Decoding Global AI Risk Perception Evolution: Intercultural drivers and Public Discourse Patterns in Algorithmic Societies

Decoding Global AI Risk Perception Evolution: Intercultural drivers and Public Discourse Patterns in Algorithmic Societies | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Decoding Global AI Risk Perception Evolution: Intercultural drivers and Public Discourse Patterns in Algorithmic Societies Zhichao Li, Mengxue Lyu, Lujia Li, Jilin Huang This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6468789/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract While artificial intelligence (AI) has become a transformative force across societies in the 21st century, its rapid advancement has sparked significant public discourse about potential risks and societal implications. This study investigates global patterns in public risk perceptions of artificial intelligence, examining three critical dimensions: spatial distribution characteristics, intercultural determinants, and temporal evolutionary trends. Through an integrated methodological framework combining natural language processing and convex hull analysis of 4.3 million social media posts from 2010-2023, we identify three dominant risk clusters: algorithmic bias in decision systems, data privacy violations , and labor market disruptions. The findings reveal significant geographic disparities, with privacy concerns increasing 23.7% annually in developed nations since 2018. Five key determinants emerge: national income levels, infrastructure completeness, health security index, regulatory quality, and electoral democracy index. This research offers valuable insights for evidence-based policymaking and contributes to the growing discourse on responsible AI development, ultimately supporting the sustainable integration of AI technologies into society. Social science/Cultural and media studies Social science/Science technology and society Artificial Intelligence Risk Perceptions Global Patterns Intercultural Determinants Algorithmic Bias Evolutionary Trends Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 8 Figure 9 Figure 10 1 Introduction The advent of Artificial Intelligence (AI) has precipitated a paradigm shift in 21st-century technological innovation while simultaneously amplifying intercultural ethical tensions in algorithmic governance (Hongladarom & Bandasak, 2024 ). As AI systems become increasingly integrated into critical decision-making processes and automation frameworks, their societal impact has grown exponentially. However, this rapid advancement has also generated significant public concerns regarding potential risks, particularly in areas such as algorithmic bias in decision-making systems, large-scale privacy violations, AI ethics (Zhang, B., & Dafoe, A., 2019 ) and the disruptive effects of automation on employment markets. Risk perception research in the field of Artificial Intelligence is crucial for developing effective risk communication strategies for informed decision making (Krieger, J. B. et al., 2024), as trust is a key factor in the acceptance and adoption of real-life technologies (Afroogh, S. et al., 2024 ). Further efforts are required to scrutinize the ethical, legal, and technical solutions proposed for AI governance through critical perspectives (Chang H., 2023 ). While AI's borderless nature demands international cooperation, our analysis of 4.3 million social media posts across 164 jurisdictions reveals fundamental disparities in risk perception patterns shaped by cultural cognition frameworks (Hofstede, 1980 ; O'Regan & Ferri, 2025). It is our aspiration that this research will lay the groundwork for tackling the pivotal issue of transitioning toward effective and principled AI governance. 1.1 Public Risk Perception of AI and Socio-cultural Influences The literature on public perception of AI risks has evolved across multiple domains, with significant contributions examining AI applications in healthcare (Bologheanu et al., 2023 ; Robleto, E. et al., 2024 ; Kerstan, S. et al., 2024 ; Chow, J. C. L. et al., 2024; Huo, W. et al.,2024), information communication (Zhang, W. et al., 2022 ), and agriculture (Kaiping Chen et al., 2023 ). Some researchers have focused on AI's impact on daily life through small-scale sampling (Neyazi, T. A. et al., 2023; Bedué, P., & Fritzsche, A., 2022 ; Rebekka S. et al., 2023), others have conducted national population-based studies (Gerlich M., 2023 ; Bao, L. et al., 2022 ; Liu, Y., & Lyu, Z., 2024 ). As AI adoption becomes increasingly widespread, understanding public risk perception has emerged as a crucial research area, particularly given its influence on technology acceptance and implementation (Yigitcanlar, T. et al., 2024 ). However, despite AI's growing ubiquity, research examining public risk perception remains relatively limited, with existing studies often prioritizing technical aspects over societal implications (Gerlich M., 2023 ). Building on established theoretical frameworks including the Technology Acceptance Model (Davis, 1989 ) and Risk Perception Heuristics (Slovic, 2000), recent studies demonstrate paradoxical findings. While 68% of Americans express concerns about AI-driven job displacement (Rainie, L. et al., 2022 ), only 23% of Chinese respondents prioritize employment risks (Bao et al., 2022 ). Existing research demonstrates that public trust plays a pivotal role in determining the successful adoption and diffusion of AI technologies. Trust in AI reduces perceived risk (Ajenaghughrure et al., 2021 ; Choudhury, A., 2022 ). The emergence of "trustworthy AI" as a fundamental principle in development and regulation underscores the importance of fairness, accountability, and transparency in AI systems (Liu, H. et al., 2022 ). When public trust erodes or fear of AI risks intensifies, it can impede technological innovation, decelerate adoption rates, and potentially trigger societal resistance, thereby forfeiting valuable economic and social benefits (Zuiderwijk, A. et al., 2021 ). Our research inquiry focuses on examining cross-cultural variations in public trust toward artificial intelligence systems, particularly investigating how these trust differentials manifest and evolve across distinct sociocultural value orientations. As the public's perception of AI risk is multifaceted and dynamic, monitoring these changes allows policymakers and technologists to anticipate potential concerns and to address them proactively (Zhang, B., & Dafoe, A., 2019 ). A significant body of research highlights the ethical challenges posed by AI systems. Studies have documented how AI-powered decision-making systems can perpetuate existing biases, leading to discriminatory outcomes across various domains including employment, criminal justice, and financial services (O'Reilly-Shah, V. N. et al., 2020). Additionally, concerns about privacy violations have intensified, particularly in regions with limited data protection frameworks (Ferrer, X. et al., 2021 ). These ethical considerations expose the tension between universalist AI governance models and culturally situated epistemologies. The observed 'ethics-washing' phenomenon in standardized guidelines (Hongladarom & Bandasak, 2024 ) suggests current frameworks inadequately address depth ontology differences in intercultural reasoning about algorithmic fairness. Public perceptions of AI risks emerge from complex interactions between technological characteristics, socioeconomic contexts, and culturally embedded value systems. Recent analyses of non-Western AI ethics frameworks reveal superficial adherence to universal principles masking deeper cultural ontologies that inform ethical reasoning (Hongladarom & Bandasak, 2024 ; O'Regan & Ferri, 2025). Public perceptions of AI risks are shaped by complex interactions between technological characteristics and broader socioeconomic contexts. Research has demonstrated significant variations in AI perception across different countries and regions, influenced by factors such as educational attainment, income levels, and political systems (Bengio, Y. et al., 2023 ). The relationship between personal AI experience and risk perception has been explored through the familiarity hypothesis, which suggests that increased exposure to AI technologies correlates with lower risk perception (Chang H., 2023 ). However, this relationship is moderated by broader societal factors and individual circumstances. Introverted traits and heightened general correlate with more positive attitudes towards AI (Schepman et al., 2022). Notable patterns emerge: societies with robust privacy regulations often exhibit heightened concerns about data misuse, while regions prioritizing social stability tend to focus more on employment-related risks from automation (McDuff, D. et al., 2018 ; Bao, L. et al., 2022 ). Enhanced security perceptions of private data in e-commerce reduces the perceived risk of financial AI services (Al-Gasawneh et al., 2022 ). Public perception is significantly shaped by media representation and cultural narratives surrounding AI. Research indicates that media portrayals—whether emphasizing AI's transformative potential or highlighting potential threats to personal freedom and security—can substantially influence public concerns (Wissing, B. G., & Reinhard, M. A., 2018; Sartori, L., & Bocca, G., 2023 ). These narratives contribute to broader societal uncertainty about AI's implications for social structures, employment patterns, and governance systems. 1.2 Research objectives In conventional studies that rely on regional questionnaire data, we frequently encounter fragmented insights into the public's perception of AI risks within a specific region or across regions (Bao, L. et al., 2022 ; Gerlich M., 2023 ; Neyazi, T. A. et al., 2023; Liu, Y., & Lyu, Z., 2024 ). However, such isolated information is inherently limited and fails to furnish policymakers with a comprehensive understanding of the broader trends in public perception (Zhang, B., & Dafoe, A., 2019 ). Even in existing studies that characterize public risk perception based on Twitter users' tweets, there is a significant geographical skew, with approximately 70% of the sample data set originating from the United States (Paté-Cornell, E., 2024). This overrepresentation poses a challenge, as it limits the breadth of global intercultural perspectives and makes it difficult to garner a comprehensive understanding of how the public in other countries perceives AI risks. The diverse social and economic factors that contribute to these varying risk perceptions are substantial, and they cannot be accurately assessed solely based on the content of tweets. A broader perspective is critical to fully understand the different and often conflicting views of AI (Krieger, J. B. et al., 2024). This would provide a more nuanced understanding of the factors that shape risk perception across different contexts. More critically, it falls short in elucidating the temporal dynamics and evolution of the cognitive structures that underpin these perceptions. To bridge this gap, there is a pressing need for a more holistic and longitudinal approach to analyzing public sentiment towards AI. This approach should aim to synthesize dispersed data points into a cohesive narrative that reveals not only the current state of public perception but also its trajectory over time. It enables policymakers to anticipate potential areas of concern, develop targeted educational initiatives to enlighten the public on the nuances of AI technologies, and establish regulatory frameworks that protect against the misuse of AI while promoting its beneficial applications. This study addresses three fundamental research questions: RQ1 : What are the global characteristics of public perception of AI risks in a intercultural context? RQ2 : What factors influence the public's perception of AI risks? RQ3 : How have AI risk perception patterns evolved over time? Our research makes several contributions to the growing body of literature on public perceptions of emerging technologies: We provide a comprehensive analysis of AI risk perceptions at a global scale, illuminating patterns across diverse geographic regions. Our findings offer valuable insights for policymakers and stakeholders in developing strategies that promote technological advancement while maintaining public trust and acceptance. The research promotes the understanding of factors for the sustainable development of AI technology in a intercultural context. Methodologically, we employed a sophisticated analytical framework combining multiple natural language processing tools (including Gensim and Perspective) with systematic manual review procedures. This approach was enhanced by implementing a multidimensional mathematical convex hull method for geospatial data analysis, enabling detailed examination of spatiotemporal distribution patterns, regional heterogeneity, and potential driving factors in global public perceptions of AI risks. Our analysis reveals that public concerns primarily concentrate on three key areas: algorithmic bias and unfairness in decision support systems, data privacy violations, and labor market disruptions due to automation. Notably, privacy-related concerns have exhibited a significant upward trajectory since 2018. Furthermore, our findings identify five critical factors that significantly influence public perceptions of AI risks: national income levels, infrastructure completeness, health security index, regulatory quality, and electoral democracy index. 1.3 Research Gaps Despite growing scholarly interest in public perceptions of AI, significant gaps remain in our understanding, particularly regarding global-scale empirical studies. It is worth exploring in more depth the determinants of human risk perception of AI (Krieger, J. B. et al., 2024). Although existing literature has made attempts to study the public's perception of artificial intelligence (AI) based on big data from social media, the researchers have acknowledged that their studies have two main limitations. Firstly, the data used only comes from the Twitter (X), thus overlooking the differences in the usage habits of netizens from different countries. Secondly, the sample tweets are predominantly in English. Despite English being one of the most widely used languages in the world, there are still objective limitations in applying sentiment analysis to texts in different languages (Demirel, S. et al., 2024 ). On the one hand, our study has significantly expanded the sample size, incorporating 430 pieces of data from both the Twitter (X) and Reddit. On the other hand, our data sources cover users of multiple languages from around the globe, and we have employed a comprehensive analytical approach to enhance the robustness of the results. In addition to the single factor of income level (Demirel, S. et al., 2024 ), our study also takes into account the contributions of factors such as educational level, religious control, and political trust to the public's perception of AI risks. Therefore, our study not only includes data from multiple social platforms but also examines the social impacts and ethics of AI technology, providing a spatiotemporal sequence analysis within a multilingual and multicultural context. 2 Data and Methodology This study employs an comprehensive theoretical framework that integrates concepts from sociology, psychology, and computational social science to deeply explore the complexity and multilayered nature of Artificial Intelligence (AI) risk perception. The methodology is grounded in the following core conceptual model: the discourse analysis framework. Based on discourse analysis, this study views AI risk perception as a social construct shaped by discursive activities. This framework reveals how public perceptions of potential AI risks are gradually formed through collective expressions in media, policy discussions, and social platforms. These perceptions are not static; rather, they are dynamically generated through multilayered social interactions, involving the linguistic construction of risk, symbolic meaning, and cultural formation. Thus, AI risk perception can be understood as "embedded in discursive practices," where individuals and groups continuously assign new meanings to risks through discursive exchanges. Consequently, this paper develops a multidimensional and dynamic model of risk perception that captures the complex role of social, cultural, and geographical factors in shaping AI risk perception. This theoretical integration provides a novel perspective on how AI technologies are perceived and accepted across diverse global social contexts. 2.1 Data Collection This study captures public perceptions of AI risks through multiple online platforms, with the main data sources being Twitter and Reddit. From 2010 to 2023, 4.3 million related social media posts were collected. These platforms were selected due to their extensive user bases and the high frequency of AI-related discussions. The data covers multiple languages and global regions. To further analyze the relationship between AI risk perception and social context, supplemental data was gathered from sources such as the World Bank and Global Data Lab, encompassing variables like income levels, education indices, infrastructure, and political stability. Supplemental data included intercultural competence indices derived from Hofstede's cultural dimensions and the Global Ethics Survey, recognizing that risk perception constructs vary significantly across collectivist versus individualist societies (O'Regan & Ferri, 2025). Table 1 reports the descriptive statistics of the sample data, which include mean, standard deviation, maximum and minimum values. Table 1 Descriptive Statistics Variable Name Mean Std. Dev Min Max Income (USD) 28,512.37 14,123.57 512.345 214,982.784 Education Index 0.478 0.117 0.112 0.665 Infrastructure Completeness (%) 64.53% 10.57% 40.15% 94.68% Global Health Security Index 0.458 0.084 0.105 0.679 Political Stability Index 0.525 0.157 0.124 0.85 Night Light Index 35.789 12.487 10.356 78.123 Government Effectiveness Index 0.652 0.124 0.214 0.919 Electoral Democracy Index 0.478 0.125 0.275 0.674 Quality of Regulations Index 0.654 0.129 0.212 0.797 Gini Coefficient 0.421 0.117 0.209 0.654 Religious Influence Index 0.553 0.185 0.101 0.873 Corruption Control Index 0.625 0.153 0.223 0.879 Liberal Component Index 0.451 0.134 0.257 0.674 Life Expectancy (years) 62.487 8.521 55.013 85.471 Population aged 15 to 64 (%) 63.23% 6.24% 50.10% 75.79% Transparency Concerns (%) 23.54% 5.78% 5.23% 45.01% Innovation Control Concerns (%) 21.71% 4.69% 4.14% 38.48% Ethical Concerns (%) 19.87% 5.14% 6.05% 42.21% Decision Support Concerns (%) 18.26% 4.74% 3.91% 35.68% Job Market Risks Concerns (%) 25.31% 5.57% 8.51% 47.82% Privacy Concerns (%) 24.89% 5.62% 5.15% 45.92% Security Vulnerabilities (%) 22.34% 4.82% 3.05% 40.47% 2.2 Data Preprocessing To accurately classify public perceptions of AI risks, this study combined automated tools with manual annotation. Natural Language Processing (NLP) techniques—including Python’s NLTK, spaCy, and Gensim—were used to automatically identify AI-related sentiments and topics. Manual annotations were completed by 8 experts to ensure classification accuracy, achieving a Fleiss' Kappa reliability score of 0.77. Topic modeling was then used to categorize the content into multiple themes, such as privacy concerns, system bias, and job displacement. Each category’s influence on the public and society was weighted based on the literature. To facilitate global comparisons, the composite index of public perceptions of AI risks for each geographical location is normalized. As mentioned earlier, this study was achieved using min-max scaling. The normalized indices are then visually represented on a map through color gradients, where different colors correspond to varying levels of public perception of AI risk indices, making it easier to identify regions of high and low impact. 