose early-warning indicator questions (cruxes) for the probability of an AI-related catastrophe by 2100. AI experts saw greater threats from AI than did expert forecasters, and neither group changed its long-term risk estimates, but they did preregister cruxes whose resolution by 2030 would sway their views on long-term risk. These persistent differences shrank as questioning moved across centuries, from 2100 to 2500 and beyond, by which time both groups put the risk of extreme negative outcomes from AI at 30%-40%. Future research should address the generalizability of these results beyond our sample to alternative samples of experts, and beyond the topic area of AI to other questions and time frames.

Fugitive dust from concentrated animal feeding operations (CAFOs) is a potential source of produce contamination with human pathogens. Our objective was to develop a general framework and methodology for predicting preharvest produce contamination with human pathogens arising from fugitive dust derived from a nearby CAFO. We applied this framework to a case study of lettuce grown in proximity to a bovine CAFO. We implemented the EPA's AERMOD dispersion model at two locations, assessing dust dispersion and deposition over a 30-day period across 100 km² surrounding a 10,000-animal CAFO. E. coli O157:H7 contaminated lettuce servings grown on fields within the 100 km² were predicted using a risk assessment approach, integrating data about dust deposition, pathogen contamination in cattle manure, and pathogen survival on crops. To contextualize the results, infectious servings were predicted based on the average number of E. coli O157:H7 per serving and the E. coli O157:H7 dose-response relationship. Dust from CAFOs has the potential to deposit across at least 100 km². E. coli O157:H7 dispersion and deposition are impacted by wind direction and velocity, emission factor, and prevalence and concentration in dust. Mean E. coli O157:H7 concentrations on preharvest lettuce were predicted across the 100 km² and declined considerably with distance from the CAFO. Surviving E. coli O157:H7 on preharvest lettuce arise primarily from dust deposited in the 2 weeks before harvest. Our modeling approach provides a flexible framework that can be adapted to any location, providing quantitative information to inform foodborne outbreak investigations and the development of prevention strategies.

Recent literature has examined the role of misinformation, biases, and other factors in contributing to the integrity of a risk study. These types of social and cognitive dynamics-referred to as narratives-comprise concern and value in a risk study. These narratives may appear to undermine aspects of objectivity in a scientific sense, but they may also shed light on aspects of a risk study that involve perceived scientific truths, related risk concerns, and values. The narratives can inform overall risk perception and the perception of quality for the risk study. As a result, understanding and classifying those narratives provides additional evidence that can potentially inform decisions for the design and implementation of a risk study. In this article, we develop a classification system that can be used to understand and address narratives that can influence a risk study and how various stakeholders perceive the risk study. This article will be of interest to risk analysts, policymakers, and risk communicators.

In this work, we introduce JointLIME, a novel interpretation method for explaining black-box survival (BBS) models with endogenous time-varying covariates (TVCs). Existing interpretation methods, like SurvLIME, are limited to BBS models only with time-invariant covariates. To fill this gap, JointLIME leverages the Local Interpretable Model-agnostic Explanations (LIME) framework to apply the joint model to approximate the survival functions predicted by the BBS model in a local area around a new individual. To achieve this, JointLIME minimizes the distances between survival functions predicted by the black-box survival model and those derived from the joint model. The outputs of this minimization problem are the coefficient values of each covariate in the joint model, serving as explanations to quantify their impact on survival predictions. JointLIME uniquely incorporates endogenous TVCs using a spline-based model coupled with the Monte Carlo method for precise estimations within any specified prediction period. These estimations are then integrated to formulate the joint model in the optimization problem. We illustrate the explanation results of JointLIME using a US mortgage data set and compare them with those of SurvLIME.

The Chinese public is increasingly experiencing the local impacts of climate change, whereas the government downplays its domestic effects and critical opinions on environmental governance. As climate change perceptions are crucial for individual risk management, adaptation, and collective climate actions, it is vital to explore how these perceptions are shaped. Given the increasing significance of social media in climate change discourse, this study employs survey data from the 2021 Environmental Risk Perceptions and Environmental Behaviors of Urban Residents Project to investigate how social media exposure influences risk perceptions of climate change among the Chinese public. Drawing on the social amplification of risk framework, this study examines the effect of exposure to environmental information, exposure to opinion diversity, individuals' social media network ties to environmental opinion leaders, and the interaction between social media exposure and cultural values. The results indicate that in the contexts where climate change is neither politically divisive nor openly debated, social media exposure to diverse opinions and social media network ties to environmental scholars positively predict risk perceptions. Additionally, egalitarianism and fatalism are found to moderate the effect of these connections with environmental scholars. This study extends previous research, which focuses largely on the association between the frequency of social media exposure and risk perceptions of climate change, by revealing a more comprehensive and nuanced process that links social media exposure to climate change perceptions.

