Risk Analysis最新文献

Amidst mass immunization efforts to curb COVID-19 transmission, policy mandates enforced minimum physical distancing. Concerns arose regarding risk compensation, where individuals might reduce adherence to distancing if benefiting from multiple risk-reducing interventions. This study used an online natural experiment to examine the association of vaccination status and vaccine efficacy beliefs with social distancing preferences. Participants completed a distance-matching task, positioning avatars in stylized scenarios drawn from a 2 (location) $\times$ 3 (activity) factorial design. Data were collected in July 2021 during the vaccine rollout program in the United Kingdom. Contrary to risk compensation expectations, this study found no strong evidence of reduced distancing at a population level. However, stronger vaccine efficacy beliefs were associated with slightly reduced distancing among fully vaccinated individuals-a small effect size. In contrast, partially vaccinated and unvaccinated individuals with stronger vaccine beliefs maintained greater distance, suggesting a nuanced relationship between perceptions of vaccine efficacy and distancing behavior. Subjective risk perceptions did not significantly alter these patterns. Additionally, partially vaccinated individuals behaved similarly to the unvaccinated despite expressing higher perceived infection risk, and unvaccinated participants who intended to vaccinate showed lower distancing preferences. The study also identified an in-group bias in perceptions of vaccine distribution. While these findings were collected during a specific phase of the COVID-19 pandemic-when vaccination uptake and policy measures were rapidly changing-they underscore the importance of investigating how vaccine beliefs shape protective behaviors. Given the modest effect sizes observed, further research is warranted to clarify the evolving role of vaccine perceptions in public health strategies.

This study explores the spatiotemporal characteristics of social media users' emotional responses to natural disasters, providing valuable insights for government agencies to guide public sentiment, enhance emergency responses, and facilitate post-disaster reconstruction. It uses the 6.2 magnitude earthquake in Jishishan, Gansu, as a case study, collecting Weibo postings, and applying Snow NLP for sentiment analysis, and using DUTIR method for sentiment classification. This study examines the dynamics of public emotional expression over time and their spatial distribution during the disaster. Key findings indicate that the volume of social media posts about the Jishishan earthquake has shown a fluctuating downward trend, predominantly characterized by positive emotional expressions. The posting volume and the nature of emotional expression are influenced by various factors, including economic and social conditions, the progress of rescue efforts, the frequency of disasters, the extent of impact experienced by those affected, personal experiences of the disaster, and collective memory of the disaster, all exhibiting temporal and regional variations. The spatial distribution of these emotional expressions showed a negative correlation with the severity of the disaster impact, although this pattern was not evident in the epicenter region. Areas with memories of past disasters exhibited a higher prevalence of "sadness", regions more severely affected by the disaster displayed a greater proportion of "disgust", and the epicenter region had a higher volume of posts expressing "fear". As a case study, this research provides insights for decision-makers and the government to better understand public sentiments during disasters.

Climate change, along with its associated extreme and abnormal temperature events, poses risks to human health. We examine the impact of temperatures on health insurance decisions using a proprietary dataset that links critical illness insurance records with long-term, daily meteorological data. We show that both heat and cold increase health insurance purchases. The effect of heat is driven by both heightened physical health risks and the salience of unexpected, abnormal heatwaves in health insurance decisions. However, the heat impact decays with prior experience to abnormal heat events. In contrast, we find no evidence that cold temperatures either increase physical health risks or trigger a salience effect. Risk preference changes and business cycles do not explain our findings. Air conditioning mitigates heat-induced insurance demand and centralized heating system mitigates cold-induced insurance demand. Males, the elderly, and outdoor workers are more sensitive to heat compared to females, the young, and indoor workers. This research uniquely quantifies the effects of abnormal temperatures on insurance purchases, highlighting the salience effect in insurance decision-making.

Cross-organizational governance for extreme disaster risk represents a critical challenge for modern society. This study develops an integrated theoretical framework examining how emerging technologies transform collaborative governance for extreme disaster risks through complex adaptive mechanisms. Employing an innovative methodological triangulation approach combining qualitative comparative analysis (QCA), machine learning (XGBoost with SHAP), and agent-based modeling-systems dynamics (ABM-SD), we analyze disaster cases in China to identify and validate key technology-organization configurations that enhance system resilience. Initially, QCA analysis of 12 representative cases reveals that data analysis precision and inter-organizational links are necessary foundations for high-performance collaborative governance, with three distinct configuration pathways identified: non-pressure-responsive type, pressure-state type, and pressure-responsive type. Machine learning validation across an expanded sample of 120 cases confirms the robustness of these configurations while revealing their temporal evolution from network-dominated to data-driven patterns. The ABM-SD simulation demonstrates that proactive policies with cyclical technological upgrading significantly enhance system resilience, while loosely coupled networks with high heterogeneity better prevent "complexity traps" during extreme events. This research makes unique contributions by (1) establishing a systematic framework for analyzing technology-organization interactions in disaster contexts; (2) identifying equifinal pathways to effective collaborative governance; and (3) developing a theoretical model that illustrates how technological empowerment and organizational collaboration dynamically interact across threshold conversion areas to generate system emergence and reconstruction under varying pressure levels. Practical implications include configuration selection strategies for policy-makers based on regional development levels and disaster characteristics. Study limitations include the focus on Chinese cases, which may limit generalizability to different institutional contexts, and the need for longitudinal studies to further validate the proposed adaptation mechanisms.

