The relationship between artificial intelligence and civil servants’ work performance: a chain mediation model of self-efficacy and innovation ability

The relationship between artificial intelligence and civil servants’ work performance: a chain mediation model of self-efficacy and innovation ability | 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 Research Article The relationship between artificial intelligence and civil servants’ work performance: a chain mediation model of self-efficacy and innovation ability Zongyang Li, Dongming Gu This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-6526907/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 9 You are reading this latest preprint version Abstract Background: As direct providers of public services and key agents of social governance, civil servants have experienced significant changes in their roles and work behaviors under the wave of digital transformation in the public sector, driven by modern information technologies such as artificial intelligence(AI). However, there remain few studies exploring how AI has influenced civil servants’ work performance and through what mechanisms these effects occur. Therefore, this study investigates the relationship between AI application and civil servants’ work performance, focusing on the mediating roles of self-efficacy and innovation ability, as well as the moderating role of willingness to change. Methods: A quantitative research design was adopted in this study. Data were collected through an online questionnaire from 296 civil servants. The empirical analysis employs a structural equation model, chain mediation model, moderation model, and bootstrap method, with all statistical procedures conducted using Stata 16. Results: The findings confirmed that: AI application was positively correlated with civil servants’ work performance; self-efficacy and innovation ability mediated the relationship between AI application and work performance; A chain-mediating effect through self-efficacy and innovation ability was observed in this relationship; Willingness to change exerted a moderating effect on the relationships between AI application and innovation ability, as well as between self-efficacy and innovation ability. Conclusion: This study offers a novel perspective for understanding the behavioral motivations and underlying logic of civil servants amid the digital transformation in the public sector, by emphasizing the impact of AI application, the mediating effects of self-efficacy and innovation ability, and the moderating role of willingness to change. These findings provide beneficial insights for civil servants and policymakers, underscoring the importance of prioritizing and accelerating the learning of AI and the enforcement of AI-related competencies. Further, the study offers practical recommendations from the perspectives of fostering an innovation-oriented institutional environment and stimulating individual motivation for technology adoption. AI application self-efficacy innovation ability work performance willingness to change civil servant Figures Figure 1 Figure 2 Figure 3 Figure 4 1 Introduction Artificial intelligence (AI) refers to a set of technologies and methodologies that enable computer systems to simulate human cognitive processes such as perception, comprehension, reasoning, learning, and decision-making (Jackson, 2019 ). In recent years, with continuous advancements and maturation of AI technologies, AI has been widely applied across various domains—including healthcare (Beam et al., 2023), traffic management (Caianiello, 2021 ), education (Hwang & Wu, 2025 ), urban governance (Lartey & Law, 2025 ), and social security systems (Lexer & Scarcella, 2019 )—and has profoundly reshaped service delivery models and governance paradigms within the public sector (Wirtz et al., 2019 ). Since the emergence of the New Public Management movement in the 1980s, public sector organizations have gradually undergone a transformation toward service-oriented and efficiency-driven structures (Hood, 1995 ). Improving the quality of public services and enhancing administrative efficiency have become key objectives. In the 21st century, the rapid advancement of digital technologies—including the internet, big data, and AI—has served as a critical driver for digital transformation in the public sector, representing a pivotal strategy for achieving the effectiveness and responsiveness of public services (Maxwell et al., 2019 ). From a macro perspective, the digitization of the public sector refers to the deep integration of digital technologies into public sector governance, led by governments, encompassing transformations in governance concepts, organizational structures, and institutional mechanisms. This process drove initiatives such as “digital government” (Janowski, 2015). Specifically, it involves efforts to strengthen the integration of administrative information systems and the sharing of data resources, ultimately advancing the intelligence and responsiveness of governance systems. From a micro perspective, public sector digitization is reflected in the use of digital tools—such as data analytics, intelligent decision-support systems, and online government service platforms—by civil servants in daily administrative practices. These tools help streamline workflows, improve service efficiency, and strengthen government responsiveness to public demands, thereby facilitating the precise and personalized provision of public services (Bilan et al., 2023 ; Seckelmann & Catakli, 2025 ). At present, AI is increasingly becoming a key driver in reshaping public service delivery models and reconstructing policy implementation mechanisms. As the primary providers of public services and executors of governance, civil servants are also undergoing significant changes in their role positioning and competency structures (Mehr et al., 2017 ). Moreover, while AI contributes to optimizing governance processes and enhancing administrative efficiency, the deep integration of emerging technologies such as AI has simultaneously transformed the logic of government service provision. This shift compels civil servants to evolve from traditional task-oriented executors into compound governance professionals equipped with technological literacy and data competence (Lexer & Scarcella, 2019 ). However, AI also imposes institutional transformations and professional disruptions upon the civil service, which are particularly salient amid China’s ongoing digital transformation in the public sector. In 2025, the breakout success of DeepSeek marked a new stage of maturity for China's large language model technologies. Building on these advancements, Shenzhen Municipality in Guangdong Province launched a pilot program featuring “AI-powered digital civil servants” within its administrative system, aiming to automate routine tasks such as administrative consultation, information entry, and process approvals, thereby improving operational efficiency in government services. Against this backdrop of accelerated governmental digital transformation, the traditional professional identity and skill structure of civil servants are facing mounting challenges brought about by technological change. In the context of the widespread application of AI in governmental systems, shifts in identity and behavioral choices among civil servants have become a critical component of the government’s digital transformation. Research in environmental psychology suggests that individual behavior is not only influenced by objective environmental changes but also shaped by subjective perceptions of those changes. When significant technological transformations occur, especially those that impact occupational stability and role expectations, individuals may experience psychological stress responses and engage in adaptive behaviors. As a structurally stable force within the national governance system, civil servants’ engagement with AI—reflected in their acceptance, risk perception, and behavioral responses—is crucial not only for their professional adaptation but also for the sustainability of digital governance and the effective implementation of AI-driven administration. In recent years, scholarly interest in how AI influences the civil service has deepened, with increasing attention paid to its role in enhancing administrative efficiency and reshaping competency structures. Some studies affirm the positive contributions of AI in improving government operations. In studies on AI practices in the United States, it has been found that AI enables public authorities to analyze public sentiment through data analytics, thereby allowing governments to more precisely identify societal needs and deliver targeted public services, enhancing citizen satisfaction (Mehr et al., 2017 ). With the ongoing evolution of generative AI technologies, intelligent systems are increasingly capable of assisting civil servants with routine administrative tasks such as document drafting and information archiving, thereby freeing them from repetitive labor and allowing them to focus on more strategic and creative aspects of governance (Caianiello, 2021 ). In addition, AI supports public decision-making by offering simulations and probabilistic estimations that help officials handle complex public issues more effectively. Based on these perspectives, scholars argue that the digital transformation driven by AI substantially enhances civil servants' work efficiency. However, a contrasting body of literature expresses concern. Some researchers argue that deep integration of AI