2.3 Analytical Techniques This study opts for machine learning methods based on ensemble models, including XGBoost, Random Forest, Extra Trees, and AdaBoost, due to their multifaceted advantages. First, ensemble methods significantly enhance model prediction performance. By combining the predictions of multiple base models, ensemble models effectively reduce the variance of the model, thereby improving the accuracy of overall predictions. Second, ensemble methods reduce the risk of overfitting. As ensemble models consider the opinions of multiple base models, they possess strong resistance to overfitting, producing more robust and generalizable prediction outcomes. Additionally, ensemble methods provide an assessment of feature importance, aiding in identifying key factors influencing model predictions, offering vital clues for further analysis. Lastly, ensemble methods can handle imbalanced data, are applicable to various data types, and offer a degree of model interpretability, providing researchers with explanations and understandings of model prediction results. By selecting appropriate base models, training them on the training set, and obtaining their prediction results, these results serve as input values for a meta-model, which synthesizes these base model predictions. The final effectiveness of the ensemble model is determined by predicting and evaluating model performance on the test set. Through this ensemble approach, attribution analysis and contribution analysis of public perceptions of AI risks and types across different countries and regions are conducted, revealing the contributions of factors such as income level, education level, religious control, and political trust to public perceptions of AI risks. By leveraging the strengths of each base model, a more comprehensive and accurate understanding and assessment of factors affecting AI risk perceptions and their impacts are achieved. 2.4 Validation and Extrapolation To ensure the robustness and generalizability of the study’s findings, convex hull validation techniques were applied, covering over 90% of global socioeconomic variability. Cross-validation was employed to ensure the continuity and accuracy of the model’s predictions. This study provides a solid foundation for understanding the spatiotemporal patterns of public AI risk perceptions and helps to identify key factors influencing risk perception and their global distribution. 3 Results This section presents the findings from the spatiotemporal analysis of global public risk perceptions of AI. The results are organized into three main subsections: the identification of key patterns in AI risk perceptions, the factors influencing these perceptions, and the evolution of public concerns over time. 3.1 Global Patterns in Public Perceptions of AI Risks To assess the interpolation and extrapolation capabilities of the public perception of AI risk index, namely the representativeness of the dataset in the multidimensional covariate space, this study employed a method based on Principal Component Analysis (PCA). It includes 16 principal components, covering global education levels and other specific variables (for detailed variables, refer to the data sources). These covariates are mean-centered, standardized, and mapped to the same PCA space using eigenvectors. In the significant principal components, convex hulls for each pair of variables are constructed, which account for over 90% of the sample space variability. Based on the coordinates of these convex hulls, each data point of the public perception of AI risk index is categorized to determine whether it falls inside or outside these convex hulls. The study finds that 86.4% of the risk perception index data points are within the convex hull space, covering 95% or more of global regions. This indicates that in most global areas, the sample's representativeness and extrapolation capabilities perform well. Spatial analysis reveals distinct risk perception clusters correlating with cultural value orientations (Fig. 1 ). Table 2 shows how these perceptions are distributed across continents. Regional development levels and social structure differences reflect the public's concerns and apprehensions about AI technology in various aspects. Table 2 Perception of various types of AI technologies Continent Information Transparency (%) Innovation Control (%) Ethical Concerns (%) Decision Support (%) Job Market (%) Privacy (%) Security Vulnerabilities (%) Asia 17.95 17.62 10.27 8.9 29.48 20.79 23.2 Africa 21.81 15.96 17.94 12.94 18.46 17.89 11.91 Europe 11.19 12.48 19.87 16.02 15.93 18.71 22.16 North America 9.44 9.56 18.79 27.23 8.11 14.86 21.64 South America 14.47 28.51 11.06 21.19 17.54 13.74 8.32 Oceania 25.14 15.87 22.07 13.72 10.48 14.01 12.77 3.2 Factors Influencing Public Risk Perceptions In Fig. 2 , the contributing factors to public perceptions of AI risks are depicted, which directly or indirectly influence the public's understanding and concerns about technology, especially AI technology. This study utilizes machine learning methods based on ensemble models, including XGBoost, Random Forest, Extreme Trees, and AdaBoost. These models assess and analyze the impact of various social, economic, political, and technological factors on the public risk perception index. The length of the bars represents the contribution of each indicator in the model, while the irregular curves within the bars reflect the partial correlation relationships between the indicators and AI risk perception, indicating the correlation between individual variables and public perception of AI risks; as the indicator values within the bars increase or decrease, the trend of AI risk perception correspondingly rises or falls. The model is built on representative impact variables identified within the conceptual framework, presenting the relative contribution of 16 predictive variables after excluding certain collinear and statistically insignificant variables. Overall, income, infrastructure completeness, health security index, quality of regulations, and electoral democracy index are significant influencing variables on the public's perception of AI risks. Specifically, economic factors such as income level, infrastructure completeness, Gini coefficient, and night-time light index make important contributions to the model. Figure 3 provides a clearer and more concise visualization by aggregating minor features, identifying the main drivers of model predictions by examining the vertical distribution (density) of each feature point and comparing the relative importance of features. Features with denser clusters of points with higher absolute SHAP values are considered more influential than features with points close to zero. Results indicate that indicators such as "income," "infrastructure completeness," and "global health security index" have a strong and varied impact on model predictions; political indicators like "regulatory quality" and "electoral democracy index" also have an impact, mostly positive; whereas demographic features have a non-significant impact on predictions of AI risk perception. The description of individual indicators for different continents is shown in Appendix Table A.1. and Appendix Table A.2. The specific impact of each indicator on AI risk perception is shown in Appendix Table A.3. 3.3 Temporal Evolution of AI Risk Perceptions The longitudinal analysis of the data from 2010 to 2023 shows a clear upward trend in public concern regarding AI risks, particularly after 2018. This spike correlates with several key events, including advancements in AI capabilities, widespread media coverage of AI-related controversies, and high-profile data breaches involving AI technologies. From 2010 to 2017, public concerns predominantly focused on labor market disruptions, with 68% of discussions emphasizing automation risks. However, post-2018 marked a paradigm shift: privacy violations (42%) and algorithmic bias (35%) emerged as dominant themes, coinciding with high-profile data breaches (e.g., Facebook-Cambridge Analytica) and AI ethics debates. This shift can be attributed to growing awareness of AI's role in influencing key societal functions, such as hiring practices, law enforcement, and financial services (Rebekka S. et al., 2023). Public discourse around these issues was further amplified by media reports on discriminatory outcomes produced by AI systems, leading to heightened concern over fairness and transparency. Furthermore, the analysis revealed significant regional variation in the evolution of these concerns. In North America and Europe, privacy issues have shown the steepest rise, while in regions such as Southeast Asia and Latin America, concerns over employment and automation have remained more prominent. This divergence reflects differing regional priorities, with more developed regions focusing on data protection and ethical AI, and less developed regions prioritizing economic stability and job security (Wissing, B. G., & Reinhard, M. A., 2018). 4 Discussion In this section, we discuss the implications of these results for policymakers, researchers, and industry stakeholders, with a focus on understanding intercultural differences, the role of income and infrastructure, and the evolving nature of AI-related concerns. 4.1 Regional Variations in Risk Perception The analysis of public perceptions across the dataset revealed several dominant themes in AI-related risk concerns. One of the most striking results of this study is the clear regional differentiation in AI risk perceptions. Appendix Table A.3 presents the factors driving the differences in public risk perception across continents, the impact on local societies and policies, and the likely continued impact of AI in the future. Our analysis is based on the results presented in Fig. 1 and Appendix Table A.3. The divergence in risk priorities between individualist (e.g., North America) and collectivist societies (e.g., Southeast Asia) aligns with Hofstede’s cultural dimensions theory (Hofstede, G., 1980 ). In individualist contexts, personal autonomy amplifies concerns over data sovereignty, whereas collectivist emphasis on economic stability heightens sensitivity to labor market shocks. In North America and Europe, public concerns are predominantly centered around privacy violations and data misuse. This is likely driven by a combination of factors, including widespread use of digital services, stricter data protection regulations, and greater public awareness of the risks associated with AI-driven data collection. These findings align with previous research that highlights how privacy issues are becoming increasingly salient in more digitally connected societies (McDuff, D., 2018). These regions, with advanced technological infrastructure and highly digitized lifestyles, involve massive collection, storage, and analysis of personal data covering aspects from social media activities to financial transactions. This aligns with studies indicating that regions with advanced digital infrastructures are more attuned to the risks posed by widespread data collection and surveillance (Ferrer, X. et al., 2021 ). Our result has also similarities with the findings of Bao et al.(2022). In their study, about one-third of the U.S. public in the sample worried about the unintended consequences of AI and expressed serious concerns about the low level of control over the use of the data, and the possibility that AI could exacerbate discrimination on the basis of race, income, health risks, gender, religion, and sexual orientation, and about nearly thirty percent of the sample expressed concerns about discrimination based on AI, while recognizing the morality and usefulness of AI. In contrast, regions with emerging economies, exhibited heightened concerns over the impact of AI on employment. The fear of job displacement due to automation was especially prominent in countries with less robust labor markets. Respondents in these regions often expressed concern over the potential of AI to exacerbate unemployment and deepen economic inequality. These concerns reflect the more immediate and tangible impact of AI technologies on labor markets that are already facing challenges related to economic instability and high unemployment rates (Bao, L. et al., 2022 ). ‘Trust’ significantly influences their risk-benefit perceptions and acceptance (Siegrist, M., 2021 ). Interestingly, as the largest developing country, China's risk perception attitude toward AI is divided into two main views: trust and transparency, and security vulnerabilities, which are widely distributed in some less developed regions, but China's situation is more special. Since our data comes mainly from Twitter (X) and reddit, communities where Chinese posters may themselves be among those more receptive to emerging technologies, the following analyses are based on existing posts in the database. Although the distribution of these two views does not follow the law of economic and social development level, in general, people in the more economically developed eastern region are more worried about security loopholes, and people in the western inland region are more worried about trust and transparency. In a nationwide sampling questionnaire survey on Chinese people, Yina Z. et al., (2021) found that the level of trust in regulators is lower than in scientists, but the strength of impact of trust in regulators is higher than that in scientists, and even than knowledge of artificial intelligence, which is measured by perceived familiarity of artificial intelligence (Yina Zhu, & Guangxi He., 2021). With regard to the different dimensions of trust in scientists, the competence-based trust has much stronger influence than value-based one. This means that social trust mechanisms are more important than rational cognitive mechanisms, and trust in management authority (expert) systems is more important than trust in knowledge authority (expert) systems. Therefore, within China, the impact of political trust on people’s AI risk perception is well worth further investigation. Moreover, while non-Western AI ethics frameworks adopt Western terminology such as ‘fairness’ and ‘transparency’ in their formal structures (e.g., Thailand's Digital Ethics Guidelines), substantial divergences emerge in their underlying cultural logics (Hongladarom, S., & Bandasak, J.,2024). For instance, China’s Governance Principles for New Generation Artificial Intelligence emphasize the Confucian value of ‘harmonious coexistence’ (Laskai & Webster, 2019), whereas the ‘Be ethical’ provision in Thailand’s guidelines deliberately refrains from explicit definition of ethical parameters, implicitly invoking the Theravada Buddhist conception of karma (Hongladarom, 2021 ). This phenomenon underscores the necessity for cross-cultural technology ethics to critically examine how cultural traditions engender differentiated interpretations of ostensibly ‘universal’ principles. 