Risk factors at different stages of COVID-19 may interact with each other, forming a risk network. Identifying the key risk factors within this network and their interrelationships is crucial for reducing the overall risk of COVID-19. We constructed three Bayesian Belief Network (BBN) models by combining data-driven approaches with expert validation. Using the Tree-Augmented Naive Bayes (TAN) algorithm, we developed the INFORM COVID-19 Risk BBN model and the COVID-19 Regional Safety Assessment BBN model. The joint BBN model was established using the Greedy Thick Thinning (GTT) algorithm. Parameter learning was performed through maximum likelihood estimation. Expert validation, 10-fold cross-validation, and model performance metrics were employed to comprehensively assess the overall performance of the models. Additionally, mutual information analysis and sensitivity analysis were used to explore the importance of risk factors at each stage and their interdependencies. "INFORM Vulnerability" and "INFORM Lack of Coping Capacity" were identified as the two key risk factors influencing the risk of early outbreak. In the mid-to-late stages of the pandemic, "Emergency Preparedness" and "Monitoring and Detection" had the greatest impact on regional safety and control measures. Furthermore, the joint BBN model indicated that the most important risk factors affecting the overall COVID-19 risk were "Lack of Coping Capacity," "Government Risk Management Efficiency," and "Regional Resiliency," while the influence of other variables was relatively minor. The main contribution of this study lies in identifying the key risk factors at different stages of the pandemic and their interdependencies, providing policymakers with valuable insights for the rational allocation of limited health resources and the formulation of appropriate and effective prevention and control policies.

Recent government and commercial initiatives have pushed for the use of the automated, artificial intelligence (AI)-based, analysis of cyber threat intelligence. The potential bias that might be present when evaluating threat intelligence coming from human and AI sources has to be better understood before deploying automated solutions to production. We present a controlled experiment with $n=57$ master students who had a mix of experience in security and machine learning to measure the bias introduced by the source of intelligence (human vs. AI). Each participant analyzed eight threat intelligence reports from the Dutch National Cyber Security Center where the source of the final recommendation was manipulated as for coming from a human expert or an AI algorithm. Our findings revealed that participants tended to disagree with the recommendation when it was coming from AI. While expertise on ML did not have any impact, we found that participants with more security expertise tended to agree with the recommendation. In contrast, we found that the perceives bias was statistically equivalent (TOST) whether the recommendation was coming from a human or from an AI. The only (expected) factor which had an impact on perceived bias was when participants disagreed with the recommendation (irrespective whether it was human or AI). These results provide insight on the possible impact of introduction on AI on rank-and-file Tier 1 SOC analysts. The generalization of our results to professional practice requires more experiments with experienced security professionals.

This study examines how question order influences responses in multidimensional risk perception measurement. Through a randomized between-subjects experiment (N = 1352) manipulating the sequence of risk perception dimensions, we identified systematic question order effects. When a general risk question followed specific dimensional questions, responses showed significant assimilation effects (i.e., general risk aligned more closely with preceding specific dimension ratings). Consequence dimension responses (severity, affect) showed assimilation effects when preceded by probability dimensions (exposure, susceptibility), while probability dimensions remained stable regardless of ordering. Within subdimensions, severity ratings were influenced by preceding affect questions, and susceptibility ratings were influenced by preceding exposure questions, both displaying assimilation patterns. Testing how individual differences in cognitive sophistication moderate susceptibility to order effects, contrary to our predictions, we found that individuals higher in analytical thinking style demonstrated stronger order effects for general risk questions than those lower in analytical thinking. These findings reveal an asymmetrical pattern where judgments requiring more analytic specificity tend to anchor evaluations that are relatively global, affective, or self-focused.

There is a lack of rigorous studies addressing the theory life cycle model in disaster management. Thus, this study aimed to review the theory life cycle to improve disaster management practices. The study employed a systematic literature review, guided by the Preferred Reporting Items for Systematic Reviews and Meta-Analyses. A reductionist model was proposed, including (1) theory inception, (2) theory scrutiny, and (3) theory termination (X) or establishment (O). This model was applied to four theories: suicide rate (X1), risk perception (X2), redundancy (O1), and all hazards (O2). In pursuing the reductionist model, the field must consider disaster characteristics, the advantages and disadvantages of various theories, the changing environment, a hybridization perspective, emergency education and training, and continuous improvement. This study emphasizes the question of adaptive relevance more than previous research.

The deployment of radiological dispersal devices (RDDs) via drones presents a novel security challenge, necessitating advanced tools for consequence assessment and response planning. We developed an integrated framework combining physics-based dispersion modeling, constrained optimization, and machine learning to evaluate such threats. Using a Monte Carlo approach, 2000 synthetic scenarios were generated incorporating five radionuclides (Cs-137, I-131, Co-60, Sr-90, and Am-241), meteorological variability, and geospatial risk zones. A constrained optimization routine based on the Limited-memory Broyden-Fletcher-Goldfarb-Shanno algorithm with bound constraints (L-BFGS-B) identified adversarial scenarios that maximize contaminated area (>10 km²) while minimizing energy use and detection risk, revealing nonlinear trade-offs between dispersal effectiveness and operational stealth. Consequence modeling with Health Physics Code (HotSpot) and Java-based Real-time Online Decision Support system (JRODOS) showed systematic differences, with HotSpot predicting higher total effective dose (TED) and time-integrated air concentration (TIAC). I-131 posed the greatest acute thyroid risk, whereas Am-241 dominated long-term exposure. Protective action analysis demonstrated that reinforced sheltering reduces cumulative dose by up to two orders of magnitude compared to outdoor exposure. Finally, the machine learning framework achieved accurate and rapid predictions (R² = 0.975), with distance as the dominant predictor. These findings provide actionable guidance for emergency preparedness against drone-based RDD threats.