may lead to over-reliance on technology, thereby weakening civil servants' independent judgment and problem-solving capacities (Ivić et al., 2022 ). Furthermore, the hyper-rationality introduced by AI systems may conflict with ethical and moral dimensions inherent in public governance, potentially leading to more complex governance challenges. When facing diverse and context-sensitive governance issues, standardized technical solutions are insufficient to replace human judgment, contextual awareness, and adaptive decision-making, potentially undermining civil servants’ creativity and adaptability (Abioye et al., 2021 ). Additionally, the technological shift caused by AI threatens to displace repetitive, basic, and routine roles within the public sector, heightening concerns over job security and professional anxiety among civil servants (Sari, 2025 ). In sum, despite ongoing debates regarding the impact of AI on the civil service, it is undeniable that the emergence of AI has introduced profound changes to both public sector operations and the work performance of civil servants. Accordingly, academic inquiry has increasingly focused on how and through which mechanisms AI affects civil servants’ work performance. Among the various explanatory factors, self-efficacy and innovative ability have received widespread attention due to their critical role in accounting for individual performance differences. Some scholars argued that the promotion of AI competencies can enhance civil servants’ self-efficacy, thereby enabling them to approach their tasks with greater confidence and competence (Ramadian et al., 2025 ). Meanwhile, innovation capability, as a key dimension of individual resource integration and problem-solving, plays a vital role in advancing organizational performance. Meanwhile, innovation ability, as a key dimension of individual resource integration and problem-solving, is also a crucial factor in enhancing organizational performance. Civil servants’ proficiency in AI technologies is positively associated with their innovative capabilities, and those with stronger innovation ability tend to perform better in terms of task execution and responsiveness to public needs. Taking into account the intricate interconnections among these variables and the identified gaps in existing literature, this study aims to reveal the complex connections among AI application, self-efficacy, innovation ability, work performance, and willingness to change. To be more exact, this study sought to explore how AI application affect civil servants’ work performance, the role self-efficacy plays in this process, the extent to which innovation ability mediates this relationship, and how willingness to change contributes to the enhancement of innovation ability through AI application and self-efficacy. To achieve this goal, this study collected 296 completed questionnaires from MPA(Master of Public Administration) students who are employed in various public sector organizations. Based on this, a series of empirical analyses were conducted, including a structural equation model (SEM), chain mediation model, moderation model, and bootstrap method. These analyses were designed to identify the logical relationships and influence mechanisms among the five core latent variables. The research results of this study are beneficial for examining the chain-mediating pathway and the moderating role of willingness to change, which provide theoretical insights and empirical evidence for empowering AI technologies to enhance the work performance of civil servants. 2 Research hypotheses 2.1 AI application and work performance According to the Technology Acceptance Model (TAM), individuals’ technology usage behavior is primarily influenced by their Perceived Usefulness and Perceived Ease of Use of the technology, both of which subsequently exert a direct influence on performance outcomes (Marangunić & Granić, 2015 ). In the context of accelerating digital transformation within the public sector, artificial intelligence technologies—particularly generative AI(AIGC)—have been widely adopted in routine tasks such as document drafting, information retrieval, data processing, and decision support. These technologies provide important and substantial assistance to civil servants by improving task completion speed, reducing administrative burden, and enhancing information processing efficiency (Casalino et al., 2020a , b ). Existing studies have shown that effective use of AI can optimize task workflows, facilitate knowledge acquisition, and improve information management, thereby contributing to both individual productivity and overall organizational performance (Rusho & Chan, 2023 ). In China’s digital government initiatives, civil servants are often required to process large volumes of structured and unstructured data. The application of AI technologies helps alleviate these cognitive and labor-intensive demands, thereby enhancing their capacity to handle complex tasks (Shahzad et al., 2023 ). In summary, the application of AI by civil servants can significantly enhance their capabilities in information acquisition, document processing, and assisted decision-making. This, in turn, reduces routine workload and contributes to higher execution efficiency and job performance in complex administrative environments. Therefore, this paper proposes the following hypothesis: Hypothesis H1: AI application is positively associated with work performance. 2.2 The mediating role of self-efficacy Self-efficacy refers to individual subjective judgment regarding their ability to accomplish specific tasks, directly influencing behavioral motivation, task engagement, and performance levels (Bandura, 1982 ). The TAM also emphasizes that technology use is not solely determined by the attributes of the technology itself, but also affects work performance by shaping individuals’ cognitive and psychological states (Chen et al., 2011 ). In the context of AI integration into public governance practices, the application of AI by civil servants contributes to their perception of task feasibility and convenience, which in turn enhances their confidence in their abilities—that is, their self-efficacy. For instance, when civil servants use generative AI to assist in drafting official documents, retrieving information, or processing data, they may experience a tangible sense of improved efficiency. This positive experiential feedback enhances their perception of their work capabilities, which in turn fosters greater motivation and initiative in task execution. Moreover, existing research has confirmed that self-efficacy significantly enhances work performance by strengthening an individual’s goal orientation and increasing their willingness to undertake challenging tasks, which in turn improves the quality and efficiency of task completion (Bandura, 1982 ). In addition, self-efficacy has been found to enhance psychological resilience in the face of pressure or obstacles, reduce feelings of burnout, and help maintain high levels of performance (Shoji et al., 2016 ). Taken together, technology use facilitates the development and maintenance of self-efficacy, which itself is a critical psychological factor influencing individual work performance (Shank & Cotten, 2014 ). Therefore, self-efficacy serves as a psychological mechanism that links AI application with work performance. Based on this, the following hypotheses are proposed: Hypothesis H2a: The wide application of AI would increase civil servants’ self-efficacy. Hypothesis H3: The increase in self-efficacy due to the use of AI would mediate the improvement of work performance. 2.3 The mediating role of innovation ability Innovation ability refers to the individual capacity to identify problems, generate new ideas, integrate resources, and develop solutions within specific contexts. It is a critical factor influencing one’s creative contributions and performance within an organization (Prajogo & Ahmed, 2006 ). In the era of rapidly evolving information technologies, artificial intelligence offers civil servants novel tools for knowledge acquisition, task generation, and cognitive expansion. Its capabilities in multimodal content creation, language structuring, and data analysis provide algorithmic support for stimulating innovative thinking. The application of AI disrupts traditional path-dependent approaches to task execution by offering more flexible and efficient methods of working (Zhao & Cao, 2024 ). This encourages civil servants to adopt more creative and proactive approaches to problem identification, strategic decision-making, and resource allocation. Studies have shown that technology adoption can enhance innovation awareness and behavior by improving information processing capabilities and reducing the cost of trial-and-error in innovation (Benson, 2019 ; Lemke, 2010 ). Meanwhile, innovation ability—considered a high-level cognitive and practical competence—influences work performance not only by transforming ways of thinking but also through direct contributions to task effectiveness. Civil servants with strong innovation capabilities are better equipped to develop novel solutions and overcome existing limitations when dealing with complex or unstructured problems, thus improving task quality and organizational adaptability (Laviolette et al., 2016 ). Besides, innovation ability also contributes to optimizing service processes and public resource allocation, ultimately enhancing governance efficiency and citizen satisfaction. Empirical research has also confirmed a significant positive correlation between innovation capacity and both individual and organizational performance (Prajogo & Ahmed, 2006 ). Accordingly, innovation ability may serve as both a cognitive and skill-based outcome of AI utilization, as well as a key mediating variable that links AI use with improved work performance among civil servants. Based on this, the following hypotheses are proposed: Hypothesis H2b: The wide application of AI would increase civil servants’ innovation ability. Hypothesis H4: The increase in innovation ability due to the use of AI would mediate the improvement of work performance. 