4.2 The Role of Intercultural Factors Machine learning models were employed to assess the socio-economic and political variables that influence public perceptions of AI risks. Our analysis identified several key factors that had significant impacts on risk perception, including income levels, infrastructure completeness, health security, and the electoral democracy index. Appendix Table A.4 presents several core factors that contribute to risk perception heterogeneity. Macroeconomic structures and income. Economic inequality, pressures for industrial transformation, income disparities and economic restructuring have heightened concerns about AI replacing traditional labour, especially in developing countries. Income level emerged as the most influential variable, accounting for approximately 39.4% of the variance in public AI risk perceptions. People in developed countries tend to have greater concerns about privacy violations, which may be due to their greater use of digital technologies and greater emphasis on data protection (McDuff, D. et al., 2018 ). Conversely, in lower-income regions, the primary concern was the impact of automation on job security, as automation technologies were perceived as a direct threat to livelihoods in economies with more vulnerable employment sectors (Zhang, W. et al., 2021). Income emerged as the most significant socio-economic factor influencing AI risk perceptions. This is in line with the “digital divide” hypothesis (Zhang, W. et al., 2021). Governance and regulatory systems. Policy consistency and government credibility: public trust in AI technologies is more stable in countries with high political trust and clear policies, especially in terms of privacy and security protection. Government policy stability and credibility affect the public's attitude towards AI. If the government is able to formulate clear and reasonable policies to regulate the development of AI and effectively safeguard the public's privacy and security, the public will be more willing to embrace AI technologies. As data collection and use become more frequent with the widespread adoption of AI technologies, the public in these regions are more concerned about the potential for misuse of personal data, and therefore have higher expectations of privacy and data security measures for AI. Social and cultural capital. Manifested in differences in education and technological literacy, as well as conflicts in ethical and social values. Technical education and socio-cultural context influence public understanding and trust in AI technologies. Regions with high quality education are more open to AI applications. The public in Oceania (22.07%) and Europe (19.87%) are concerned about the ethical challenges posed by AI, particularly the issues of algorithmic bias and transparency in decision-making. Social values and ethics in these regions have led to a higher demand for fairness and morality in AI, and they are concerned that AI systems may produce unfair results due to irrational design of algorithms or lack transparency in the decision-making process, which may affect the interests of individuals and society. In some regions, religion and traditional culture deeply influence the public's moral judgement and acceptance of AI, especially in developing countries. Religious beliefs and traditional cultural values cause people to look at AI from the perspective of ethics and social responsibility, and consider whether it is compatible with the values and ethical standards of the society, which may lead to discussions about ethical standards for AI. Technology development path. It is manifested as technology maturity and infrastructure expansion, decision support and innovation incentive. Advanced technology ecosystems and infrastructure development support AI deployment and increase public recognition of its capabilities. For example, in some technologically advanced regions, such as North America and parts of Asia, well-developed network infrastructure and advanced technology R&D environments enable AI technologies to be better applied and demonstrate their benefits, thus making it easier for the public to recognise the potential value of AI. The public in North America (27.23 per cent) and South America (21.19 per cent) have high expectations of the role of AI in decision support and improving governance efficiency, seeing it as a potentially transformative tool. Geopolitical and environmental factors. Highly urbanised areas are more receptive to AI applications, while geopolitical conflicts may lead to greater resistance to AI introduction in certain areas. At the same time, regions with a high degree of urbanisation usually have a better technological base and application environment, and are more likely to accept and promote AI technologies. Cross-border technology standards and international cooperation have facilitated the global proliferation of AI, but they have also exacerbated concerns about technology dependence and autonomy in developing countries. In the context of globalisation, AI technology is spreading rapidly across the globe, but developing countries may worry about losing their autonomy in the process of technology introduction and becoming overly reliant on foreign technology, as well as not being able to participate in the formulation of global technology standards, thus affecting their own development in the field of AI. 4.3 The Evolving Nature of AI Risk Perceptions A temporal analysis of AI risk perceptions over the past decade has uncovered a dynamic shift in public concerns, particularly from 2018 onwards. This surge in concern correlates with several pivotal events, including advancements in AI capabilities, extensive media coverage of AI-related controversies, and high-profile data breaches involving AI technologies. Initially, concerns were predominantly focused on the economic impacts of automation. However, in recent years, there has been a marked increase in anxiety over privacy violations and algorithmic bias. This evolution is closely linked to both technological progress and heightened media scrutiny of AI-related issues, including high-profile data breaches and instances of discriminatory outcomes from AI systems (Rebekka S. et al. 2023). From a macro perspective, AI systems, being based on historical data, are prone to reproduce biases and discriminations present in human history in new and opaque ways. Bias and discriminatory systems also have meso-level effects, as negative impacts are not evenly distributed but are concentrated among those who are already marginalized (Henrik S. S., 2023). In government services and finance, there is a growing emphasis on algorithms determining "digital destinies," which become primary determinants of individuals' life opportunities, leading to concerns about the marginalization of certain groups. Furthermore, public apprehension regarding AI decision support arises from its lack of transparency and interpretability. AI systems are often perceived as black boxes, making it difficult to understand their specific reasoning and judgment processes (Barocas, S., & Selbst, A. D., 2016). Despite the competitive advantages offered by AI-powered systems, their black-box nature lacks transparency and hinders the explanation of their decisions (Minh, D. et al., 2022 ). The design and application of AI systems should be grounded in public reasons and universally accepted principles, rather than catering exclusively to the needs of specific interest groups. This approach helps ensure that AI systems serve the interests of the entire society (Buccella, A., 2023 ). Additionally, concerns about "the impact of automation on the job market" have shown steady growth, reflecting apprehensions about the effects of automation technology on unemployment. This study further elucidates these concerns. With the development and application of AI technology, jobs are being replaced by automation to some extent, leading to issues of unemployment or difficulty in career transition. The capability of AI is currently expanding beyond mechanical and repetitive tasks to analytical and thinking tasks (Huang, M.-H. et al., 2019), with workers in lower-skilled jobs bearing the brunt, but managers in decision-making positions are increasingly vulnerable. Futurists predict that by 2025, a third of jobs that exist today could be taken by Smart Technology, Artificial Intelligence, Robotics, and Algorithms (STARA). The emergence of STARA could spell the end of successful career planning, reinforcing the turbulent changes in borderless careers that are likely to become more common in the future due to technological change (Brougham, D., & J. Haar., 2018). Moreover, while AI technology may create new job opportunities, these opportunities often require higher levels of technical proficiency and specialized knowledge, which may be challenging for certain segments of the population to adapt to, exacerbating uncertainty and unfairness in the job market. The increasing complexity of AI applications, particularly in fields such as law enforcement, finance, and healthcare, has led to growing public scrutiny. Concerns over fairness, transparency, and accountability have become more pronounced as AI technologies are increasingly embedded in decision-making processes that have direct consequences for individuals and communities (O'Reilly-Shah, V. N. et al., 2020). This trend suggests that as AI continues to evolve and permeate more aspects of society, public perceptions of its risks will also become more multifaceted, requiring policymakers to address a broader range of ethical and regulatory challenges. 4.4 Implications for Policy and Governance The findings of this study offer important implications for the governance of AI technologies. First, policymakers need to recognize the intercultural differences in AI risk perceptions and tailor their regulatory frameworks accordingly. In regions where privacy concerns are dominant, strengthening data protection laws and ensuring greater transparency in AI applications will be key to building public trust. In contrast, regions where job displacement and economic instability are more pressing concerns may require policies that focus on workforce retraining and education to mitigate the negative impacts of automation (Gao, S. et al., 2020 ). Additionally, as AI systems become more integrated into everyday life, there is a growing need for regulatory mechanisms that address the ethical implications of AI, particularly in areas such as bias and fairness. Public concern over algorithmic bias and the lack of transparency in AI decision-making processes indicates a need for clearer standards and oversight to ensure that AI systems are developed and deployed in a manner that is both fair and accountable (Siegrist, M., 2021 ). 5 Conclusion and limitations 5.1 Conclusion The multidimensional mathematical convex hull approach was utilized to extrapolate public geospatial data, offering an in-depth analysis of the spatiotemporal distribution characteristics, regional heterogeneity, and potential influencing factors of global public perceptions towards AI risks. Concerns about AI exacerbating bias in decision support systems have been the public's most significant perceived risks from AI since 2015. The public in regions with higher levels of economic development, education, and infrastructure are increasingly concerned about the potential invasion of personal privacy by the advancement of artificial intelligence. These concerns have seen a significant rise since 2018. At the same time, our findings indicate that public perceptions of AI risks are heavily influenced by sociocultural, economic, and political factors. Income, infrastructure completeness, health security index, quality of regulations, and electoral democracy index are significant influencing variables on the public's perception of AI risks. Economic factors such as income level, infrastructure completeness, Gini coefficient, and night-time light index also make important contributions to the model. Furthermore, our results demonstrate a growing public apprehension regarding AI bias, transparency, and fairness over time. 5.2 Limitations Although this study attempts to capture a broad perception of AI risks among the global public, balancing depth and breadth has always been a challenge. It's difficult to delve into every detail when exploring such a wide topic. Particularly, there's insufficient depth in the interpretation of AI risk perceptions across different cultural contexts. Cultural values, historical backgrounds, and social structures are significant factors influencing public perceptions. Our study’s Western-centric social media data (70% from Twitter (X)/Reddit) may marginalize Global South perspectives—a limitation demanding participatory ethnography in future work. Additionally, there's a certain gap in linking public perceptions with policymaking. While identifying public concerns about AI risks is an important first step, translating these concerns into effective policies and practical measures is equally crucial. This requires interdisciplinary efforts to integrate knowledge from technology, law, ethics, and social sciences, exploring bridges from perception to policy implementation. For instance, research could explore which types of public education programs are most effective or how to design policies to address the public's most concerned AI risk issues. Declarations Disclosure of interest The authors report there are no competing interests to declare. Author Contribution Z.L.: Original Draft; Formal analysis; SupervisionM.L.: Original Draft; Review & EditingL.L.: Review & EditingJ.H.: Original Draft; Methodology; Formal analysis; VisualizationAll authors reviewed the manuscript. Data Availability The data that support the findings of this study are openly available in figshare at https://doi.org/10.6084/m9.figshare.25305244.v2, reference number. References Afroogh S, Akbari A, Malone E, Kargar M, Alambeigi H (2024) Trust in AI: progress, challenges, and future directions. Humanit Social Sci Commun 11(1568). https://doi.org/10.1057/s41599-024-04044-8 Ajenaghughrure I, Sousa S, Lamas D (2021) Psychophysiological modeling of trust in technology: Influence of feature selection methods. Proceedings of the ACM on Human-Computer Interaction, 5(EICS), 1–25. https://doi.org/10.1145/3459745 Al-Gasawneh J, Al-Hawamleh A, Alorfi A, Al-Rawashdeh G (2022) Moderating the role of the perceived security and endorsement on the relationship between perceived risk and intention to use the artificial intelligence in financial services. Int J Data Netw Sci 6(3):743–752. https://doi.org/10.5267/j.ijdns.2022.3.007 Bao L, Krause NM, Calice MN, Scheufele DA, Wirz CD, Brossard D, Newman TP, Xenos MA (2022) Whose AI? How different publics think about AI and its social impact. Comput Hum Behav 130:107182. https://doi.org/10.1016/j.chb.2022.107182 Barocas S, Selbst AD (2016) Big data's disparate impact. Calif Law Rev 671–732. https://doi.org/10.15779/Z38BG31 Bengio Y, Hinton G, Yao A, Song D, Abbeel P, Harari YN, Mindermann S (2023) Managing ai risks in an era of rapid progress. arXiv preprint arXiv :231017688 Bedué P, Fritzsche A (2022) Can we trust AI? An empirical investigation of trust requirements and guide to successful AI adoption. J Enterp Inform Manage 35(2):530–549. https://doi.org/10.1108/JEIM-06-2020-0233 Brougham D, Haar J (2018) Smart Technology, Artificial Intelligence, Robotics, and Algorithms (STARA): Employees’ perceptions of our future workplace. J Manage Organ 24(2):239–257. https://doi.org/10.1017/jmo.2016.55 Bologheanu R, Kapral L, Laxar D, Maleczek M, Dibiasi C, Zeiner S, Agibetov A et al (2023) Development of a reinforcement learning algorithm to optimize corticosteroid therapy in critically ill patients with sepsis. J Clin Med 12:1513. https://doi.org/10.3390/jcm12041513 Buccella A (2023) AI for all’ is a matter of social justice. AI Ethics 3(4):1143–1152. https://doi.org/10.1007/s43681-022-00222-z Chang H (2023) Analysis of sentiment towards artificial intelligent industry using hybrid natural language processing technique. In 2023 IEEE 8th International Conference On Software Engineering and Computer Systems (ICSECS) (pp. 363–370). IEEE. https://doi.org/10.1109/ICSECS58457.2023.10256397 Chen K, Cate A, Cheren H (2023) Communicating agriculture AI technologies: How American agricultural producers’ perception of trustworthiness, risk perception, and emotion affect their likelihood of adopting artificial intelligence in food systems. Environ Communication 17(8):1004–1019. https://doi.org/10.1080/17524032.2023.2211746 Choudhury A (2022) Factors influencing clinicians’ willingness to use an AI-based clinical decision support system. Front Digit Health 4:920662. https://doi.org/10.3389/fdgth.2022.920662 Chow JCL, Wong V, Li K (2024) Generative pre-trained transformer-empowered healthcare conversations: Current trends, challenges, and future directions in large language model-enabled medical chatbots. BioMedInformatics 4(1):837–852. https://doi.org/10.3390/biomedinformatics4010047 Davis FD (1989) Perceived Usefulness, Perceived Ease of Use, and User Acceptance of Information Technology. MIS Q 13(3):319–340. https://doi.org/10.2307/249008 Demirel S, Kahraman-Gokalp E, Gündüz U (2024) From Optimism to Concern: Unveiling Sentiments and Perceptions Surrounding ChatGPT on Twitter. Int J Human–Computer Interact 1–23. https://doi.org/10.1080/10447318.2024.2392964 Ferrer X, Nuenen Tv, Such JM, Coté M, Criado N (2021) Bias and discrimination in AI: A cross-disciplinary perspective. IEEE Technol Soc Mag 40(2):72–80. https://doi.org/10.1109/MTS.2021.3056293 Gao S, He L, Chen Y, Li D, Lai K (2020) Public perception of artificial intelligence in medical care: Content analysis of social media. J Med Internet Res 22(7):e16649. https://doi.org/10.2196/16649 Gerlich M (2023) Perceptions and acceptance of artificial intelligence: A multi-dimensional study. Social Sci 12(9):502. https://doi.org/10.3390/socsci12090502 Henrik Skaug Sætra (2023) Generative AI: Here to stay, but for good? Technology in Society, 75, 102372. https://doi.org/10.1016/j.techsoc.2023.102372 Hofstede G (1980) Culture's Consequences: International Differences in Work-Related Values. Sage Hongladarom S (2021) The Thailand national AI ethics guideline: an analysis. J Inform Communication Ethics Soc 19(4):480–491 Hongladarom S, Bandasak J (2024) Non-western AI ethics guidelines: implications for intercultural ethics of technology. AI Soc 39:2019–2032. https://doi.org/10.1007/s00146-023-01665-6 Huang J (2024) Spatial_DataSet_for_Global public risk perception of artificial intelligence. https://doi.org/10.6084/m9.figshare.25305244.v2 . figshare. Dataset Huang M-H, Rust R, Maksimovic V (2019) The Feeling Economy: Managing in the next generation of artificial intelligence (AI). Calif Manag Rev 61(4):43–65. https://doi.org/10.1177/0008125619863436 Huo W, Zhang Z, Qu J, Yan J, Yan S, Yan J, Shi B (2024) Speciesism and preference of human–artificial intelligence interaction: A study on medical artificial intelligence. Int J Human–Computer Interact 40(11):2925–2937. https://doi.org/10.1080/10447318.2023.2176985 Kerstan S, Bienefeld N, Grote G (2024) Choosing human over AI doctors? How comparative trust associations and knowledge relate to risk and benefit perceptions of AI in healthcare. Risk Anal 44:939–957. https://doi.org/10.1111/risa.14216 Krieger JB, Bouder F, Wibral M, Almeida RJ (2024) A systematic literature review on risk perception of artificial narrow intelligence. J Risk Res June 1–19. https://doi.org/10.1080/13669877.2024.2350725 Liu H, Wang Y, Fan W, Liu X, Li Y, Jain S, Tang J (2022) Trustworthy ai: A computational perspective. ACM Trans Intell Syst Technol 14(1):1–59. https://doi.org/10.1145/3546872 Liu Y, Lyu Z (2024) Changes in public perception of ChatGPT: A text mining perspective based on social media. Int J Human–Computer Interact 1–15. https://doi.org/10.1080/10447318.2024.2406966 McDuff D, Cheng R, Kapoor A (2018) Identifying bias in AI using simulation. arXiv preprint arXiv:1810.00471. Minh D, Wang HX, Li YF et al (2022) Explainable artificial intelligence: A comprehensive review. Artif Intell Rev 55:3503–3568. https://doi.org/10.1007/s10462-021-10088-y National New Generation Artificial Intelligence Governance Expert Committee (2019), June 17 Translation: Chinese Expert Group Offers ‘Governance Principles’ for ‘Responsible