Positive relationship between education level and risk perception and behavioral response: A machine learning approach

PLOS ONE | https://doi.org/10.1371/journal.pone.0321153 April 3, 2025 3 / 10 PLOS ONE Positive relationship between education level and risk perception and behavioral response chemical industry parks in China. Their findings revealed that individuals’ perceived risks played a pivotal role in determining their willingness to engage in protest or resistance activ- ities. Notably, higher levels of social trust and public engagement were associated with lower perceived risks, which subsequently reduced the likelihood of active resistance. This highlights the importance of not only comprehending risk perception but also considering how social and community factors shape behavioral responses to perceived risks. In recent years, machine learning techniques have become increasingly prevalent in the analy- sis of risk perception and disaster response behaviors. These methods provide novel opportunities to uncover complex patterns and relationships within large datasets, offering insights that tradi- tional statistical approaches may fail to detect. ML algorithms, including decision trees and sup- port vector machines (SVM), have been employed to predict individuals’ behaviors in response to disaster risks and to identify key factors such as education, age, and socioeconomic status that influence risk perception and preparedness actions. For instance, in the domain of pandemic risk perception, ML methods have been applied to analyze public attitudes and behaviors in relation to COVID-19. Vieira et al. [7] developed a Pandemic Risk Perception Scale and utilized ML

techniques to model various risk dimensions, including infection risk, emotional health risk, and health system risk, as well as their influence on public preparedness behaviors. These approaches facilitate a more nuanced understanding of how individuals from diverse backgrounds (e.g., edu- cational levels, socioeconomic status) perceive and respond to risks. ML can identify contextual variations in risk perception, allowing for the customization of disaster management strategies to specific groups. By integrating demographic data, past experiences, and risk communication, ML models can provide real-time predictions of public behavior, thereby enhancing the efficacy of risk communication and emergency preparedness efforts. Despite significant progress in under- standing the relationship between education level, risk perception, and behavioral responses, several gaps persist. Most research has concentrated on specific disaster types, such as natural hazards or pandemics, with limited cross-disaster comparisons. Future studies should explore risk perception across diverse hazard types to identify both commonalities and distinct factors influencing risk awareness and behavior. While ML holds promise for analyzing large-scale data, its application in risk perception research remains nascent. Further investigation is needed to integrate ML with theoretical frameworks, such as the TPB, to develop comprehensive models that address the complex interactions between education, risk perception, and disaster preparedness. Additionally,

more context-specific studies are required to examine how cultural and social factors, along- side education, influence risk perception and response behaviors. This study aims to address these gaps by examining the impact of education level on risk per- ception and behavioral response using ML techniques. By constructing predictive models based on 27 factors, we identify key predictors of behavioral intention and preparatory behavior, with a particular emphasis on education level. Additionally, SHAP values are employed to interpret model outputs and reveal the mechanisms through which education influences these behaviors. Our findings contribute to the growing literature on the application of ML in risk research and provide a scientific foundation for the development of targeted risk management policies. By integrating recent advancements in ML and risk research, this study provides a novel perspective on the role of education in shaping risk-related behaviors and highlights the importance of context-specific evidence in guiding policy and practice. 2. Methodology ML techniques are widely used in environmental risk assessments [8,9], such as flood suscep- tibility mapping [10], and in community and behavioral risk evaluations, including identifying