2.4 The chain-mediating role of self-efficacy and innovation ability Beyond independent mediating mechanisms, the application of AI may also exert an indirect impact on civil servants’ work performance through a sequential mediation pathway involving self-efficacy and innovation ability. According to self-efficacy theory, individuals’ beliefs in their capabilities to accomplish specific tasks significantly enhance their initiative and willingness to explore when confronted with challenging situations (Marangunić & Granić, 2015 ; Zhao et al., 2025 ). They tend to exhibit stronger capabilities in problem identification, strategy generation, and resource integration, thereby exhibiting a higher level of innovation ability (Bandura, 2013 ). Furthermore, the TAM suggests that perceived usefulness and perceived ease of use not only shape individual willingness to adopt technology but also indirectly affect performance by influencing cognitive evaluations and psychological responses (Davis et al., 2024 ). Within an administrative environment where AI-assisted work is increasingly widespread, civil servants who receive positive feedback during AI use—such as improved task efficiency and more effective information processing—are likely to experience enhanced self-efficacy. This strengthened self-efficacy, in turn, improves their ability to adapt to technological change and motivates them to proactively apply AI tools in problem-solving, thereby enhancing their innovation ability. The enhancement of innovation ability, in return, lays the groundwork for improved service delivery and work performance. Therefore, the application of AI may influence work performance not only through isolated psychological mechanisms but also through a chain mediation pathway involving self-efficacy and innovation ability, reflecting a deeper and more complex route of indirect influence. Based on this theoretical reasoning, the following hypothesis is proposed: Hypothesis H5: self-efficacy and innovation ability play a chain-mediating role in the relationship between AI application and work performance. 2.5 The moderating role of willingness to change In technology-driven organizational change, individuals’ attitudes toward change and their willingness to accept it are critical factors influencing the effectiveness of technology implementation and the development of individual innovation capabilities (Grant et al., 2020 ; Jiao & Zhao, 2014 ). Willingness to change refers to an individual proactive disposition and behavioral tendency in response to environmental shifts and organizational technological reform. It is a key psychological precondition for converting personal resources into adaptive actions (Ballet & Kelchtermans, 2008 ). Existing studies suggest that, beyond individual motivations, organizational factors—such as the climate for change, leadership orientation, and institutional support—also play a crucial role in shaping employees’ change-related behaviors (Herold et al., 2008 ). When organizational culture actively encourages technology adoption and institutional innovation, or when supervisors demonstrate clear support for change, individuals are more likely to perceive technological reform as legitimate and necessary. This recognition helps cultivate a positive psychological attitude toward change, thereby strengthening their willingness to embrace new technologies and adapt proactively. Thus, in an environment characterized by a strong change-oriented culture, civil servants with a high willingness to change are more inclined to perceive AI as an opportunity for innovation and process optimization. They are likely to engage in active learning and exploration during AI application, which in turn promotes the development of their innovation ability. In contrast, those with a low willingness to change may exhibit passive or resistant attitudes during AI implementation, lacking the motivation and intention to transform technological potential into innovative practice. Moreover, while self-efficacy enhances individuals’ confidence in task performance, whether this confidence translates into creative behavior often depends on the level of environmental support and the individual’s psychological readiness (Marangunić & Granić, 2015 ; Zhang, 2024 ). For civil servants with a high willingness to change, the motivation and confidence derived from self-efficacy are more likely to manifest as proactive change-seeking and innovation attempts, thereby strengthening their innovation capabilities. Conversely, in contexts where change motivation is weak or organizational support is lacking, even individuals with high self-efficacy may struggle to translate internal confidence into actual creative outcomes due to the absence of motivational triggers for action (Choi, 2011 ). Based on the above analysis, the following hypotheses are proposed: Hypothesis H6a: Willingness to change exerts a positive moderating influence on the relationship between AI application and innovation ability. Hypothesis H6b: Willingness to change exerts a positive moderating influence on the relationship between self-efficacy and innovation ability. The hypothesized research model, which visually represents these proposed relationships, is presented in Fig. 1. FIGURE 1 Research model map 3 Methodology 3.1 Participant and recruitment This study was approved by the Academic Committee of the South China University of Technology and was conducted in strict accordance with the guidelines of the Declaration of Helsinki, with all participants provided informed consent before their participation. A total of 327 MPA students were recruited through an online survey. These participants were employed in various public sector organizations, including government and public institutions. To ensure data quality, responses from participants who failed the attention check or displayed inconsistent answers were excluded. After data cleaning, 296 valid questionnaires were retained, resulting in an effective response rate of 90.52%. The final sample covered 134 prefecture-level cities across 29 provinces (including autonomous regions and municipalities) in China. Among the respondents, 46.6% were male and 53.4% were female. Participants' ages ranged from 22 to 64 years, with an average age of 32.2. Notably, the proportion of those aged 22 to 50 reached 94.6%. In terms of educational background, respondents ranged from high school to Ph.D, with the majority holding a bachelor’s degree. 70.9% of the participants were members of the Communist Party of China. The administrative level of the respondents ranged from staff-level(gu-ji) to department-level(chu-ji), with 98.6% being staff-level(gu-ji) and section-level (ke-ji) officers, indicating that the vast majority of respondents are grass-root civil servants. Monthly income levels varied from 1,800 yuan to 20,000 yuan, with an average monthly salary of 7,452 yuan. 3.2 Measures AI application: The scale used to measure AI application was adapted from recent research (Obenza et al., 2024 ; Wirtz et al., 2019 ), with appropriate revisions made to fit the context of this study. The scale consists of six items, including sample statements such as “I am willing to discuss recent developments and practical applications of AI with people around me” and “AI has changed the way I live and work.” In the current study, Cronbach’s alpha for scores of AI application was 0.6719. Self-efficacy: The measurement of self-efficacy among civil servants was adapted from the existing study (Xu & Xue, 2025 ). The scale includes three items, such as “I believe I am capable of performing my job well.” In the current study, Cronbach’s alpha for scores of self-efficacy was 0.7298. Innovation ability: Based on existing studies (Doroodian et al., 2014 ). The scale consists of four items, including statements such as “My writing ability for text-based tasks (official documents, reports, proposals) has improved” and “I have become more creative and have more innovative ideas.” In the current study, Cronbach’s alpha for scores of innovation ability was 0.8510. Work performance: The measurement of work performance was adapted from (Iaffaldano & Muchinsky, 1985 ), with item modifications made to reflect the specific context of this study. The scale comprises five items, including statements such as “The average time I need to complete the same tasks has significantly decreased” and “The quality of output in my work has noticeably improved.” In the current study, Cronbach’s alpha for scores of work performance was 0.7877. Willingness to change: To measure this construct, we adapted items from some research (Metselaar, 1997 ; Miller et al., 1994 ). The scale includes four items, such as “While completing my tasks with high quality, I also strive for innovation and breakthroughs” and “My department encourages innovation and change in daily work.” In the current study, Cronbach’s alpha for scores of willingness to change was 0.7528. 