AI’. Digichina. Stanford University. https://digichina.stanford.edu/work/translation-chinese-expert-group-offers-governance-principles-for-responsible-ai/ Neyazi TA, Ng SWT, Hobbs M, Yue A (2023) Understanding user interactions and perceptions of AI risk in Singapore. Big Data Soc 10(2). https://doi.org/10.1177/20539517231213823 O'Reilly - Shah VN, Gentry KR, Walters AM, Zivot J, Anderson CT, Tighe PJ (2020) Bias and ethical considerations in machine learning and the automation of perioperative risk assessment. Br J Anaesth 125(6):843–846. https://doi.org/10.1016/j.bja.2020.07.040 O’Regan J, Ferri G (2025) Artificial intelligence and depth ontology: implications for intercultural ethics. Appl Linguistics Rev 16(2):797–807. https://doi.org/10.1515/applirev-2024-0189 Paté - Cornell E (2024) Preferences in AI algorithms: The need for relevant risk attitudes in automated decisions under uncertainties. Risk Anal 44:2317–2323. https://doi.org/10.1111/risa.14268 Schwesig R, Brich I, Buder J, Huff M, Said N (2023) Using artificial intelligence (AI)? Risk and opportunity perception of AI predict people’s willingness to use AI. J Risk Res 26(10):1053–1084. https://doi.org/10.1080/13669877.2023.2249927 Rainie L, Funk C, Anderson M, Tyson A (2022) AI and Human Enhancement: American’s Openness Is Tempered by a Range of Concerns: Public views are tied to how these technologies would be used, what constraints would be in place. Pew Res Cent. http://www.jstor.org/stable/resrep57310 Robleto E, Habashi A, Kaplan MAB, Riley RL, Zhang C, Bianchi L, Shehadeh LA (2024) Medical students’ perceptions of an artificial intelligence (AI) assisted diagnosing program. Med Teach 46(9):1180–1186. https://doi.org/10.1080/0142159X.2024.2305369 Sartori L, Bocca G (2023) Minding the gap(s): Public perceptions of AI and socio-technical imaginaries. AI Soc 38:443–458. https://doi.org/10.1007/s00146-022-01422-1 Schepman A, Rodway P (2022) The General Attitudes towards Artificial Intelligence Scale (GAAIS): Confirmatory validation and associations with personality, corporate distrust, and general trust. Int J Human–Computer Interact 39(13):2724–2741. https://doi.org/10.1080/10447318.2022.2085400 Siegrist M (2021) Trust and risk perception: A critical review of the literature. Risk Anal 41:480–490. https://doi.org/10.1111/risa.13325 Wissing BG, Reinhard MA (2018) Individual differences in risk perception of artificial intelligence. Swiss J Psychol 77(4):149. https://doi.org/10.1024/1421-0185/A000214 Yigitcanlar T, Degirmenci K, Inkinen T (2024) Drivers behind the public perception of artificial intelligence: Insights from major Australian cities. AI Soc 39:833–853. https://doi.org/10.1007/s00146-022-01566-0 Zhu Y, He G (2021) Can trust reduce the publics risk perception of artificial intelligence technology? Stud Sci Sci 10:1748–1757. https://doi.org/10.16192/j.cnki.1003-2053.20210425.001 Zhang B, Dafoe A (2019) Artificial intelligence: American attitudes and trends. SSRN Electron J. https://doi.org/10.2139/ssrn.3312874 Zhang W, Sun L, Wang X, Wu A (2022) The influence of AI word - of - mouth system on consumers' purchase behaviour: The mediating effect of risk perception. Syst Res Behav Sci 39(3):516–530. https://doi.org/10.1002/sres.2871 Zuiderwijk A, Chen YC, Salem F (2021) Implications of the use of artificial intelligence in public governance: A systematic literature review and a research agenda. Government Inform Q 38(3):101577. https://doi.org/10.1016/j.giq.2021.101577 Additional Declarations No competing interests reported. Supplementary Files Appendix.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-6468789","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":489581545,"identity":"5b05eabf-8362-424c-a63e-1d3f48d2d122","order_by":0,"name":"Zhichao Li","email":"","orcid":"","institution":"Shanghai Jiao Tong University","correspondingAuthor":false,"prefix":"","firstName":"Zhichao","middleName":"","lastName":"Li","suffix":""},{"id":489581546,"identity":"50d764b4-28c5-43db-ba11-188169b38f72","order_by":1,"name":"Mengxue Lyu","email":"","orcid":"","institution":"Shanghai Jiao Tong University","correspondingAuthor":false,"prefix":"","firstName":"Mengxue","middleName":"","lastName":"Lyu","suffix":""},{"id":489581548,"identity":"73bfd4db-636f-48f0-8047-f9e286d556a6","order_by":2,"name":"Lujia Li","email":"","orcid":"","institution":"Shanghai Jiao Tong University","correspondingAuthor":false,"prefix":"","firstName":"Lujia","middleName":"","lastName":"Li","suffix":""},{"id":489581549,"identity":"b4898f17-31c5-4532-afbe-360c73f312f6","order_by":3,"name":"Jilin Huang","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA7ElEQVRIie3ROwrCQBCA4VkWtImkHRu9wtooouJVIoI2FnqDiBCbiK3ewWJtRLvItBHblIJgpRBIo52Jr84kpeD+1bDMBwMLoFL9YGzETYeZ0cgdwMebkUTYm2QMQCMFCZEJT6KJ13YC4WM2pMGaihV9EfjVK4Ge7Qm4reMPo7lLpc3sskI0CPL2WbCJm0ByFjHp7VYQEeH1BGdWMmlKzz36EWmmJS25t+FxmMB0pNuWXqaM2Olq6J7620kMKU2JgpxVa8g9HQOs1wr6uL083OKI+Z7Cb+Th72vR7HwHAMXPpDvA/LhVlUql+tvux2VXaqp4RkEAAAAASUVORK5CYII=","orcid":"","institution":"Shanghai Jiao Tong University","correspondingAuthor":true,"prefix":"","firstName":"Jilin","middleName":"","lastName":"Huang","suffix":""}],"badges":[],"createdAt":"2025-04-17 06:53:24","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6468789/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6468789/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":87605302,"identity":"ee7a484e-65a3-44cd-85ce-239bb9300079","added_by":"auto","created_at":"2025-07-25 18:07:29","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":183941,"visible":true,"origin":"","legend":"\u003cp\u003eThe main geographic distribution of global public perception of various types of AI technologies\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-6468789/v1/a657ba83018c9e655c807ac7.png"},{"id":87605010,"identity":"e5d0215d-4acd-482e-8e91-67f3c7c54ad6","added_by":"auto","created_at":"2025-07-25 17:59:29","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":174367,"visible":true,"origin":"","legend":"\u003cp\u003eFeature importance illustrates the evaluation of the significance of various influencing variables on social, economic, and technological factors\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-6468789/v1/74633b4ce9e5b441825bcf13.png"},{"id":87605009,"identity":"796b8a63-e2f0-4b81-8db5-b328e7176d75","added_by":"auto","created_at":"2025-07-25 17:59:29","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":330287,"visible":true,"origin":"","legend":"\u003cp\u003eThe contributions of different factors to the Public Perception of AI Risk Index through SHAP analysis grounded in game theory\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-6468789/v1/453cde4e1d35ee1fe7b7ac50.png"},{"id":87605006,"identity":"8df46f1f-7eb9-4197-9422-1267b84a01dc","added_by":"auto","created_at":"2025-07-25 17:59:29","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":226522,"visible":true,"origin":"","legend":"\u003cp\u003eComprehensive Tracking of Global Trends in Public Perception of AI Risk Across Different Categories from 2010 to 2023. Figure 4a provides a detailed tracking of the frequency of various categories of public perception of AI risk globally from 2010 to 2023. Figure 4b ranks the different categories of public perception of AI risk\u003c/p\u003e","description":"","filename":"4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6468789/v1/0d6439ffb06f788b7a9c4c28.jpg"},{"id":87605304,"identity":"4877ec1a-d7d3-4361-981b-5fd1db258295","added_by":"auto","created_at":"2025-07-25 18:07:29","extension":"png","order_by":8,"title":"Figure 8","display":"","copyAsset":false,"role":"figure","size":183941,"visible":true,"origin":"","legend":"\u003cp\u003eThe main geographic distribution of global public perception of various types of AI technologies\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-6468789/v1/75013741489d8295647c3f2b.png"},{"id":87605309,"identity":"b401b5c6-eda7-4479-a122-742f58605d7c","added_by":"auto","created_at":"2025-07-25 18:07:29","extension":"png","order_by":9,"title":"Figure 9","display":"","copyAsset":false,"role":"figure","size":174367,"visible":true,"origin":"","legend":"\u003cp\u003eFeature importance illustrates the evaluation of the significance of various influencing variables on social, economic, and technological factors\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-6468789/v1/f6472025df8d1aeaaafecc18.png"},{"id":87605306,"identity":"68f7a30f-91bc-417f-ae66-d52205919419","added_by":"auto","created_at":"2025-07-25 18:07:29","extension":"png","order_by":10,"title":"Figure 10","display":"","copyAsset":false,"role":"figure","size":330287,"visible":true,"origin":"","legend":"\u003cp\u003eThe contributions of different factors to the Public Perception of AI Risk Index through SHAP analysis grounded in game theory\u003c/p\u003e","description":"","filename":"floatimage3.png","url":"https://assets-eu.researchsquare.com/files/rs-6468789/v1/8283868c5b06ddb6a74a9188.png"},{"id":108048288,"identity":"0e6b5974-152f-4660-9774-64e49bb861fb","added_by":"auto","created_at":"2026-04-28 20:24:45","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":2103177,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6468789/v1/856267ae-1de3-432c-8069-943e459613c7.pdf"},{"id":87605003,"identity":"faec4fcf-12c3-40b0-aabc-242fedc3d866","added_by":"auto","created_at":"2025-07-25 17:59:29","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":37414,"visible":true,"origin":"","legend":"","description":"","filename":"Appendix.docx","url":"https://assets-eu.researchsquare.com/files/rs-6468789/v1/d27de6dee68e35f6605b8c77.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Decoding Global AI Risk Perception Evolution: Intercultural drivers and Public Discourse Patterns in Algorithmic Societies","fulltext":[{"header":"1 Introduction","content":"\u003cp\u003eThe advent of Artificial Intelligence (AI) has precipitated a paradigm shift in 21st-century technological innovation while simultaneously amplifying intercultural ethical tensions in algorithmic governance (Hongladarom \u0026amp; Bandasak, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). As AI systems become increasingly integrated into critical decision-making processes and automation frameworks, their societal impact has grown exponentially. However, this rapid advancement has also generated significant public concerns regarding potential risks, particularly in areas such as algorithmic bias in decision-making systems, large-scale privacy violations, AI ethics (Zhang, B., \u0026amp; Dafoe, A., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2019\u003c/span\u003e) and the disruptive effects of automation on employment markets. Risk perception research in the field of Artificial Intelligence is crucial for developing effective risk communication strategies for informed decision making (Krieger, J. B. et al., 2024), as trust is a key factor in the acceptance and adoption of real-life technologies (Afroogh, S. et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). Further efforts are required to scrutinize the ethical, legal, and technical solutions proposed for AI governance through critical perspectives (Chang H., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). While AI's borderless nature demands international cooperation, our analysis of 4.3\u0026nbsp;million social media posts across 164 jurisdictions reveals fundamental disparities in risk perception patterns shaped by cultural cognition frameworks (Hofstede, \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e1980\u003c/span\u003e; O'Regan \u0026amp; Ferri, 2025). It is our aspiration that this research will lay the groundwork for tackling the pivotal issue of transitioning toward effective and principled AI governance.\u003c/p\u003e\u003cdiv id=\"Sec2\" class=\"Section2\"\u003e\u003ch2\u003e1.1 Public Risk Perception of AI and Socio-cultural Influences\u003c/h2\u003e\u003cp\u003eThe literature on public perception of AI risks has evolved across multiple domains, with significant contributions examining AI applications in healthcare (Bologheanu et al., \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Robleto, E. et al., \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Kerstan, S. et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Chow, J. C. L. et al., 2024; Huo, W. et al.,2024), information communication (Zhang, W. et al., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), and agriculture (Kaiping Chen et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Some researchers have focused on AI's impact on daily life through small-scale sampling (Neyazi, T. A. et al., 2023; Bedu\u0026eacute;, P., \u0026amp; Fritzsche, A., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Rebekka S. et al., 2023), others have conducted national population-based studies (Gerlich M., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Bao, L. et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Liu, Y., \u0026amp; Lyu, Z., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). As AI adoption becomes increasingly widespread, understanding public risk perception has emerged as a crucial research area, particularly given its influence on technology acceptance and implementation (Yigitcanlar, T. et al., \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). However, despite AI's growing ubiquity, research examining public risk perception remains relatively limited, with existing studies often prioritizing technical aspects over societal implications (Gerlich M., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). Building on established theoretical frameworks including the Technology Acceptance Model (Davis, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e1989\u003c/span\u003e) and Risk Perception Heuristics (Slovic, 2000), recent studies demonstrate paradoxical findings. While 68% of Americans express concerns about AI-driven job displacement (Rainie, L. et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2022\u003c/span\u003e), only 23% of Chinese respondents prioritize employment risks (Bao et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eExisting research demonstrates that public trust plays a pivotal role in determining the successful adoption and diffusion of AI technologies. Trust in AI reduces perceived risk (Ajenaghughrure et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2021\u003c/span\u003e; Choudhury, A., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). The emergence of \"trustworthy AI\" as a fundamental principle in development and regulation underscores the importance of fairness, accountability, and transparency in AI systems (Liu, H. et al., \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). When public trust erodes or fear of AI risks intensifies, it can impede technological innovation, decelerate adoption rates, and potentially trigger societal resistance, thereby forfeiting valuable economic and social benefits (Zuiderwijk, A. et al., \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Our research inquiry focuses on examining cross-cultural variations in public trust toward artificial intelligence systems, particularly investigating how these trust differentials manifest and evolve across distinct sociocultural value orientations. As the public's perception of AI risk is multifaceted and dynamic, monitoring these changes allows policymakers and technologists to anticipate potential concerns and to address them proactively (Zhang, B., \u0026amp; Dafoe, A., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eA significant body of research highlights the ethical challenges posed by AI systems. Studies have documented how AI-powered decision-making systems can perpetuate existing biases, leading to discriminatory outcomes across various domains including employment, criminal justice, and financial services (O'Reilly-Shah, V. N. et al., 2020). Additionally, concerns about privacy violations have intensified, particularly in regions with limited data protection frameworks (Ferrer, X. et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). These ethical considerations expose the tension between universalist AI governance models and culturally situated epistemologies. The observed 'ethics-washing' phenomenon in standardized guidelines (Hongladarom \u0026amp; Bandasak, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2024\u003c/span\u003e) suggests current frameworks inadequately address depth ontology differences in intercultural reasoning about algorithmic fairness.\u003c/p\u003e\u003cp\u003ePublic perceptions of AI risks emerge from complex interactions between technological characteristics, socioeconomic contexts, and culturally embedded value systems. Recent analyses of non-Western AI ethics frameworks reveal superficial adherence to universal principles masking deeper cultural ontologies that inform ethical reasoning (Hongladarom \u0026amp; Bandasak, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; O'Regan \u0026amp; Ferri, 2025). Public perceptions of AI risks are shaped by complex interactions between technological characteristics and broader socioeconomic contexts. Research has demonstrated significant variations in AI perception across different countries and regions, influenced by factors such as educational attainment, income levels, and political systems (Bengio, Y. et al., \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). The relationship between personal AI experience and risk perception has been explored through the familiarity hypothesis, which suggests that increased exposure to AI technologies correlates with lower risk perception (Chang H., \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). However, this relationship is moderated by broader societal factors and individual circumstances. Introverted traits and heightened general correlate with more positive attitudes towards AI (Schepman et al., 2022). Notable patterns emerge: societies with robust privacy regulations often exhibit heightened concerns about data misuse, while regions prioritizing social stability tend to focus more on employment-related risks from automation (McDuff, D. et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2018\u003c/span\u003e; Bao, L. et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Enhanced security perceptions of private data in e-commerce reduces the perceived risk of financial AI services (Al-Gasawneh et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Public perception is significantly shaped by media representation and cultural narratives surrounding AI. Research indicates that media portrayals\u0026mdash;whether emphasizing AI's transformative potential or highlighting potential threats to personal freedom and security\u0026mdash;can substantially influence public concerns (Wissing, B. G., \u0026amp; Reinhard, M. A., 2018; Sartori, L., \u0026amp; Bocca, G., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). These narratives contribute to broader societal uncertainty about AI's implications for social structures, employment patterns, and governance systems.