Generative artificial intelligence (GAI), exemplified by models such as ChatGPT, is driving disruptive technological transformations. Its rapid and widespread adoption presents dual challenges: job displacement and skill renewal, placing unprecedented pressure on the public's sense of occupational security. To clarify the mechanisms through which GAI adoption affects occupational security, this study analyzes public commentary data from major social media platforms. Sentiment analysis and topic modeling are employed to identify key influencing factors, map their causal relationships, and construct a hierarchical structure. The aim is to offer targeted mitigation strategies to address the occupational security challenges arising from GAI adoption.

Public comments related to the occupational impact of GAI were primarily collected from TikTok, with additional data obtained from Weibo and bilibili, all of which are widely used social media platforms in China. After data cleaning and manual filtering, a bidirectional encoder representations from Transformers—based sentiment classification model was employed to extract comments expressing negative sentiment, resulting in a perception-based corpus focused on occupational insecurity triggered by GAI adoption. The biterm topic model was then used for topic modeling, identifying eight core themes—including employment and societal dynamics and human-AI collaboration. Semantic analysis of topic keywords distilled seven critical influencing factors. The decision-making trial and evaluation laboratory (DEMATEL) method was applied to construct an influence matrix quantifying the strength and direction of causal relationships among these factors. Finally, interpretive structural modeling (ISM) was used to build a hierarchical structure of the influencing factors, revealing their stratified distribution and transmission pathways.

The impact of GAI adoption on public occupational security was found to result from interconnected, multi-level structural factors. Based on DEMATEL analysis, the seven influencing factors were categorized into four functional zones: core driving, auxiliary support, adaptive adjustment, and comprehensive transmission. Human-AI collaboration was placed in the core driving zone, exerting a strong influence on other factors. Employment market changes, situated in the comprehensive transmission zone, showed the highest centrality. This factor was significantly influenced by other variables, indicating its pivotal role in the overall impact mechanism. ISM further revealed a three-level hierarchical structure. The foundational level included human-AI collaboration and legal/ethical and privacy safeguards, which initiated the occupational security shock. The middle level comprised factors such as future development prospects, industrial restructuring, and infrastructure development, which reflected structural adjustments driven by technological change and functioned as transitional nodes. The top level encompassed digital literacy education and employment market changes, representing the most direct pathways through which GAI impacted the public's occupational security.

Based on these findings, this study proposes a multi-level response framework from the perspectives of individuals, enterprises, social organizations, and government actors to address the occupational security challenges posed by GAI adoption. Furthermore, the analytical framework developed herein provides a theoretical foundation and practical reference for future research on occupational risk assessment and governance strategies in the era of rapid GAI advancement, thereby supporting the coordinated development of technological progress and societal stability.

Developing a strong safety culture is a critical strategy for coal enterprises to address persistent safety challenges. However, existing approaches to safety culture construction face two major limitations: ambiguity in the conceptual definition of "safety culture" and a lack of clarity regarding its interaction mechanisms with other organizational safety elements. These limitations hinder the full realization of the safety culture's preventive role in accident reduction. To overcome these issues, this study proposes a safety culture construction method based on the 24Model, followed by an empirical investigation using data collected from coal enterprises.

First, the theoretical framework of the 24Model was applied to clarify the connotation of safety culture and elucidate its interaction mechanisms with other organizational safety elements. Second, a safety culture analysis program was employed to collect and verify 3482 valid questionnaires from 19 coal enterprises in China. Third, the influence of various factors on safety culture levels was systematically examined across three dimensions: individual characteristics (gender, years of service, and education level); industry-wide safety level; and four functional position groups (managers, professionals, team leaders, and frontline workers). Based on the empirical findings, targeted strategies for improving safety culture levels were proposed.

No statistically significant difference in safety culture scores was found between male and female employees. Employees with more than 10 years of service scored significantly higher than those with ≤2 years or 2-5 years of service, and those with higher education levels scored significantly higher than those with lower education levels. The overall safety culture score of the sampled enterprises was 78.19, which is slightly higher than the national average for other industries (77.74). A comparative analysis of the evaluation results of 32 safety culture elements across four functional position groups, benchmarked against enterprises with superior and poor safety performance, revealed the following: elements such as "role of the safety department" "approach to safety performance" and "role of safety organization" consistently scored low across three groups; elements including "community safety impact" "care for injured employees" "understanding of safety performance" "relationship between safety performance and human resources" and "overall safety expectation" showed deficiencies in two groups; and elements such as "importance of safety" "perception of safety investment" "formation of safety values" "leadership responsibility" "employee participation" "training needs" "quality of safety meetings" "implementation of safety procedures" "types of safety inspections" "safety management of subsidiaries and contractors" and "emergency response capability" fell below expectations in one group.

The empirical results demonstrate that both years of service and education level are significantly and positively correlated with safety culture levels in coal enterprises. Furthermore, systematic differences were observed among the four functional position groups in their understanding of specific safety concepts. Based on these insights, a three-stage safety culture enhancement strategy and a dual-path improvement framework integrating reconstruction of the safety management system and enhancement of employee safety competence were proposed. This framework provides a systematic and practical approach for strengthening safety culture in coal enterprises.