3.3 Statistical processing To achieve robust evidence, a chain mediation model test was necessary. Thus, SEM analysis and statistical tests were conducted utilizing STATA 16. We subsequently adopted a mediation testing procedure to evaluate the mediation effects of self-efficacy and innovation ability (Baron & Kenny, 1986 ). Further, we construct a chain mediation model to examine the chain-mediating role of self-efficacy and innovation ability. Additionally, to enhance the robustness of our mediation analysis, we applied the bootstrapping method (Bolin, 2014 ). Finally, we conducted moderation effect tests for willingness to change through interaction item construction and simple slope analysis. 4 Results 4.1 Common method variance test The five key core latent variables used in this paper, which are AI application, self-efficacy, innovation ability, work performance, and willingness to change, come from the same source. To prevent the potential interference of common method variance on the reliability of our findings, Harman’s single factor test was used in this study to test the common method variance. The results shown in Table 1 revealed relatively low factor loadings of a single factor on these variables (mostly ranging between 0.3 and 0.6) accompanied by generally high uniqueness (mostly above 0.7), indicating that a single factor does not dominate the variance explanation for these items. Therefore, it can be concluded that no significant common method variance exists, ensuring a robust foundation in reliability and validity for subsequent structural equation modeling analysis. 4.2 Confirmatory factor analysis(CFA) To further verify the discriminant validity among variables, we performed a CFA using Stata 16.0 software. There are five core latent variables concerned in this study: AI application, self-efficacy, innovation ability, willingness to change, and work performance. The analytical results are shown in Table 2 . According to the results of data analysis, the single-factor model showed an inferior fit ( \(\:{}^{2}\) /df = 3.155 > 3, CFI = 0.792 0.08) compared to the five-factor model ( \(\:{}^{2}\) /df = 1.896 0.8, RMSEA = 0.055 < 0.08). To further prevent the single-factor comparison bias, we constructed three-factor ( \(\:{}^{2}\) /df = 2.557, CFI = 0.852, RMSEA = 0.073) and four-factor models ( \(\:{}^{2}\) /df = 2.171, CFI = 0.890, RMSEA = 0.063; \(\:{}^{2}\) /df = 2.361, CFI = 0.872, RMSEA = 0.068) sequentially, which did not significantly enhance fit indices. These analytical outcomes demonstrate satisfactory discriminant validity among variables, justifying further analyses. Table 1 Common method deviation test(N = 296) Bernbach Variable Factor1 Uniqueness Cronbach Variable Factor1 Uniqueness α = 0.6719 AIP1 0.4555 0.7925 α = 0.7298 WE1 0.3255 0.8941 AIP2 0.2774 0.9231 WE2 0.4417 0.8049 AIP3 0.4213 0.8225 WE3 0.5185 0.7311 AIP4 0.5103 0.7396 α = 0.7528 IR1 0.6233 0.6115 AIP5 0.6427 0.5870 IR2 0.5152 0.7346 AIP6 0.3711 0.8623 IR3 0.5427 0.7055 α = 0.7877 WP1 0.5031 0.7469 IR4 0.6061 0.6326 WP2 0.5094 0.7405 α = 0.8510 IA1 0.5392 0.7092 WP3 0.6698 0.5514 IA2 0.7395 0.4532 WP4 0.5630 0.6831 IA3 0.7184 0.4840 WP5 0.6334 0.5988 IA4 0.7350 0.4598 Table 2 Confirmatory factor analysis(N = 296) Model Model Description \(\:{}^{2}\) df \(\:{\varDelta\:}^{2}(\varDelta\:\varvec{d}\varvec{f})\) CFI TLI RMSEA SRMR 1 Five-factor Model 377.222 199 0.918 0.905 0.055 0.050 2 Four-factor Model 440.797 203 63.575 (4)*** 0.890 0.875 0.063 0.062 3 Four-factor Model 479.352 203 102.13 (4)*** 0.872 0.855 0.068 0.057 4 Three-factor Model 526.702 206 149.48 (7)*** 0.852 0.834 0.073 0.059 5 Single-factor Model 659.374 209 282.152 (10)*** 0.792 0.770 0.085 0.066 1. It is a hypothetical model. 2. Self-efficacy and innovation ability are combined as one factor. 3. Willingness to change and work performance are combined as one factor. 4. Self-efficacy, innovation ability, and willingness to change are combined as one factor. 5. All variables are combined as one factor. ∆χ2 test is relative to Model 6.*p < 0.05. **p < 0.01. ***p < 0.001. 4.3 Descriptive statistics and correlation analysis Means, standard deviations, and correlations between the research variables are presented in Table 3 . It can be seen from Table 3 that the means for AI application, self-efficacy, and innovation ability are 3.992, 4.279, and 3.814. These results indicate that civil servants show high levels of self-efficacy and creative cognition in the context of the wide application of AI technology. Further calculation based on item count shows approximate scoring rates of 79.84%, 85.59%, and 76.28% for these three primary latent variables, confirming most respondents’ acknowledgment of the positive impacts of AI applications on work performance. The correlation analysis further indicates significant positive correlations between AI application and self-efficacy (r = 0.383, p < 0.001), innovation ability (r = 0.566, p < 0.001), and work performance (r = 0.576, p < 0.001), preliminarily supporting Hypotheses H1, H2a, and H2b. Additionally, significant positive correlations were found between self-efficacy and innovation ability (r = 0.428, p < 0.001). Besides, these two variables strongly correlated with work performance (r = 0.399, p < 0.001 1 ; r = 0.590, p < 0.001 2 ), preliminarily supporting Hypotheses H3, H4, and H5. Furthermore, significant correlations were observed between willingness to change and AI application (r = 0.515, p < 0.001), self-efficacy (r = 0.495, p < 0.001), and innovation ability (r = 0.554, p < 0.001). These findings indicate the likelihood that willingness to change significantly influences various dimensions, including AI application, self-efficacy development, and individual innovation enhancement. Table 3 Descriptive statistics and correlation matrix of variables (N = 296) M SD 1 2 3 4 5 1. AI application 3.992 0.475 1.000 2. Self-efficacy 4.279 0.446 0.3837*** 1.000 3. Innovation ability 3.814 0.737 0.5659*** 0.4283*** 1.000 4. Willingness to change 3.922 0.651 0.5146*** 0.4954*** 0.5543*** 1.000 5. Work performance 4.039 0.540 0.5762*** 0.3988*** 0.5899*** 0.5138*** 1.000 1. M and SD are used to represent mean and standard deviation, respectively. 2. *p < 0.05. **p < 0.01. ***p < 0.001. (dup: 6 ?) 4.4 Chain Mediation Model Test Stata 16 was used in this study to verify Hypotheses H1, H2a, H2b, H3, and H4 by multiple linear regression analyses. As Table 4 and Fig. 2 illustrate, after incorporating all control variables, AI application showed a significant positive correlation with civil servants' work performance (Model 1, β = 0.6473, p < 0.001), supporting Hypothesis H1. This result indicates that, despite the learning cost and crowding-out effects of AI technology, its extensive application significantly enhances individual performance and output among civil servants. Then, following Baron and Kenny’s ( 1986 ) mediation test procedure, we evaluated the mediating roles of self-efficacy and Innovation Ability, the results of which are shown in Table 4 . After incorporating all control variables, we found AI application positively correlated with individual Self-efficacy (Model 2a, β = 0.3453, p < 0.001). With control applied to AI application, self-efficacy showed a significant positive correlation with work performance (Model 2b, β = 0.2550, p < 0.001), thus validating Hypotheses H2a and H3. Moreover, as Table 4 further indicates, AI application is positively associated with innovation ability (Model 3a, β = 0.8344, p < 0.001). With AI application controlled, innovation ability correlates positively with work performance (Model 3b, β = 0.2845, p < 0.001), thereby validating Hypotheses H2b and H4. This result also confirms the view of the Stimulus-Organism-Response, which suggests that external stimulus induces a behavioral response by influencing an individual’s internal mental state. AI application significantly enhance civil servants’ self-efficacy and innovation ability, which in turn enable them to perform more effectively in their work, thereby leading to better work performance. Subsequently, we tested the chain mediation model of AI application → self-efficacy → innovation ability → work performance, using Stata 16 as well. After controlling AI application, self-efficacy exhibited a significant association with innovation ability (Model 4a, β = 0.3991, p < 0.001). Furthermore, when all three core latent variables—AI application, self-efficacy, and innovation ability—were simultaneously included in the regression model, each demonstrated a significant positive relationship with work performance (Model 4b, β = 0.3812, p < 0.001; β = 0.1530, p = 0.011; β = 0.2556, p < 0.001), thereby validating Hypothesis H5. Finally, following the approach proposed by Hayes (2017), we conducted a multi-step mediation analysis using the bootstrap method. As shown in Table 5 and Fig. 2, the indirect effect of AI application on work performance through self-efficacy is 0.0881, and 95%CI = [0.03585, 0.14029], excluding 0. This result indicates that civil servants can significantly enhance their self-efficacy through extensive application of AI technologies, which in turn improves their work performance. Therefore, Hypothesis H3 is supported. Moreover, the indirect effect of AI application on work performance via innovation ability is 0.2374, and 95%CI = [0.15035, 0.32441], excluding zero. This result confirms a significant mediating role of innovation ability, thereby providing further support for Hypothesis H4. To be further, the indirect effect of AI application on work performance through self-efficacy and innovation ability is 0.0352, and 95%CI = [0.01186, 0.05858], excluding zero. This result confirms a significant chain mediation effect among the four core latent variables, offering robust support for Hypothesis H5. This finding demonstrates that the application of AI technologies significantly enhances civil servants’ self-efficacy, which in turn promotes their innovation ability, thereby ultimately improving their work performance. Table 4 Regression analysis of AI application, self-efficacy, innovation ability, and work performance(N = 296) Variables Regression coefficients Fits the index Model DV IV β t p R2 AR2 F Model 1 Work performance AI application 0.6473 11.67 < 0.001 0.3477 0.3295 19.12*** Model 2a Self-efficacy AI application 0.3453 6.66 < 0.001 0.1632 0.1399 7.00*** Model 2b Work performance AI application 0.5592 9.65 < 0.001 0.3848 0.3654 19.88*** Self-efficacy 0.2550 4.15 < 0.001 Model 3a Innovation ability AI application 0.8344 11.07 < 0.001 0.3525 0.3344 19.53*** Model 3b Work performance AI application 0.4099 6.70 < 0.001 0.4452 0.4277 25.50*** Innovation Ability 0.2845 7.09 < 0.001 Model 4a Innovation ability AI application 0.6966 8.93 < 0.001 0.4013 0.3825 21.30*** Self-efficacy 0.3991 4.83 < 0.001 Model 4b Work performance AI application 0.3812 6.18 < 0.001 0.4576 0.4385 24.04*** Self-efficacy 0.1530 2.55 0.011 Innovation Ability 0.2556 6.18 < 0.001 1. Control variables were included in all regression models. DV, Dependent Variable; IV, Independent Variable. 