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e1.2 Research objectives\u003c/h2\u003e\u003cp\u003eIn conventional studies that rely on regional questionnaire data, we frequently encounter fragmented insights into the public's perception of AI risks within a specific region or across regions (Bao, L. et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2022\u003c/span\u003e; Gerlich M., \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Neyazi, T. A. et al., 2023; Liu, Y., \u0026amp; Lyu, Z., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). However, such isolated information is inherently limited and fails to furnish policymakers with a comprehensive understanding of the broader trends in public perception (Zhang, B., \u0026amp; Dafoe, A., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Even in existing studies that characterize public risk perception based on Twitter users' tweets, there is a significant geographical skew, with approximately 70% of the sample data set originating from the United States (Pat\u0026eacute;-Cornell, E., 2024). This overrepresentation poses a challenge, as it limits the breadth of global intercultural perspectives and makes it difficult to garner a comprehensive understanding of how the public in other countries perceives AI risks. The diverse social and economic factors that contribute to these varying risk perceptions are substantial, and they cannot be accurately assessed solely based on the content of tweets. A broader perspective is critical to fully understand the different and often conflicting views of AI (Krieger, J. B. et al., 2024). This would provide a more nuanced understanding of the factors that shape risk perception across different contexts. More critically, it falls short in elucidating the temporal dynamics and evolution of the cognitive structures that underpin these perceptions.\u003c/p\u003e\u003cp\u003eTo bridge this gap, there is a pressing need for a more holistic and longitudinal approach to analyzing public sentiment towards AI. This approach should aim to synthesize dispersed data points into a cohesive narrative that reveals not only the current state of public perception but also its trajectory over time. It enables policymakers to anticipate potential areas of concern, develop targeted educational initiatives to enlighten the public on the nuances of AI technologies, and establish regulatory frameworks that protect against the misuse of AI while promoting its beneficial applications.\u003c/p\u003e\u003cp\u003eThis study addresses three fundamental research questions:\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eRQ1\u003c/b\u003e: What are the global characteristics of public perception of AI risks in a intercultural context?\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eRQ2\u003c/b\u003e: What factors influence the public's perception of AI risks?\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003e\u003cb\u003eRQ3\u003c/b\u003e: How have AI risk perception patterns evolved over time?\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003cp\u003eOur research makes several contributions to the growing body of literature on public perceptions of emerging technologies:\u003c/p\u003e\u003cp\u003e\u003cul\u003e\u003cli\u003e\u003cp\u003eWe provide a comprehensive analysis of AI risk perceptions at a global scale, illuminating patterns across diverse geographic regions.\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eOur findings offer valuable insights for policymakers and stakeholders in developing strategies that promote technological advancement while maintaining public trust and acceptance.\u003c/p\u003e\u003c/li\u003e\u003cli\u003e\u003cp\u003eThe research promotes the understanding of factors for the sustainable development of AI technology in a intercultural context.\u003c/p\u003e\u003c/li\u003e\u003c/ul\u003e\u003c/p\u003e\u003cp\u003eMethodologically, we employed a sophisticated analytical framework combining multiple natural language processing tools (including Gensim and Perspective) with systematic manual review procedures. This approach was enhanced by implementing a multidimensional mathematical convex hull method for geospatial data analysis, enabling detailed examination of spatiotemporal distribution patterns, regional heterogeneity, and potential driving factors in global public perceptions of AI risks.\u003c/p\u003e\u003cp\u003eOur analysis reveals that public concerns primarily concentrate on three key areas: algorithmic bias and unfairness in decision support systems, data privacy violations, and labor market disruptions due to automation. Notably, privacy-related concerns have exhibited a significant upward trajectory since 2018. Furthermore, our findings identify five critical factors that significantly influence public perceptions of AI risks: national income levels, infrastructure completeness, health security index, regulatory quality, and electoral democracy index.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e1.3 Research Gaps\u003c/h2\u003e\u003cp\u003eDespite growing scholarly interest in public perceptions of AI, significant gaps remain in our understanding, particularly regarding global-scale empirical studies. It is worth exploring in more depth the determinants of human risk perception of AI (Krieger, J. B. et al., 2024). Although existing literature has made attempts to study the public's perception of artificial intelligence (AI) based on big data from social media, the researchers have acknowledged that their studies have two main limitations. Firstly, the data used only comes from the Twitter (X), thus overlooking the differences in the usage habits of netizens from different countries. Secondly, the sample tweets are predominantly in English. Despite English being one of the most widely used languages in the world, there are still objective limitations in applying sentiment analysis to texts in different languages (Demirel, S. et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). On the one hand, our study has significantly expanded the sample size, incorporating 430 pieces of data from both the Twitter (X) and Reddit. On the other hand, our data sources cover users of multiple languages from around the globe, and we have employed a comprehensive analytical approach to enhance the robustness of the results. In addition to the single factor of income level (Demirel, S. et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2024\u003c/span\u003e), our study also takes into account the contributions of factors such as educational level, religious control, and political trust to the public's perception of AI risks. Therefore, our study not only includes data from multiple social platforms but also examines the social impacts and ethics of AI technology, providing a spatiotemporal sequence analysis within a multilingual and multicultural context.\u003c/p\u003e\u003c/div\u003e"},{"header":"2 Data and Methodology","content":"\u003cp\u003eThis study employs an comprehensive theoretical framework that integrates concepts from sociology, psychology, and computational social science to deeply explore the complexity and multilayered nature of Artificial Intelligence (AI) risk perception. The methodology is grounded in the following core conceptual model: the discourse analysis framework. Based on discourse analysis, this study views AI risk perception as a social construct shaped by discursive activities. This framework reveals how public perceptions of potential AI risks are gradually formed through collective expressions in media, policy discussions, and social platforms. These perceptions are not static; rather, they are dynamically generated through multilayered social interactions, involving the linguistic construction of risk, symbolic meaning, and cultural formation. Thus, AI risk perception can be understood as \"embedded in discursive practices,\" where individuals and groups continuously assign new meanings to risks through discursive exchanges. Consequently, this paper develops a multidimensional and dynamic model of risk perception that captures the complex role of social, cultural, and geographical factors in shaping AI risk perception. This theoretical integration provides a novel perspective on how AI technologies are perceived and accepted across diverse global social contexts.\u003c/p\u003e\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\u003ch2\u003e2.1 Data Collection\u003c/h2\u003e\u003cp\u003eThis study captures public perceptions of AI risks through multiple online platforms, with the main data sources being Twitter and Reddit. From 2010 to 2023, 4.3\u0026nbsp;million related social media posts were collected. These platforms were selected due to their extensive user bases and the high frequency of AI-related discussions. The data covers multiple languages and global regions. To further analyze the relationship between AI risk perception and social context, supplemental data was gathered from sources such as the World Bank and Global Data Lab, encompassing variables like income levels, education indices, infrastructure, and political stability. Supplemental data included intercultural competence indices derived from Hofstede's cultural dimensions and the Global Ethics Survey, recognizing that risk perception constructs vary significantly across collectivist versus individualist societies (O'Regan \u0026amp; Ferri, 2025). Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e reports the descriptive statistics of the sample data, which include mean, standard deviation, maximum and minimum values.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eDescriptive Statistics\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"5\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eVariable Name\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eMean\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eStd. Dev\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eMin\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eMax\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eIncome (USD)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e28,512.37\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e14,123.57\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e512.345\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e214,982.784\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEducation Index\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.478\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.117\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.112\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.665\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eInfrastructure Completeness (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e64.53%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e10.57%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e40.15%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e94.68%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGlobal Health Security Index\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.458\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.084\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.105\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.679\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePolitical Stability Index\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.525\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.157\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.124\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.85\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNight Light Index\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e35.789\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e12.487\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e10.356\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e78.123\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGovernment Effectiveness Index\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.652\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.124\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.214\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.919\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eElectoral Democracy Index\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.478\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.125\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.275\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.674\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eQuality of Regulations Index\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.654\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.129\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.212\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.797\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eGini Coefficient\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.421\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.117\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.209\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.654\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eReligious Influence Index\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.553\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.185\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.101\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.873\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCorruption Control Index\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.625\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.153\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.223\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.879\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLiberal Component Index\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e0.451\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e0.134\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e0.257\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e0.674\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eLife Expectancy (years)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e62.487\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e8.521\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e55.013\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e85.471\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePopulation aged 15 to 64 (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e63.23%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e6.24%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e50.10%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e75.79%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eTransparency Concerns (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e23.54%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e5.78%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e5.23%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e45.01%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eInnovation Control Concerns (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e21.71%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e4.69%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e4.14%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e38.48%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEthical Concerns (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e19.87%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e5.14%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e6.05%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e42.21%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eDecision Support Concerns (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e18.26%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e4.74%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e3.91%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e35.68%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eJob Market Risks Concerns (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e25.31%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e5.57%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e8.51%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e47.82%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003ePrivacy Concerns (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e24.89%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e5.62%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e5.15%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e45.92%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSecurity Vulnerabilities (%)\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e22.34%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e4.82%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e3.05%\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e40.47%\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003e2.2 Data Preprocessing\u003c/h2\u003e\u003cp\u003eTo accurately classify public perceptions of AI risks, this study combined automated tools with manual annotation. Natural Language Processing (NLP) techniques\u0026mdash;including Python\u0026rsquo;s NLTK, spaCy, and Gensim\u0026mdash;were used to automatically identify AI-related sentiments and topics. Manual annotations were completed by 8 experts to ensure classification accuracy, achieving a Fleiss' Kappa reliability score of 0.77. Topic modeling was then used to categorize the content into multiple themes, such as privacy concerns, system bias, and job displacement. Each category\u0026rsquo;s influence on the public and society was weighted based on the literature.\u003c/p\u003e\u003cp\u003eTo facilitate global comparisons, the composite index of public perceptions of AI risks for each geographical location is normalized. As mentioned earlier, this study was achieved using min-max scaling. The normalized indices are then visually represented on a map through color gradients, where different colors correspond to varying levels of public perception of AI risk indices, making it easier to identify regions of high and low impact.