According to recent Household Pulse Survey data, roughly 1.1% of households were displaced due to disasters in the United States in recent years. Although most households returned relatively quickly, 20% were displaced for longer than 1 month, and 14% had not returned by the time of the survey. Protracted displacement creates enormous hardships for affected households and communities, yet few disaster risk analyses account for the time component of displacement. Here, we propose predictive models for household displacement duration and return for practical integration within disaster risk analyses, ranging in complexity and predictive power. Two classification tree models are proposed to predict return outcomes with a minimum number of predictors: one that considers only physical factors (TreeP) and another that also considers socioeconomic factors (TreeP&S). A random forest model is also proposed (ForestP&S), improving the model's predictive power and highlighting the drivers of displacement duration and return outcomes. The results of the ForestP&S model highlight the importance of both physical factors (e.g., property damage and unsanitary conditions) and socioeconomic factors (e.g., tenure status and income per household member) on displacement outcomes. These models can be integrated within disaster risk analyses, as illustrated through a hurricane scenario analysis for Atlantic City, NJ. By integrating displacement duration models within risk analyses, we can capture the human impact of disasters more holistically and evaluate mitigation strategies aimed at reducing displacement risk.

This study investigates financial contagion during natural disasters and explores the potential advantage of environmental, social, and governance (ESG) investing in such contagion. Specifically, we propose a new edge-weighted undirected contagion network to explore disaster-driven contagion and transmission channels across sectors, asset classes, and ESG international indexes. Our empirical results demonstrate the existence of the disaster-driven contagion. Natural disasters may increase investors' risk aversion, which further magnify portfolio rebalancing behavior, leading to the spread of financial contagion. Moreover, we also find that ESG investing helps mitigate the spread of disaster-driven contagion, thereby contributing to the resilience of the financial system during natural disasters.

With the increasing global risk of floods, there is an urgent need for new adaptive emergency management (AEM) frameworks. This study aims to integrate machine learning, physical models (such as the Variable Infiltration Capacity model and InfoWorks-ICM), and social data to develop the IFloPhy (Integrated Machine Learning and River Physical Model) framework, which explores early flood risk warnings under AEM. The multidimensional integration design of IFloPhy overcomes the limitations of traditional single-warning systems, enhancing dynamic response capabilities and predictive accuracy. By integrating physical processes, IFloPhy can dynamically track the formation and development of floods, comprehensively considering natural and socio-economic factors, thereby achieving holistic and interactive flood risk assessments. The incorporation of real-time satellite data with multi-model forecast results establishes an immediate warning mechanism, significantly reducing prediction uncertainty. IFloPhy has been deployed and validated in the San Isabel Basin in South America, demonstrating exceptional performance in areas with scarce data and limited communication infrastructure. IFloPhy offers new technologies and insights for risk management and AEM, proposing novel methods for flood risk emergency management.

Disasters caused by tailings storage facilities (TSFs) have highlighted the complexity of safely managing mine tailings and the extension of consequences over time and throughout the tailings runoff. Investigations commissioned by mining companies following major failures in Mariana and Brumadinho, Brazil, primarily focused on immediate technical causes and hazards. However, for effective disaster risk reduction, the integration of technical, environmental, and social factors is needed to comprehensively address the complexity of risk management. Bow-tie models can be used for TSF's disaster analysis, as they consider causes, consequences, and preventive and mitigation controls. Here, an adapted bow-tie framework for TSF's disaster risk analysis is proposed to systematize the identification of threats and consequences and address the four disaster risk dimensions: hazard, exposure, vulnerability, and capacity. The framework was applied to the Pontal TSF, Brazil, using publicly available information, revealing gaps in the risk management, such as the lack of identification of social vulnerabilities. Our framework highlights the importance of reducing TSF's disaster risks through all dimensions and engaging multiple stakeholders. Although TSF stability control is primordial and irreplaceable, alone it is insufficient for effective disaster risk reduction.