2. *p < 0.05. **p < 0.01. ***p < 0.001. (dup: 8 ?) Table 5 Mediation effect test(N = 296) Control variables were included in all regression models. 95% CI Effect β (standardized path coefficient) SE LL UL X→M1→Y 0.0881 0.0266 0.0359 0.1402 X→M2→Y 0.2374 0.0444 0.1504 0.3244 X→M1→M2→Y 0.0352 0.0119 0.0119 0.0586 4.5 Moderation Effect Test Table 6 delineates that the interplay between willingness to change and AI application robustly fosters innovation ability(Model 1, β = 0.316, p < 0.001). Referencing Fig. 3, the interaction effects and simple slopes suggest that at subdued levels of willingness to change, AI application imparts a significant and positive influence on innovation ability(simple slope = 1.111, p < 0.001). Conversely, at elevated willingness to change levels, this influence markedly intensifies (simple slope = 0.419, p = 0.012), corroborating that willingness to change acts as an effective moderator within the nexus of AI application and innovation ability, thereby validating Hypothesis H6a. Similarly, the results shown in Table 6 also demonstrate that the synergy between willingness to change and self-efficacy significantly propels innovation ability(Model 2, β = 0.269, p < 0.001). Furthermore, a simple slope analysis is conducted on the above moderating effects, and the results are shown in Fig. 4. Under a lower level of willingness to change, self-efficacy has a mild promoting effect on the improvement of innovation ability (simple slope = 0.420, p = 0.019). Whereas, with the increase of willingness to change, the promoting effect of self-efficacy on the innovation ability is significantly increased(simple slope = 0.899, p = 0.01), confirming that willingness to change constructively regulates the interrelation between self-efficacy and innovation ability, hence affirming Hypothesis H6b. Table 6 Moderation effects test(N = 296) Variables Regression coefficients Fits the index Model DV IV β t p R2 AR2 F Model 1 Innovation ability AI application 0.618 7.73 < 0.001 0.4344 0.4286 74.76*** Willingness to change 0.453 7.55 < 0.001 AI application ×willingness to change 0.316 3.22 < 0.001 Model 2 Innovation ability Self-efficacy 0.443 4.51 < 0.001 0.3538 0.3471 53.29*** Willingness to change 0.536 8.65 < 0.001 Self-efficacy ×willingness to change 0.269 2.62 < 0.001 1. Control variables were included in all regression models. 2. *p < 0.05. **p < 0.01. ***p < 0.001. 5 Conclusion and discussion 5.1 Study conclusion Our study provides valuable insights into AI application and its impact on civil servants’ work performance. Based on the structural equation model, chain mediation model, moderation model, and bootstrap method of questionnaire data from 296 civil servants, this study found strong evidence of AI’s positive influence on work performance. Firstly, AI application had a significant direct effect on work performance (β = 0.6473, p < 0.001). This result indicated that the extensive use of AI can effectively improve the ability of civil servants in information acquisition, document processing, and decision-making, thus improving their work performance. Secondly, self-efficacy plays a mediating role between AI application and work performance (β = 0.2550, p < 0.001), as well as innovation ability notably mediated the relationship between AI application and work performance (β = 0.2845, p < 0.001). This confirmed that self-efficacy and innovation ability are important transmission mechanisms of AI application, playing a notable role in bridging the gap between AI application and work performance. Thirdly, this study further found that self-efficacy and innovation ability play a chain mediating role between AI application and work performance of civil servants (β = 0.1530, p < 0.001; β = 0.2556, p < 0.001). This conclusion suggests that AI application elevate self-efficacy of civil servants, and then further improves their innovation ability, which in turn promotes their work performance. Fourthly, willingness to change emerged as a positive moderator, refining the relationship between the AI application and innovation ability (β = 0.316, p < 0.001), underscoring the paramount importance of acceptance willingness and action tendency of new technology in fostering effectiveness of AI application and improving innovation ability. Besides, willingness to change also amplified the promoting effect of self-efficacy on innovation ability (β = 0.269, p < 0.001), emphasizing the transformative power of change orientation on the relationship between psychological perception and innovative practice. 5.2 Theoretical and practical implications 5.2.1 Theoretical Based on the preceding theoretical framework and empirical analysis, this study examined the relationship between AI application and civil servants’ Work Performance. In particular, it explored the mediating roles of self-efficacy and innovation ability, as well as the moderating role of willingness to change. The main theoretical contributions of this study can be summarized as follows. First, this study focuses on the actual usage behavior of AI technologies among civil servants, introducing the TAM into the research context of the public sector, thereby extending the model’s scope of applicability. Compared with the original TAM proposed, which has been primarily applied to analyze technology adoption behaviors among corporate employees (Park et al., 2014 ) and educators (Teo, 2009 ), this study applied the TAM framework to the public sector—a context marked by high institutional rigidity and significant heterogeneity in technological adaptability—namely, civil servants. This approach responds to the practical need for research on technology adoption mechanisms in public organizations under the current “digital government” transformation. By incorporating key mediating variables—namely self-efficacy and innovation ability—this study further uncovers the cognitive and capability transformation processes that underlie AI application. It enriches the explanatory power of the TAM framework in mapping the process of “perception-intention-behavior-performance”, broadening the model’s theoretical boundary and providing theoretical support for understanding individual technology adoption within complex organizational environments. Secondly, this study constructs and empirically validates a chain mediation model that integrates Bandura’s self-efficacy theory with organizational innovation theory to examine how AI application influences civil servants’ work performance through the sequential mediators of self-efficacy and innovation ability. This study agrees with some researchers that self-efficacy has a positive impact on an individual’s ability to innovate (Tierney & Farmer, 2011 ; Yu, 2013 ). Building on this, this study further demonstrates that, in the context of digital government, the application of AI can effectively enhance civil servants’ self-efficacy, which in turn promotes their innovation ability and ultimately improves their work performance. Our findings contribute a psychological perspective to understanding the behavioral mechanisms of civil servants in the context of digital transformation. Thirdly, this study takes an innovative approach by measuring willingness to change from both individual and environmental dimensions, introducing it as a moderating variable. Traditional research based on the TAM has primarily focused on the impact of technological attributes on individual willingness to use. Such studies assume a relatively uniform adaptive tendency among users when facing new technologies, often overlooking potential differences in individuals’ cognitive readiness and psychological receptiveness (Legris et al., 2003 ; Venkatesh & Davis, 2000 ). By fully considering the individual heterogeneity in willingness to change, this study empirically confirms its moderating effects on two critical paths: from AI application to innovation ability, and from self-efficacy to innovation ability. These findings offer a beneficial extension of the TAM by incorporating psychological and contextual variability into the analysis of technology adoption. 