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\u003ch2\u003e2.3 Analytical Techniques\u003c/h2\u003e\u003cp\u003eThis study opts for machine learning methods based on ensemble models, including XGBoost, Random Forest, Extra Trees, and AdaBoost, due to their multifaceted advantages. First, ensemble methods significantly enhance model prediction performance. By combining the predictions of multiple base models, ensemble models effectively reduce the variance of the model, thereby improving the accuracy of overall predictions. Second, ensemble methods reduce the risk of overfitting. As ensemble models consider the opinions of multiple base models, they possess strong resistance to overfitting, producing more robust and generalizable prediction outcomes. Additionally, ensemble methods provide an assessment of feature importance, aiding in identifying key factors influencing model predictions, offering vital clues for further analysis. Lastly, ensemble methods can handle imbalanced data, are applicable to various data types, and offer a degree of model interpretability, providing researchers with explanations and understandings of model prediction results.\u003c/p\u003e\u003cp\u003eBy selecting appropriate base models, training them on the training set, and obtaining their prediction results, these results serve as input values for a meta-model, which synthesizes these base model predictions. The final effectiveness of the ensemble model is determined by predicting and evaluating model performance on the test set. Through this ensemble approach, attribution analysis and contribution analysis of public perceptions of AI risks and types across different countries and regions are conducted, revealing the contributions of factors such as income level, education level, religious control, and political trust to public perceptions of AI risks. By leveraging the strengths of each base model, a more comprehensive and accurate understanding and assessment of factors affecting AI risk perceptions and their impacts are achieved.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e2.4 Validation and Extrapolation\u003c/h2\u003e\u003cp\u003eTo ensure the robustness and generalizability of the study\u0026rsquo;s findings, convex hull validation techniques were applied, covering over 90% of global socioeconomic variability. Cross-validation was employed to ensure the continuity and accuracy of the model\u0026rsquo;s predictions. This study provides a solid foundation for understanding the spatiotemporal patterns of public AI risk perceptions and helps to identify key factors influencing risk perception and their global distribution.\u003c/p\u003e\u003c/div\u003e"},{"header":"3 Results","content":"\u003cp\u003eThis section presents the findings from the spatiotemporal analysis of global public risk perceptions of AI. The results are organized into three main subsections: the identification of key patterns in AI risk perceptions, the factors influencing these perceptions, and the evolution of public concerns over time.\u003c/p\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003e3.1 Global Patterns in Public Perceptions of AI Risks\u003c/h2\u003e\u003cp\u003eTo assess the interpolation and extrapolation capabilities of the public perception of AI risk index, namely the representativeness of the dataset in the multidimensional covariate space, this study employed a method based on Principal Component Analysis (PCA). It includes 16 principal components, covering global education levels and other specific variables (for detailed variables, refer to the data sources). These covariates are mean-centered, standardized, and mapped to the same PCA space using eigenvectors. In the significant principal components, convex hulls for each pair of variables are constructed, which account for over 90% of the sample space variability. Based on the coordinates of these convex hulls, each data point of the public perception of AI risk index is categorized to determine whether it falls inside or outside these convex hulls. The study finds that 86.4% of the risk perception index data points are within the convex hull space, covering 95% or more of global regions. This indicates that in most global areas, the sample's representativeness and extrapolation capabilities perform well. Spatial analysis reveals distinct risk perception clusters correlating with cultural value orientations (Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e shows how these perceptions are distributed across continents. Regional development levels and social structure differences reflect the public's concerns and apprehensions about AI technology in various aspects.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003ePerception of various types of AI technologies\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"8\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eContinent\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eInformation Transparency (%)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eInnovation Control (%)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eEthical Concerns (%)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c5\"\u003e\u003cp\u003eDecision Support (%)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c6\"\u003e\u003cp\u003eJob Market (%)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c7\"\u003e\u003cp\u003ePrivacy (%)\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c8\"\u003e\u003cp\u003eSecurity Vulnerabilities (%)\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAsia\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e17.95\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e17.62\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e10.27\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e8.9\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e29.48\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e20.79\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e23.2\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eAfrica\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e21.81\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e15.96\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e17.94\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e12.94\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e18.46\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e17.89\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e11.91\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eEurope\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e11.19\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e12.48\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e19.87\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e16.02\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e15.93\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e18.71\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e22.16\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eNorth America\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e9.44\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e9.56\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e18.79\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e27.23\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e8.11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e14.86\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e21.64\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eSouth America\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e14.47\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e28.51\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e11.06\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e21.19\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e17.54\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e13.74\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e8.32\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003eOceania\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e\u003cp\u003e25.14\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e15.87\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e\u003cp\u003e22.07\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c5\"\u003e\u003cp\u003e13.72\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c6\"\u003e\u003cp\u003e10.48\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c7\"\u003e\u003cp\u003e14.01\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c8\"\u003e\u003cp\u003e12.77\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec12\" class=\"Section2\"\u003e\u003ch2\u003e3.2 Factors Influencing Public Risk Perceptions\u003c/h2\u003e\u003cp\u003eIn Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, the contributing factors to public perceptions of AI risks are depicted, which directly or indirectly influence the public's understanding and concerns about technology, especially AI technology. This study utilizes machine learning methods based on ensemble models, including XGBoost, Random Forest, Extreme Trees, and AdaBoost. These models assess and analyze the impact of various social, economic, political, and technological factors on the public risk perception index. The length of the bars represents the contribution of each indicator in the model, while the irregular curves within the bars reflect the partial correlation relationships between the indicators and AI risk perception, indicating the correlation between individual variables and public perception of AI risks; as the indicator values within the bars increase or decrease, the trend of AI risk perception correspondingly rises or falls. The model is built on representative impact variables identified within the conceptual framework, presenting the relative contribution of 16 predictive variables after excluding certain collinear and statistically insignificant variables. Overall, income, infrastructure completeness, health security index, quality of regulations, and electoral democracy index are significant influencing variables on the public's perception of AI risks. Specifically, economic factors such as income level, infrastructure completeness, Gini coefficient, and night-time light index make important contributions to the model.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eFigure \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e provides a clearer and more concise visualization by aggregating minor features, identifying the main drivers of model predictions by examining the vertical distribution (density) of each feature point and comparing the relative importance of features. Features with denser clusters of points with higher absolute SHAP values are considered more influential than features with points close to zero. Results indicate that indicators such as \"income,\" \"infrastructure completeness,\" and \"global health security index\" have a strong and varied impact on model predictions; political indicators like \"regulatory quality\" and \"electoral democracy index\" also have an impact, mostly positive; whereas demographic features have a non-significant impact on predictions of AI risk perception. The description of individual indicators for different continents is shown in Appendix Table A.1. and Appendix Table A.2. The specific impact of each indicator on AI risk perception is shown in Appendix Table A.3.\u003c/p\u003e\u003cp\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\u003ch2\u003e3.3 Temporal Evolution of AI Risk Perceptions\u003c/h2\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eThe longitudinal analysis of the data from 2010 to 2023 shows a clear upward trend in public concern regarding AI risks, particularly after 2018. This spike correlates with several key events, including advancements in AI capabilities, widespread media coverage of AI-related controversies, and high-profile data breaches involving AI technologies.\u003c/p\u003e\u003cp\u003eFrom 2010 to 2017, public concerns predominantly focused on labor market disruptions, with 68% of discussions emphasizing automation risks. However, post-2018 marked a paradigm shift: privacy violations (42%) and algorithmic bias (35%) emerged as dominant themes, coinciding with high-profile data breaches (e.g., Facebook-Cambridge Analytica) and AI ethics debates. This shift can be attributed to growing awareness of AI's role in influencing key societal functions, such as hiring practices, law enforcement, and financial services (Rebekka S. et al., 2023). Public discourse around these issues was further amplified by media reports on discriminatory outcomes produced by AI systems, leading to heightened concern over fairness and transparency.\u003c/p\u003e\u003cp\u003eFurthermore, the analysis revealed significant regional variation in the evolution of these concerns. In North America and Europe, privacy issues have shown the steepest rise, while in regions such as Southeast Asia and Latin America, concerns over employment and automation have remained more prominent. This divergence reflects differing regional priorities, with more developed regions focusing on data protection and ethical AI, and less developed regions prioritizing economic stability and job security (Wissing, B. G., \u0026amp; Reinhard, M. A., 2018).\u003c/p\u003e\u003c/div\u003e"},{"header":"4 Discussion","content":"\u003cp\u003eIn this section, we discuss the implications of these results for policymakers, researchers, and industry stakeholders, with a focus on understanding intercultural differences, the role of income and infrastructure, and the evolving nature of AI-related concerns.\u003c/p\u003e\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003e4.1 Regional Variations in Risk Perception\u003c/h2\u003e\u003cp\u003eThe analysis of public perceptions across the dataset revealed several dominant themes in AI-related risk concerns. One of the most striking results of this study is the clear regional differentiation in AI risk perceptions. Appendix Table A.3 presents the factors driving the differences in public risk perception across continents, the impact on local societies and policies, and the likely continued impact of AI in the future. Our analysis is based on the results presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e and Appendix Table A.3.\u003c/p\u003e\u003cp\u003eThe divergence in risk priorities between individualist (e.g., North America) and collectivist societies (e.g., Southeast Asia) aligns with Hofstede\u0026rsquo;s cultural dimensions theory (Hofstede, G., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e1980\u003c/span\u003e). In individualist contexts, personal autonomy amplifies concerns over data sovereignty, whereas collectivist emphasis on economic stability heightens sensitivity to labor market shocks. In North America and Europe, public concerns are predominantly centered around privacy violations and data misuse. This is likely driven by a combination of factors, including widespread use of digital services, stricter data protection regulations, and greater public awareness of the risks associated with AI-driven data collection. These findings align with previous research that highlights how privacy issues are becoming increasingly salient in more digitally connected societies (McDuff, D., 2018). These regions, with advanced technological infrastructure and highly digitized lifestyles, involve massive collection, storage, and analysis of personal data covering aspects from social media activities to financial transactions. This aligns with studies indicating that regions with advanced digital infrastructures are more attuned to the risks posed by widespread data collection and surveillance (Ferrer, X. et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Our result has also similarities with the findings of Bao et al.(2022). In their study, about one-third of the U.S. public in the sample worried about the unintended consequences of AI and expressed serious concerns about the low level of control over the use of the data, and the possibility that AI could exacerbate discrimination on the basis of race, income, health risks, gender, religion, and sexual orientation, and about nearly thirty percent of the sample expressed concerns about discrimination based on AI, while recognizing the morality and usefulness of AI.\u003c/p\u003e\u003cp\u003eIn contrast, regions with emerging economies, exhibited heightened concerns over the impact of AI on employment. The fear of job displacement due to automation was especially prominent in countries with less robust labor markets. Respondents in these regions often expressed concern over the potential of AI to exacerbate unemployment and deepen economic inequality. These concerns reflect the more immediate and tangible impact of AI technologies on labor markets that are already facing challenges related to economic instability and high unemployment rates (Bao, L. et al., \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2022\u003c/span\u003e).