5.2.2 Management implications This study reveals the multiple pathways through which AI influences civil servants’ work performance, particularly highlighting its role in enhancing self-efficacy and innovation ability, as well as the moderating function of willingness to change. Based on these findings, this paper offers the following three recommendations for the future transformation of public administration: Firstly, AI competency development should be integrated into the institutional training system for civil servants. With the deepening application of AI technologies in public service delivery, traditional competency frameworks for civil servants are no longer adequate to meet the demands of intelligent governance. It is recommended that digital competencies, such as AI literacy, be integrated into civil service career development frameworks. Structured training programs should be implemented to improve civil servants’ technological comprehension and application skills, enabling them to competently engage in data governance, algorithmic management, and intelligent decision-making within a digital government environment. This shift will support the transformation of civil servants from “technology users” to “intelligent governance actors”. Secondly, innovation-oriented organizational culture and institutional environment should be fostered. Innovation ability, as the key mediating variable connecting technology adoption and performance improvement, requires strong organizational and institutional support to be fully realized. Public organizations should transition from an “execution-oriented” to an “innovation-driven” model. Institutional design should increase tolerance for trial-and-error and promote workflow optimization to create an exploratory and supportive work atmosphere that encourages civil servants to adapt and innovate in diverse scenarios. Additionally, organizational innovation capacity should be incorporated into performance evaluation and cadre assessment systems. A structured incentive and accountability framework can help cultivate sustained innovation and shift governance paradigms from reactive adaptation to proactive innovation. Thirdly, greater emphasis should be placed on stimulating individual motivation for change and cultivating a supportive atmosphere for technology adoption. The findings suggest that willingness to change significantly moderates the impact of AI use on innovation ability, indicating that the effectiveness of technology adoption depends not only on the functional advantages of AI but also on users’ psychological perceptions and value alignment. Therefore, activating civil servants’ intrinsic motivation for change is essential for advancing intelligent governance. On the one hand, public organizations should establish robust support systems for AI adoption to boost civil servants’ confidence and sense of psychological safety in the face of technological change. On the other hand, they should build inclusive and participatory organizational mechanisms that empower civil servants to take ownership in the transformation process, thereby promoting deeper integration between technological and organizational change. 5.3 Limitations and directions for future research Although this study offers a systematic examination of how AI affects the work performance of civil servants, it is subject to several limitations. Firstly, the sample primarily comprises grass-root civil servants, whose administrative ranks are mainly concentrated at the staff and section levels. As the operational core of China’s public administration, this group is highly involved in the execution of routine governmental functions and public affairs, thus holding a certain degree of representativeness. However, the absence of mid-level and high-level civil servants (e.g., department-level and above) constrains the study’s capacity to capture the heterogeneous characteristics of AI application among leadership in China's public sector. Future research should consider incorporating a broader range of administrative ranks to allow for hierarchical comparisons in AI adoption behaviors. Secondly, all empirical data were collected in China. Given China’s distinctive institutional settings and administrative culture, the findings may not be readily generalizable to other national contexts. Subsequent studies are encouraged to adopt a cross-national perspective by conducting comparative analyses across different countries Declarations The data that support the findings of this study are available on request from the corresponding author. Ethics approval and consent to participate The Ethics Committee of the South China University of Technology provided ethical approval for this study in strict accordance with the guidelines of the Declaration of Helsinki. All participants were informed of the nature and purpose of the study and provided their informed consent prior to participation. Consent for publication No applicable. Availability of data and material The datasets generated and analyzed during the current study are not publicly available due to participant confidentiality and institutional data protection policies but are available from the corresponding author on reasonable request. Competing interests The authors declare no competing interests. Funding No applicable. Authors' contributions L.Z contributed to conceptualization, research design, data processing, empirical analysis, software implementation, figure and table preparation, writing-original draft, and writing review & editing. G.D contributed to survey design, data collection, writing-original draft, and writing review & editing. All authors ead and approved the final manuscript. Acknowledgements No applicable. Authors' information 1 Zongyang Li, South China University of Technology, School of Public Administration, Guangzhou, Guangdong 510641, People’s Republic of China. 2 Dongming Gu, South China University of Technology, School of Public Administration, Guangzhou, Guangdong 510641, People’s Republic of China. References Abioye SO, Oyedele LO, Akanbi L, Ajayi A, Delgado JMD, Bilal M, Akinade OO, Ahmed A. Artificial intelligence in the construction industry: A review of present status, opportunities and future challenges. J Building Eng. 2021;44:103299. https://doi.org/doi:10.1016/j.jobe.2021.103299 . Ballet K, Kelchtermans G. Workload and willingness to change: Disentangling the experience of intensification. J curriculum Stud. 2008;40(1):47–67. https://doi.org/doi:10.1080/00220270701516463 . Bandura A. Self-efficacy mechanism in human agency. Am Psychol. 1982;37(2):122. https://doi.org/doi:10.1037/0003-066x.37.2.122 . Bandura A. Self-efficacy: The foundation of Agency1. Control of human behavior, mental processes, and consciousness. Psychology; 2013. pp. 16–30. Baron RM, Kenny DA. The moderator–mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. J Personal Soc Psychol. 1986;51(6):1173. https://doi.org/doi:10.1037/0022-3514.51.6.1173 . Benson T. (2019). Digital innovation evaluation: user perceptions of innovation readiness, digital confidence, innovation adoption, user experience and behaviour change. BMJ health care Inf, 26(1), e000018. Bilan Y, Mishchuk H, Samoliuk N. Digital skills of civil servants: Assessing readiness for successful interaction in e-society. Acta Polytech Hungarica. 2023;20(3):155–74. Bolin JH. Introduction to mediation, moderation, and conditional process analysis: a regression-based approach. In: JSTOR; 2014. Caianiello M. Dangerous liaisons. Potentialities and risks deriving from the interaction between artificial intelligence and preventive justice. Eur J Crime Criminal Law Criminal Justice. 2021;29(1):1–23. https://doi.org/doi:10.1163/15718174-29010001 . Casalino N, Saso T, Borin B, Massella E, Lancioni F. In: Agrifoglio R, Lamboglia R, Mancini D, Ricciardi F, editors. Digital Competences for Civil Servants and Digital Ecosystems for More Effective Working Processes in Public Organizations. Digital Business Transformation Cham; 2020a. Casalino N, Saso T, Borin B, Massella E, Lancioni F. Digital competences for civil servants and digital ecosystems for more effective working processes in public organizations. Digital Business Transformation: Organizing, Managing and Controlling in the Information Age; 2020b. Chen S-C, Shing-Han L, Chien-Yi L. Recent related research in technology acceptance model: A literature review. Australian J Bus Manage Res. 2011;1(9):124. Choi M. Employees' attitudes toward organizational change: A literature review. Hum Resour Manag. 2011;50(4):479–500. https://doi.org/doi:10.1002/hrm.20434 . Davis FD, Granić A, Marangunić N. The technology acceptance model: 30 years of TAM. Springer; 2024. Doroodian M, Ab Rahman MN, Kamarulzaman Y, Muhamad N. Designing and validating a model for measuring innovation capacity construct. Adv Decis Sci. 2014. https://doi.org/doi:10.1155/2014/576596 . Grant D, Jha PP, Wanjiru R, Bhalla A. Evolving willingness and ability interfaces: An innovation led transformation journey. Int J Innov Stud. 2020;4(3):69–75. https://doi.org/doi:10.1016/j.ijis.2020.06.001 . Herold DM, Fedor DB, Caldwell S, Liu Y. The effects of transformational and change leadership on employees' commitment to a change: a multilevel study. J Appl Psychol. 2008;93(2):346. https://doi.org/doi:10.1037/0021-9010.93.2.346 . Hood C. The new public management in the 1980s: Variations on a theme. Acc Organ Soc. 1995;20(2–3):93–109. https://doi.org/doi:10.1016/0361-3682(93)E0001-W . Hwang Y, Wu Y. The influence of generative artificial intelligence on creative cognition of design students: a chain mediation model of self-efficacy and anxiety. Front Psychol. 