\u003c/p\u003e\u003cp\u003e\u0026lsquo;Trust\u0026rsquo; significantly influences their risk-benefit perceptions and acceptance (Siegrist, M., \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Interestingly, as the largest developing country, China's risk perception attitude toward AI is divided into two main views: trust and transparency, and security vulnerabilities, which are widely distributed in some less developed regions, but China's situation is more special. Since our data comes mainly from Twitter (X) and reddit, communities where Chinese posters may themselves be among those more receptive to emerging technologies, the following analyses are based on existing posts in the database. Although the distribution of these two views does not follow the law of economic and social development level, in general, people in the more economically developed eastern region are more worried about security loopholes, and people in the western inland region are more worried about trust and transparency. In a nationwide sampling questionnaire survey on Chinese people, Yina Z. et al., (2021) found that the level of trust in regulators is lower than in scientists, but the strength of impact of trust in regulators is higher than that in scientists, and even than knowledge of artificial intelligence, which is measured by perceived familiarity of artificial intelligence (Yina Zhu, \u0026amp; Guangxi He., 2021). With regard to the different dimensions of trust in scientists, the competence-based trust has much stronger influence than value-based one. This means that social trust mechanisms are more important than rational cognitive mechanisms, and trust in management authority (expert) systems is more important than trust in knowledge authority (expert) systems. Therefore, within China, the impact of political trust on people\u0026rsquo;s AI risk perception is well worth further investigation. Moreover, while non-Western AI ethics frameworks adopt Western terminology such as \u0026lsquo;fairness\u0026rsquo; and \u0026lsquo;transparency\u0026rsquo; in their formal structures (e.g., Thailand's Digital Ethics Guidelines), substantial divergences emerge in their underlying cultural logics (Hongladarom, S., \u0026amp; Bandasak, J.,2024). For instance, China\u0026rsquo;s Governance Principles for New Generation Artificial Intelligence emphasize the Confucian value of \u0026lsquo;harmonious coexistence\u0026rsquo; (Laskai \u0026amp; Webster, 2019), whereas the \u0026lsquo;Be ethical\u0026rsquo; provision in Thailand\u0026rsquo;s guidelines deliberately refrains from explicit definition of ethical parameters, implicitly invoking the Theravada Buddhist conception of karma (Hongladarom, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). This phenomenon underscores the necessity for cross-cultural technology ethics to critically examine how cultural traditions engender differentiated interpretations of ostensibly \u0026lsquo;universal\u0026rsquo; principles.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\u003ch2\u003e4.2 The Role of Intercultural Factors\u003c/h2\u003e\u003cp\u003eMachine learning models were employed to assess the socio-economic and political variables that influence public perceptions of AI risks. Our analysis identified several key factors that had significant impacts on risk perception, including income levels, infrastructure completeness, health security, and the electoral democracy index. Appendix Table A.4 presents several core factors that contribute to risk perception heterogeneity.\u003c/p\u003e\u003cp\u003e\u003cb\u003eMacroeconomic structures and income.\u003c/b\u003e Economic inequality, pressures for industrial transformation, income disparities and economic restructuring have heightened concerns about AI replacing traditional labour, especially in developing countries. Income level emerged as the most influential variable, accounting for approximately 39.4% of the variance in public AI risk perceptions. People in developed countries tend to have greater concerns about privacy violations, which may be due to their greater use of digital technologies and greater emphasis on data protection (McDuff, D. et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). Conversely, in lower-income regions, the primary concern was the impact of automation on job security, as automation technologies were perceived as a direct threat to livelihoods in economies with more vulnerable employment sectors (Zhang, W. et al., 2021). Income emerged as the most significant socio-economic factor influencing AI risk perceptions. This is in line with the \u0026ldquo;digital divide\u0026rdquo; hypothesis (Zhang, W. et al., 2021).\u003c/p\u003e\u003cp\u003e\u003cb\u003eGovernance and regulatory systems.\u003c/b\u003e Policy consistency and government credibility: public trust in AI technologies is more stable in countries with high political trust and clear policies, especially in terms of privacy and security protection. Government policy stability and credibility affect the public's attitude towards AI. If the government is able to formulate clear and reasonable policies to regulate the development of AI and effectively safeguard the public's privacy and security, the public will be more willing to embrace AI technologies. As data collection and use become more frequent with the widespread adoption of AI technologies, the public in these regions are more concerned about the potential for misuse of personal data, and therefore have higher expectations of privacy and data security measures for AI.\u003c/p\u003e\u003cp\u003e\u003cb\u003eSocial and cultural capital.\u003c/b\u003e Manifested in differences in education and technological literacy, as well as conflicts in ethical and social values. Technical education and socio-cultural context influence public understanding and trust in AI technologies. Regions with high quality education are more open to AI applications. The public in Oceania (22.07%) and Europe (19.87%) are concerned about the ethical challenges posed by AI, particularly the issues of algorithmic bias and transparency in decision-making. Social values and ethics in these regions have led to a higher demand for fairness and morality in AI, and they are concerned that AI systems may produce unfair results due to irrational design of algorithms or lack transparency in the decision-making process, which may affect the interests of individuals and society. In some regions, religion and traditional culture deeply influence the public's moral judgement and acceptance of AI, especially in developing countries. Religious beliefs and traditional cultural values cause people to look at AI from the perspective of ethics and social responsibility, and consider whether it is compatible with the values and ethical standards of the society, which may lead to discussions about ethical standards for AI.\u003c/p\u003e\u003cp\u003e\u003cb\u003eTechnology development path.\u003c/b\u003e It is manifested as technology maturity and infrastructure expansion, decision support and innovation incentive. Advanced technology ecosystems and infrastructure development support AI deployment and increase public recognition of its capabilities. For example, in some technologically advanced regions, such as North America and parts of Asia, well-developed network infrastructure and advanced technology R\u0026amp;D environments enable AI technologies to be better applied and demonstrate their benefits, thus making it easier for the public to recognise the potential value of AI. The public in North America (27.23 per cent) and South America (21.19 per cent) have high expectations of the role of AI in decision support and improving governance efficiency, seeing it as a potentially transformative tool.\u003c/p\u003e\u003cp\u003e\u003cb\u003eGeopolitical and environmental factors.\u003c/b\u003e Highly urbanised areas are more receptive to AI applications, while geopolitical conflicts may lead to greater resistance to AI introduction in certain areas. At the same time, regions with a high degree of urbanisation usually have a better technological base and application environment, and are more likely to accept and promote AI technologies. Cross-border technology standards and international cooperation have facilitated the global proliferation of AI, but they have also exacerbated concerns about technology dependence and autonomy in developing countries. In the context of globalisation, AI technology is spreading rapidly across the globe, but developing countries may worry about losing their autonomy in the process of technology introduction and becoming overly reliant on foreign technology, as well as not being able to participate in the formulation of global technology standards, thus affecting their own development in the field of AI.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003e4.3 The Evolving Nature of AI Risk Perceptions\u003c/h2\u003e\u003cp\u003eA temporal analysis of AI risk perceptions over the past decade has uncovered a dynamic shift in public concerns, particularly from 2018 onwards. This surge in concern correlates with several pivotal events, including advancements in AI capabilities, extensive media coverage of AI-related controversies, and high-profile data breaches involving AI technologies. Initially, concerns were predominantly focused on the economic impacts of automation. However, in recent years, there has been a marked increase in anxiety over privacy violations and algorithmic bias. This evolution is closely linked to both technological progress and heightened media scrutiny of AI-related issues, including high-profile data breaches and instances of discriminatory outcomes from AI systems (Rebekka S. et al. 2023).\u003c/p\u003e\u003cp\u003eFrom a macro perspective, AI systems, being based on historical data, are prone to reproduce biases and discriminations present in human history in new and opaque ways. Bias and discriminatory systems also have meso-level effects, as negative impacts are not evenly distributed but are concentrated among those who are already marginalized (Henrik S. S., 2023). In government services and finance, there is a growing emphasis on algorithms determining \"digital destinies,\" which become primary determinants of individuals' life opportunities, leading to concerns about the marginalization of certain groups. Furthermore, public apprehension regarding AI decision support arises from its lack of transparency and interpretability. AI systems are often perceived as black boxes, making it difficult to understand their specific reasoning and judgment processes (Barocas, S., \u0026amp; Selbst, A. D., 2016). Despite the competitive advantages offered by AI-powered systems, their black-box nature lacks transparency and hinders the explanation of their decisions (Minh, D. et al., \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). The design and application of AI systems should be grounded in public reasons and universally accepted principles, rather than catering exclusively to the needs of specific interest groups. This approach helps ensure that AI systems serve the interests of the entire society (Buccella, A., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eAdditionally, concerns about \"the impact of automation on the job market\" have shown steady growth, reflecting apprehensions about the effects of automation technology on unemployment. This study further elucidates these concerns. With the development and application of AI technology, jobs are being replaced by automation to some extent, leading to issues of unemployment or difficulty in career transition. The capability of AI is currently expanding beyond mechanical and repetitive tasks to analytical and thinking tasks (Huang, M.-H. et al., 2019), with workers in lower-skilled jobs bearing the brunt, but managers in decision-making positions are increasingly vulnerable. Futurists predict that by 2025, a third of jobs that exist today could be taken by Smart Technology, Artificial Intelligence, Robotics, and Algorithms (STARA). The emergence of STARA could spell the end of successful career planning, reinforcing the turbulent changes in borderless careers that are likely to become more common in the future due to technological change (Brougham, D., \u0026amp; J. Haar., 2018). Moreover, while AI technology may create new job opportunities, these opportunities often require higher levels of technical proficiency and specialized knowledge, which may be challenging for certain segments of the population to adapt to, exacerbating uncertainty and unfairness in the job market.\u003c/p\u003e\u003cp\u003eThe increasing complexity of AI applications, particularly in fields such as law enforcement, finance, and healthcare, has led to growing public scrutiny. Concerns over fairness, transparency, and accountability have become more pronounced as AI technologies are increasingly embedded in decision-making processes that have direct consequences for individuals and communities (O'Reilly-Shah, V. N. et al., 2020). This trend suggests that as AI continues to evolve and permeate more aspects of society, public perceptions of its risks will also become more multifaceted, requiring policymakers to address a broader range of ethical and regulatory challenges.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e\u003ch2\u003e4.4 Implications for Policy and Governance\u003c/h2\u003e\u003cp\u003eThe findings of this study offer important implications for the governance of AI technologies. First, policymakers need to recognize the intercultural differences in AI risk perceptions and tailor their regulatory frameworks accordingly. In regions where privacy concerns are dominant, strengthening data protection laws and ensuring greater transparency in AI applications will be key to building public trust. In contrast, regions where job displacement and economic instability are more pressing concerns may require policies that focus on workforce retraining and education to mitigate the negative impacts of automation (Gao, S. et al., \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e2020\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eAdditionally, as AI systems become more integrated into everyday life, there is a growing need for regulatory mechanisms that address the ethical implications of AI, particularly in areas such as bias and fairness. Public concern over algorithmic bias and the lack of transparency in AI decision-making processes indicates a need for clearer standards and oversight to ensure that AI systems are developed and deployed in a manner that is both fair and accountable (Siegrist, M., \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2021\u003c/span\u003e).\u003c/p\u003e\u003c/div\u003e"},{"header":"5 Conclusion and limitations","content":"\u003cdiv id=\"Sec20\" class=\"Section2\"\u003e\u003ch2\u003e5.1 Conclusion\u003c/h2\u003e\u003cp\u003eThe multidimensional mathematical convex hull approach was utilized to extrapolate public geospatial data, offering an in-depth analysis of the spatiotemporal distribution characteristics, regional heterogeneity, and potential influencing factors of global public perceptions towards AI risks. Concerns about AI exacerbating bias in decision support systems have been the public's most significant perceived risks from AI since 2015. The public in regions with higher levels of economic development, education, and infrastructure are increasingly concerned about the potential invasion of personal privacy by the advancement of artificial intelligence. These concerns have seen a significant rise since 2018. At the same time, our findings indicate that public perceptions of AI risks are heavily influenced by sociocultural, economic, and political factors. Income, infrastructure completeness, health security index, quality of regulations, and electoral democracy index are significant influencing variables on the public's perception of AI risks. Economic factors such as income level, infrastructure completeness, Gini coefficient, and night-time light index also make important contributions to the model. Furthermore, our results demonstrate a growing public apprehension regarding AI bias, transparency, and fairness over time.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec21\" class=\"Section2\"\u003e\u003ch2\u003e5.2 Limitations\u003c/h2\u003e\u003cp\u003eAlthough this study attempts to capture a broad perception of AI risks among the global public, balancing depth and breadth has always been a challenge. It's difficult to delve into every detail when exploring such a wide topic. Particularly, there's insufficient depth in the interpretation of AI risk perceptions across different cultural contexts. Cultural values, historical backgrounds, and social structures are significant factors influencing public perceptions. Our study\u0026rsquo;s Western-centric social media data (70% from Twitter (X)/Reddit) may marginalize Global South perspectives\u0026mdash;a limitation demanding participatory ethnography in future work.\u003c/p\u003e\u003cp\u003eAdditionally, there's a certain gap in linking public perceptions with policymaking. While identifying public concerns about AI risks is an important first step, translating these concerns into effective policies and practical measures is equally crucial. This requires interdisciplinary efforts to integrate knowledge from technology, law, ethics, and social sciences, exploring bridges from perception to policy implementation. For instance, research could explore which types of public education programs are most effective or how to design policies to address the public's most concerned AI risk issues.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003eDisclosure of interest\u003c/p\u003e\n\u003cp\u003eThe authors report there are no competing interests to declare.\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eZ.L.: Original Draft; Formal analysis; SupervisionM.L.: Original Draft; Review \u0026amp; EditingL.L.: Review \u0026amp; EditingJ.H.: Original Draft; Methodology; Formal analysis; VisualizationAll authors reviewed the manuscript.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe data that support the findings of this study are openly available in figshare at https://doi.org/10.6084/m9.figshare.25305244.v2, reference number.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAfroogh S, Akbari A, Malone E, Kargar M, Alambeigi H (2024) Trust in AI: progress, challenges, and future directions. Humanit Social Sci Commun 11(1568). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1057/s41599-024-04044-8\u003c/span\u003e\u003cspan address=\"10.1057/s41599-024-04044-8\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAjenaghughrure I, Sousa S, Lamas D (2021) Psychophysiological modeling of trust in technology: Influence of feature selection methods. Proceedings of the ACM on Human-Computer Interaction, 5(EICS), 1\u0026ndash;25. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1145/3459745\u003c/span\u003e\u003cspan address=\"10.1145/3459745\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eAl-Gasawneh J, Al-Hawamleh A, Alorfi A, Al-Rawashdeh G (2022) Moderating the role of the perceived security and endorsement on the relationship between perceived risk and intention to use the artificial intelligence in financial services. Int J Data Netw Sci 6(3):743\u0026ndash;752. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.5267/j.ijdns.2022.3.007\u003c/span\u003e\u003cspan address=\"10.5267/j.ijdns.2022.3.007\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBao L, Krause NM, Calice MN, Scheufele DA, Wirz CD, Brossard D, Newman TP, Xenos MA (2022) Whose AI? How different publics think about AI and its social impact. Comput Hum Behav 130:107182. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.chb.2022.107182\u003c/span\u003e\u003cspan address=\"10.1016/j.chb.2022.107182\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBarocas S, Selbst AD (2016) Big data's disparate impact. Calif Law Rev 671\u0026ndash;732. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.15779/Z38BG31\u003c/span\u003e\u003cspan address=\"10.15779/Z38BG31\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBengio Y, Hinton G, Yao A, Song D, Abbeel P, Harari YN, Mindermann S (2023) Managing ai risks in an era of rapid progress. arXiv preprint arXiv :231017688\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBedu\u0026eacute; P, Fritzsche A (2022) Can we trust AI? An empirical investigation of trust requirements and guide to successful AI adoption. J Enterp Inform Manage 35(2):530\u0026ndash;549. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1108/JEIM-06-2020-0233\u003c/span\u003e\u003cspan address=\"10.1108/JEIM-06-2020-0233\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBrougham D, Haar J (2018) Smart Technology, Artificial Intelligence, Robotics, and Algorithms (STARA): Employees\u0026rsquo; perceptions of our future workplace. J Manage Organ 24(2):239\u0026ndash;257. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1017/jmo.2016.55\u003c/span\u003e\u003cspan address=\"10.1017/jmo.2016.55\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBologheanu R, Kapral L, Laxar D, Maleczek M, Dibiasi C, Zeiner S, Agibetov A et al (2023) Development of a reinforcement learning algorithm to optimize corticosteroid therapy in critically ill patients with sepsis. J Clin Med 12:1513. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/jcm12041513\u003c/span\u003e\u003cspan address=\"10.3390/jcm12041513\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBuccella A (2023) AI for all\u0026rsquo; is a matter of social justice. AI Ethics 3(4):1143\u0026ndash;1152. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s43681-022-00222-z\u003c/span\u003e\u003cspan address=\"10.1007/s43681-022-00222-z\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChang H (2023) Analysis of sentiment towards artificial intelligent industry using hybrid natural language processing technique. In 2023 IEEE 8th International Conference On Software Engineering and Computer Systems (ICSECS) (pp. 363\u0026ndash;370). IEEE. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/ICSECS58457.2023.10256397\u003c/span\u003e\u003cspan address=\"10.1109/ICSECS58457.2023.10256397\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen K, Cate A, Cheren H (2023) Communicating agriculture AI technologies: How American agricultural producers\u0026rsquo; perception of trustworthiness, risk perception, and emotion affect their likelihood of adopting artificial intelligence in food systems. Environ Communication 17(8):1004\u0026ndash;1019. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/17524032.2023.2211746\u003c/span\u003e\u003cspan address=\"10.1080/17524032.2023.2211746\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChoudhury A (2022) Factors influencing clinicians\u0026rsquo; willingness to use an AI-based clinical decision support system. Front Digit Health 4:920662. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3389/fdgth.2022.920662\u003c/span\u003e\u003cspan address=\"10.3389/fdgth.2022.920662\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChow JCL, Wong V, Li K (2024) Generative pre-trained transformer-empowered healthcare conversations: Current trends, challenges, and future directions in large language model-enabled medical chatbots. BioMedInformatics 4(1):837\u0026ndash;852. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/biomedinformatics4010047\u003c/span\u003e\u003cspan address=\"10.3390/biomedinformatics4010047\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDavis FD (1989) Perceived Usefulness, Perceived Ease of Use, and User Acceptance of Information Technology. MIS Q 13(3):319\u0026ndash;340. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2307/249008\u003c/span\u003e\u003cspan address=\"10.2307/249008\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDemirel S, Kahraman-Gokalp E, G\u0026uuml;nd\u0026uuml;z U (2024) From Optimism to Concern: Unveiling Sentiments and Perceptions Surrounding ChatGPT on Twitter. Int J Human\u0026ndash;Computer Interact 1\u0026ndash;23. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/10447318.2024.2392964\u003c/span\u003e\u003cspan address=\"10.1080/10447318.2024.2392964\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eFerrer X, Nuenen Tv, Such JM, Cot\u0026eacute; M, Criado N (2021) Bias and discrimination in AI: A cross-disciplinary perspective. IEEE Technol Soc Mag 40(2):72\u0026ndash;80. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/MTS.2021.3056293\u003c/span\u003e\u003cspan address=\"10.1109/MTS.2021.3056293\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGao S, He L, Chen Y, Li D, Lai K (2020) Public perception of artificial intelligence in medical care: Content analysis of social media. J Med Internet Res 22(7):e16649. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2196/16649\u003c/span\u003e\u003cspan address=\"10.2196/16649\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGerlich M (2023) Perceptions and acceptance of artificial intelligence: A multi-dimensional study. Social Sci 12(9):502. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/socsci12090502\u003c/span\u003e\u003cspan address=\"10.3390/socsci12090502\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHenrik Skaug S\u0026aelig;tra (2023) Generative AI: Here to stay, but for good? Technology in Society, 75, 102372. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.techsoc.2023.102372\u003c/span\u003e\u003cspan address=\"10.1016/j.techsoc.2023.102372\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHofstede G (1980) Culture's Consequences: International Differences in Work-Related Values. Sage\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHongladarom S (2021) The Thailand national AI ethics guideline: an analysis. J Inform Communication Ethics Soc 19(4):480\u0026ndash;491\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHongladarom S, Bandasak J (2024) Non-western AI ethics guidelines: implications for intercultural ethics of technology. AI Soc 39:2019\u0026ndash;2032. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00146-023-01665-6\u003c/span\u003e\u003cspan address=\"10.1007/s00146-023-01665-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHuang J (2024) Spatial_DataSet_for_Global public risk perception of artificial intelligence. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.6084/m9.figshare.25305244.v2\u003c/span\u003e\u003cspan address=\"10.6084/m9.figshare.25305244.v2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. figshare. Dataset\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHuang M-H, Rust R, Maksimovic V (2019) The Feeling Economy: Managing in the next generation of artificial intelligence (AI). Calif Manag Rev 61(4):43\u0026ndash;65. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/0008125619863436\u003c/span\u003e\u003cspan address=\"10.1177/0008125619863436\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHuo W, Zhang Z, Qu J, Yan J, Yan S, Yan J, Shi B (2024) Speciesism and preference of human\u0026ndash;artificial intelligence interaction: A study on medical artificial intelligence. Int J Human\u0026ndash;Computer Interact 40(11):2925\u0026ndash;2937. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/10447318.2023.2176985\u003c/span\u003e\u003cspan address=\"10.1080/10447318.2023.2176985\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKerstan S, Bienefeld N, Grote G (2024) Choosing human over AI doctors? How comparative trust associations and knowledge relate to risk and benefit perceptions of AI in healthcare. Risk Anal 44:939\u0026ndash;957. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/risa.14216\u003c/span\u003e\u003cspan address=\"10.1111/risa.14216\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eKrieger JB, Bouder F, Wibral M, Almeida RJ (2024) A systematic literature review on risk perception of artificial narrow intelligence. J Risk Res June 1\u0026ndash;19. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/13669877.2024.2350725\u003c/span\u003e\u003cspan address=\"10.1080/13669877.2024.2350725\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLiu H, Wang Y, Fan W, Liu X, Li Y, Jain S, Tang J (2022) Trustworthy ai: A computational perspective. ACM Trans Intell Syst Technol 14(1):1\u0026ndash;59. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1145/3546872\u003c/span\u003e\u003cspan address=\"10.1145/3546872\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLiu Y, Lyu Z (2024) Changes in public perception of ChatGPT: A text mining perspective based on social media. Int J Human\u0026ndash;Computer Interact 1\u0026ndash;15. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/10447318.2024.2406966\u003c/span\u003e\u003cspan address=\"10.1080/10447318.2024.2406966\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMcDuff D, Cheng R, Kapoor A (2018) Identifying bias in AI using simulation. arXiv preprint arXiv:1810.00471.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMinh D, Wang HX, Li YF et al (2022) Explainable artificial intelligence: A comprehensive review. Artif Intell Rev 55:3503\u0026ndash;3568. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10462-021-10088-y\u003c/span\u003e\u003cspan address=\"10.1007/s10462-021-10088-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNational New Generation Artificial Intelligence Governance Expert Committee (2019), June 17 Translation: Chinese Expert Group Offers \u0026lsquo;Governance Principles\u0026rsquo; for \u0026lsquo;Responsible AI\u0026rsquo;. Digichina. Stanford University. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://digichina.stanford.edu/work/translation-chinese-expert-group-offers-governance-principles-for-responsible-ai/\u003c/span\u003e\u003cspan address=\"https://digichina.stanford.edu/work/translation-chinese-expert-group-offers-governance-principles-for-responsible-ai/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eNeyazi TA, Ng SWT, Hobbs M, Yue A (2023) Understanding user interactions and perceptions of AI risk in Singapore. Big Data Soc 10(2). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/20539517231213823\u003c/span\u003e\u003cspan address=\"10.1177/20539517231213823\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eO'Reilly - Shah VN, Gentry KR, Walters AM, Zivot J, Anderson CT, Tighe PJ (2020) Bias and ethical considerations in machine learning and the automation of perioperative risk assessment. Br J Anaesth 125(6):843\u0026ndash;846. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.bja.2020.07.040\u003c/span\u003e\u003cspan address=\"10.1016/j.bja.2020.07.040\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eO\u0026rsquo;Regan J, Ferri G (2025) Artificial intelligence and depth ontology: implications for intercultural ethics. Appl Linguistics Rev 16(2):797\u0026ndash;807. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1515/applirev-2024-0189\u003c/span\u003e\u003cspan address=\"10.1515/applirev-2024-0189\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePat\u0026eacute; - Cornell E (2024) Preferences in AI algorithms: The need for relevant risk attitudes in automated decisions under uncertainties. Risk Anal 44:2317\u0026ndash;2323. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/risa.14268\u003c/span\u003e\u003cspan address=\"10.1111/risa.14268\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSchwesig R, Brich I, Buder J, Huff M, Said N (2023) Using artificial intelligence (AI)? Risk and opportunity perception of AI predict people\u0026rsquo;s willingness to use AI. J Risk Res 26(10):1053\u0026ndash;1084. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/13669877.2023.2249927\u003c/span\u003e\u003cspan address=\"10.1080/13669877.2023.2249927\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRainie L, Funk C, Anderson M, Tyson A (2022) AI and Human Enhancement: American\u0026rsquo;s Openness Is Tempered by a Range of Concerns: Public views are tied to how these technologies would be used, what constraints would be in place. Pew Res Cent. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttp://www.jstor.org/stable/resrep57310\u003c/span\u003e\u003cspan address=\"http://www.jstor.org/stable/resrep57310\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRobleto E, Habashi A, Kaplan MAB, Riley RL, Zhang C, Bianchi L, Shehadeh LA (2024) Medical students\u0026rsquo; perceptions of an artificial intelligence (AI) assisted diagnosing program. Med Teach 46(9):1180\u0026ndash;1186. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/0142159X.2024.2305369\u003c/span\u003e\u003cspan address=\"10.1080/0142159X.2024.2305369\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSartori L, Bocca G (2023) Minding the gap(s): Public perceptions of AI and socio-technical imaginaries. AI Soc 38:443\u0026ndash;458. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00146-022-01422-1\u003c/span\u003e\u003cspan address=\"10.1007/s00146-022-01422-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSchepman A, Rodway P (2022) The General Attitudes towards Artificial Intelligence Scale (GAAIS): Confirmatory validation and associations with personality, corporate distrust, and general trust. Int J Human\u0026ndash;Computer Interact 39(13):2724\u0026ndash;2741. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/10447318.2022.2085400\u003c/span\u003e\u003cspan address=\"10.1080/10447318.2022.2085400\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSiegrist M (2021) Trust and risk perception: A critical review of the literature. Risk Anal 41:480\u0026ndash;490. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/risa.13325\u003c/span\u003e\u003cspan address=\"10.1111/risa.13325\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWissing BG, Reinhard MA (2018) Individual differences in risk perception of artificial intelligence. Swiss J Psychol 77(4):149. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1024/1421-0185/A000214\u003c/span\u003e\u003cspan address=\"10.1024/1421-0185/A000214\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYigitcanlar T, Degirmenci K, Inkinen T (2024) Drivers behind the public perception of artificial intelligence: Insights from major Australian cities. AI Soc 39:833\u0026ndash;853. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00146-022-01566-0\u003c/span\u003e\u003cspan address=\"10.1007/s00146-022-01566-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhu Y, He G (2021) Can trust reduce the publics risk perception of artificial intelligence technology? Stud Sci Sci 10:1748\u0026ndash;1757. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.16192/j.cnki.1003-2053.20210425.001\u003c/span\u003e\u003cspan address=\"10.16192/j.cnki.1003-2053.20210425.001\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang B, Dafoe A (2019) Artificial intelligence: American attitudes and trends. SSRN Electron J. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2139/ssrn.3312874\u003c/span\u003e\u003cspan address=\"10.2139/ssrn.3312874\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang W, Sun L, Wang X, Wu A (2022) The influence of AI word - of - mouth system on consumers' purchase behaviour: The mediating effect of risk perception. Syst Res Behav Sci 39(3):516\u0026ndash;530. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/sres.2871\u003c/span\u003e\u003cspan address=\"10.1002/sres.2871\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZuiderwijk A, Chen YC, Salem F (2021) Implications of the use of artificial intelligence in public governance: A systematic literature review and a research agenda. Government Inform Q 38(3):101577. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.giq.2021.101577\u003c/span\u003e\u003cspan address=\"10.1016/j.giq.2021.101577\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Artificial Intelligence, Risk Perceptions, Global Patterns, Intercultural Determinants, Algorithmic Bias, Evolutionary Trends","lastPublishedDoi":"10.21203/rs.3.rs-6468789/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6468789/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"While artificial intelligence (AI) has become a transformative force across societies in the 21st century, its rapid advancement has sparked significant public discourse about potential risks and societal implications. This study investigates global patterns in public risk perceptions of artificial intelligence, examining three critical dimensions: spatial distribution characteristics, intercultural determinants, and temporal evolutionary trends. Through an integrated methodological framework combining natural language processing and convex hull analysis of 4.3 million social media posts from 2010-2023, we identify three dominant risk clusters: algorithmic bias in decision systems, data privacy violations , and labor market disruptions. The findings reveal significant geographic disparities, with privacy concerns increasing 23.7% annually in developed nations since 2018. Five key determinants emerge: national income levels, infrastructure completeness, health security index, regulatory quality, and electoral democracy index. This research offers valuable insights for evidence-based policymaking and contributes to the growing discourse on responsible AI development, ultimately supporting the sustainable integration of AI technologies into society.","manuscriptTitle":"Decoding Global AI Risk Perception Evolution: Intercultural drivers and Public Discourse Patterns in Algorithmic Societies","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-07-25 17:59:24","doi":"10.21203/rs.3.rs-6468789/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"d62dca2f-c2ff-47ba-9d07-dce9ca512c64","owner":[],"postedDate":"July 25th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":51975981,"name":"Social science/Cultural and media studies"},{"id":51975982,"name":"Social science/Science technology and society"}],"tags":[],"updatedAt":"2026-04-28T20:24:00+00:00","versionOfRecord":[],"versionCreatedAt":"2025-07-25 17:59:24","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6468789","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6468789","identity":"rs-6468789","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}