2025;15:1455015. https://doi.org/doi:10.3389/fpsyg.2024.1455015 . Iaffaldano MT, Muchinsky PM. Job satisfaction and job performance: A meta-analysis. Psychol Bull. 1985;97(2):251. https://doi.org/doi:10.1037/0033-2909.97.2.251 . Ivić A, Milićević A, Krstić D, Kozma N, Havzi S. (2022). The challenges and opportunities in adopting AI, IoT and blockchain technology in e-government: A systematic literature review. 2022 International Conference on Communications, Information, Electronic and Energy Systems (CIEES). Jackson PC. Introduction to artificial intelligence. Courier Dover; 2019. Jiao H, Zhao G. When will employees embrace managers' technological innovations? The mediating effects of employees' perceptions of fairness on their willingness to accept change and its legitimacy. J Prod Innov Manage. 2014;31(4):780–98. https://doi.org/doi:10.1111/jpim.12123 . Lartey D, Law KM. Artificial intelligence adoption in urban planning governance: A systematic review of advancements in decision-making, and policy making. Landsc Urban Plann. 2025;258:105337. https://doi.org/doi:10.1016/j.landurbplan.2025.105337 . Laviolette EM, Redien-Collot R, Teglborg A-C. Open innovation from the inside: Employee-driven innovation in support of absorptive capacity for inbound open innovation. Int J Entrepreneurship Innov. 2016;17(4):228–39. https://doi.org/doi:10.1177/146575031667049 . Legris P, Ingham J, Collerette P. Why do people use information technology? A critical review of the technology acceptance model. Inf Manag. 2003;40(3):191–204. https://doi.org/doi:10.1016/S0378-7206(01)00143-4 . Lemke C. (2010). Innovation through technology. 21st century skills: Rethinking how students learn, 243–272. Lexer MG, Scarcella L. Artificial intelligence and labor markets. A critical analysis of solution models from a tax law and social security law perspective. Rivista italiana di informatica e diritto. 2019;1(1):53–73. Marangunić N, Granić A. Univ Access Inf Soc. 2015;14:81–95. https://doi.org/doi:10.1007/s10209-014-0348-1 . Technology acceptance model: a literature review from 1986 to 2013. Maxwell L, Taner E, Jonathan GM. Digitalisation in the public sector: Determinant factors. Int J IT/Business Alignment Gov (IJITBAG). 2019;10(2):35–52. https://doi.org/doi:10.4018/IJITBAG.2019070103 . Mehr H, Ash H, Fellow D. Artificial intelligence for citizen services and government. Ash Cent. Democr. Gov. Innov. Harvard Kennedy Sch. no August. 2017;1:12. Metselaar EE. Assessing the willingness to change. Construction and validation of the DINAMO; 1997. Miller VD, Johnson JR, Grau J. (1994). Antecedents to willingness to participate in a planned organizational change. Obenza BN, Salvahan A, Rios AN, Solo A, Alburo RA, Gabila RJ. University students' perception and use of ChatGPT: Generative artificial intelligence (AI) in higher education. Int J Hum Comput Stud. 2024;5(12):5–18. Park N, Rhoads M, Hou J, Lee KM. Understanding the acceptance of teleconferencing systems among employees: An extension of the technology acceptance model. Comput Hum Behav. 2014;39:118–27. https://doi.org/doi:10.1016/j.chb.2014.05.048 . Prajogo DI, Ahmed PK. Relationships between innovation stimulus, innovation capacity, and innovation performance. R&d Manage. 2006;36(5):499–515. https://doi.org/doi:10.1111/j.1467-9310.2006.00450.x . Ramadian A, Chairuddin C, Judijanto L, Rachmawati R, Sopandi E, ARTIFICIAL INTELLIGENCE COMPETENCIES, ORGANIZATIONAL SUPPORT, AND EMPLOYEE SELF-EFFICACY IN PREDICTING GOVERNMENT EMPLOYEE PERFORMANCE: A MEDIATION ANALYSIS WITH WORK ENGAGEMENT. Jurnal Manajemen dan Kewirausahaan. 2025;27(1):22–32. https://doi.org/doi:10.9744/jmk.27.1.22-32 . THE ROLE OF. Rusho MA, Chan MP. Analysis of Artificial Intelligence and its Impactful Implementation on Job Performance. Dinkum J Econ Managerial Innovations. 2023;2(10):571–6. Sari RF. (2025). Exploring Artificial Intelligence Adoption in the Public Sector: Perceptions, Attitudes and Factors Influencing AI Adoption Among Government Public Relations Practitioners in Indonesia University of Twente]. Seckelmann M, Catakli D. Digital Competencies in the Civil Service. The Civil Service in Europe. Routledge; 2025. pp. 662–80. Shahzad MF, Xu S, Naveed W, Nusrat S, Zahid I. Investigating the impact of artificial intelligence on human resource functions in the health sector of China: A mediated moderation model. Heliyon. 2023;9(11). https://doi.org/doi:10.1016/j.heliyon.2023.e21818 . Shank DB, Cotten SR. Does technology empower urban youth? The relationship of technology use to self-efficacy. Comput Educ. 2014;70:184–93. https://doi.org/doi:10.1016/j.compedu.2013.08.018 . Shoji K, Cieslak R, Smoktunowicz E, Rogala A, Benight CC, Luszczynska A. Associations between job burnout and self-efficacy: A meta-analysis. Anxiety Stress Coping. 2016;29(4):367–86. https://doi.org/doi:10.1080/10615806.2015.1058369 . Teo T. Modelling technology acceptance in education: A study of pre-service teachers. Comput Educ. 2009;52(2):302–12. https://doi.org/doi:10.1016/j.compedu.2008.08.006 . Tierney P, Farmer SM. Creative self-efficacy development and creative performance over time. J Appl Psychol. 2011;96(2):277. https://doi.org/. Venkatesh V, Davis FD. A theoretical extension of the technology acceptance model: Four longitudinal field studies. Manage Sci. 2000;46(2):186–204. https://doi.org/doi:10.1287/mnsc.46.2.186.11926 . Wirtz BW, Weyerer JC, Geyer C. Artificial intelligence and the public sector—applications and challenges. Int J Public Adm. 2019;42(7):596–615. https://doi.org/doi:10.1080/01900692.2018.1498103 . Xu G, Xue M. The relationship between proactive personality and migrant workers’ perception of technical unemployment risk under the impact of artificial intelligence in China. Front Psychol. 2025;16:1474639. https://doi.org/doi:10.3389/fpsyg.2025.1474639 . Yu C. The relationship between undergraduate students’ creative self-efficacy, creative ability and career self-management. Int J Acad Res Progressive Educ Dev. 2013;2(2):181–93. Zhang T. Effects of self-regulation strategies on EFL learners’ language learning motivation, willingness to communication, self-efficacy, and creativity. BMC Psychol. 2024;12(1):75. https://doi.org/10.1186/s40359-024-01567-2 . Zhao J, Li X, Wei J, Long X, Gao Z. Understanding the psychological pathways to translation technology competence: emotional intelligence, self-esteem, and innovation capability among EFL students. BMC Psychol. 2025;13(1):66. https://doi.org/10.1186/s40359-025-02400-0 . Zhao P, Cao S. To participate or not to participate? Influence mechanism of artificial intelligence on Chinese college students’ willingness to participate in online politics. BMC Psychol. 2024;12(1):525. https://doi.org/10.1186/s40359-024-02009-9 . Footnotes Self-efficacy and Work Performance. Innovation Ability and Work Performance. Additional Declarations No competing interests reported. Supplementary Files Database.xlsx Cite Share Download PDF Status: Under Review Version 1 posted Reviews received at journal 06 Aug, 2025 Reviewers agreed at journal 01 Aug, 2025 Reviewers agreed at journal 27 Jul, 2025 Reviewers agreed at journal 26 Jul, 2025 Reviewers invited by journal 25 Jul, 2025 Editor assigned by journal 23 Jul, 2025 Editor invited by journal 09 May, 2025 Submission checks completed at journal 06 May, 2025 First submitted to journal 06 May, 2025 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-6526907","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":491449157,"identity":"f40023d7-2e8f-4980-a7ac-4256a9e27b11","order_by":0,"name":"Zongyang Li","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA6UlEQVRIie3PMWvCQBTA8QcH53Ka9QUk/QongUxCv8odkUwptItkcLiSkgyt+lWcxDEhcC4nXTOab1A3Cx3aXenFzeF+8/vz3gNwnDtEB+uuO/0ge/zM66PIFvZkxAwJfTUNoNUxPxptTwJMKQ5VEkKVRn73Rnocxg4V+rtGvioTZVJR8Mp3YfllJfiLaWROinkrd2NAc9hYtlRcIG1kAc22lYYCxydLgoJX7C/5ABE9y4L0SdKJGhZJiDCLoF/CdEzQTAOOOkZhNLP+8lDm+++vDBn38vp0zhaBVy7/Ty6w28Ydx3Gcq34B6DFNsrSBOXIAAAAASUVORK5CYII=","orcid":"","institution":"South China University of Technology","correspondingAuthor":true,"prefix":"","firstName":"Zongyang","middleName":"","lastName":"Li","suffix":""},{"id":491449158,"identity":"ab219669-240a-42b6-8bfb-ecd4d4018bf7","order_by":1,"name":"Dongming Gu","email":"","orcid":"","institution":"South China University of Technology","correspondingAuthor":false,"prefix":"","firstName":"Dongming","middleName":"","lastName":"Gu","suffix":""}],"badges":[],"createdAt":"2025-04-25 08:23:28","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-6526907/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-6526907/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":87742628,"identity":"77561f07-4b36-4893-9a7e-47f6ba5792f1","added_by":"auto","created_at":"2025-07-28 13:46:12","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":30772,"visible":true,"origin":"","legend":"\u003cp\u003eResearch model map\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-6526907/v1/079d96ee4ff809a8425afe29.png"},{"id":87744074,"identity":"52a846c8-d92b-49f4-8b37-4b15b83ec258","added_by":"auto","created_at":"2025-07-28 13:54:13","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":34217,"visible":true,"origin":"","legend":"\u003cp\u003eRelation model map.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-6526907/v1/579a42df2a6ea387aad53444.png"},{"id":87742632,"identity":"0f92c33e-876e-4122-a7b1-779a15dcefc8","added_by":"auto","created_at":"2025-07-28 13:46:13","extension":"jpg","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":36235,"visible":true,"origin":"","legend":"\u003cp\u003eModerating effect of willingness to change between AI application and innovation ability.\u003c/p\u003e","description":"","filename":"3.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6526907/v1/66c3c5750f94764a5c7a036b.jpg"},{"id":87742633,"identity":"051f407e-9ec1-44c6-84da-73fd5c2e17c5","added_by":"auto","created_at":"2025-07-28 13:46:13","extension":"jpg","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":34663,"visible":true,"origin":"","legend":"\u003cp\u003eModerating effect of willingness to change between self-efficacy and innovation ability.\u003c/p\u003e","description":"","filename":"4.jpg","url":"https://assets-eu.researchsquare.com/files/rs-6526907/v1/08ab72dce46dea9d92c30865.jpg"},{"id":87746261,"identity":"2cde6aaf-0014-4964-9b85-fcbbc64050ce","added_by":"auto","created_at":"2025-07-28 14:18:14","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1314931,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-6526907/v1/36e54ab3-42c2-404b-847f-c197ddc2a87a.pdf"},{"id":87744073,"identity":"4ab1d296-e477-4f50-9f91-f61a0c95cb54","added_by":"auto","created_at":"2025-07-28 13:54:13","extension":"xlsx","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":84238,"visible":true,"origin":"","legend":"","description":"","filename":"Database.xlsx","url":"https://assets-eu.researchsquare.com/files/rs-6526907/v1/f710a0aa87273a7c0a8f3aac.xlsx"}],"financialInterests":"No competing interests reported.","formattedTitle":"The relationship between artificial intelligence and civil servants’ work performance: a chain mediation model of self-efficacy and innovation ability","fulltext":[{"header":"1 Introduction","content":"\u003cp\u003eArtificial intelligence (AI) refers to a set of technologies and methodologies that enable computer systems to simulate human cognitive processes such as perception, comprehension, reasoning, learning, and decision-making (Jackson, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). In recent years, with continuous advancements and maturation of AI technologies, AI has been widely applied across various domains\u0026mdash;including healthcare (Beam et al., 2023), traffic management (Caianiello, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2021\u003c/span\u003e), education (Hwang \u0026amp; Wu, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), urban governance (Lartey \u0026amp; Law, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), and social security systems (Lexer \u0026amp; Scarcella, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2019\u003c/span\u003e)\u0026mdash;and has profoundly reshaped service delivery models and governance paradigms within the public sector (Wirtz et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eSince the emergence of the New Public Management movement in the 1980s, public sector organizations have gradually undergone a transformation toward service-oriented and efficiency-driven structures (Hood, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e1995\u003c/span\u003e). Improving the quality of public services and enhancing administrative efficiency have become key objectives. In the 21st century, the rapid advancement of digital technologies\u0026mdash;including the internet, big data, and AI\u0026mdash;has served as a critical driver for digital transformation in the public sector, representing a pivotal strategy for achieving the effectiveness and responsiveness of public services (Maxwell et al., \u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eFrom a macro perspective, the digitization of the public sector refers to the deep integration of digital technologies into public sector governance, led by governments, encompassing transformations in governance concepts, organizational structures, and institutional mechanisms. This process drove initiatives such as \u0026ldquo;digital government\u0026rdquo; (Janowski, 2015). Specifically, it involves efforts to strengthen the integration of administrative information systems and the sharing of data resources, ultimately advancing the intelligence and responsiveness of governance systems. From a micro perspective, public sector digitization is reflected in the use of digital tools\u0026mdash;such as data analytics, intelligent decision-support systems, and online government service platforms\u0026mdash;by civil servants in daily administrative practices. These tools help streamline workflows, improve service efficiency, and strengthen government responsiveness to public demands, thereby facilitating the precise and personalized provision of public services (Bilan et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2023\u003c/span\u003e; Seckelmann \u0026amp; Catakli, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eAt present, AI is increasingly becoming a key driver in reshaping public service delivery models and reconstructing policy implementation mechanisms. As the primary providers of public services and executors of governance, civil servants are also undergoing significant changes in their role positioning and competency structures (Mehr et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). Moreover, while AI contributes to optimizing governance processes and enhancing administrative efficiency, the deep integration of emerging technologies such as AI has simultaneously transformed the logic of government service provision. This shift compels civil servants to evolve from traditional task-oriented executors into compound governance professionals equipped with technological literacy and data competence (Lexer \u0026amp; Scarcella, \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). However, AI also imposes institutional transformations and professional disruptions upon the civil service, which are particularly salient amid China\u0026rsquo;s ongoing digital transformation in the public sector. In 2025, the breakout success of DeepSeek marked a new stage of maturity for China's large language model technologies. Building on these advancements, Shenzhen Municipality in Guangdong Province launched a pilot program featuring \u0026ldquo;AI-powered digital civil servants\u0026rdquo; within its administrative system, aiming to automate routine tasks such as administrative consultation, information entry, and process approvals, thereby improving operational efficiency in government services. Against this backdrop of accelerated governmental digital transformation, the traditional professional identity and skill structure of civil servants are facing mounting challenges brought about by technological change.\u003c/p\u003e\u003cp\u003eIn the context of the widespread application of AI in governmental systems, shifts in identity and behavioral choices among civil servants have become a critical component of the government\u0026rsquo;s digital transformation. Research in environmental psychology suggests that individual behavior is not only influenced by objective environmental changes but also shaped by subjective perceptions of those changes. When significant technological transformations occur, especially those that impact occupational stability and role expectations, individuals may experience psychological stress responses and engage in adaptive behaviors. As a structurally stable force within the national governance system, civil servants\u0026rsquo; engagement with AI\u0026mdash;reflected in their acceptance, risk perception, and behavioral responses\u0026mdash;is crucial not only for their professional adaptation but also for the sustainability of digital governance and the effective implementation of AI-driven administration.\u003c/p\u003e\u003cp\u003eIn recent years, scholarly interest in how AI influences the civil service has deepened, with increasing attention paid to its role in enhancing administrative efficiency and reshaping competency structures. Some studies affirm the positive contributions of AI in improving government operations. In studies on AI practices in the United States, it has been found that AI enables public authorities to analyze public sentiment through data analytics, thereby allowing governments to more precisely identify societal needs and deliver targeted public services, enhancing citizen satisfaction (Mehr et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2017\u003c/span\u003e). With the ongoing evolution of generative AI technologies, intelligent systems are increasingly capable of assisting civil servants with routine administrative tasks such as document drafting and information archiving, thereby freeing them from repetitive labor and allowing them to focus on more strategic and creative aspects of governance (Caianiello, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). In addition, AI supports public decision-making by offering simulations and probabilistic estimations that help officials handle complex public issues more effectively. Based on these perspectives, scholars argue that the digital transformation driven by AI substantially enhances civil servants' work efficiency. However, a contrasting body of literature expresses concern. Some researchers argue that deep integration of AI may lead to over-reliance on technology, thereby weakening civil servants' independent judgment and problem-solving capacities (Ivić et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2022\u003c/span\u003e). Furthermore, the hyper-rationality introduced by AI systems may conflict with ethical and moral dimensions inherent in public governance, potentially leading to more complex governance challenges. When facing diverse and context-sensitive governance issues, standardized technical solutions are insufficient to replace human judgment, contextual awareness, and adaptive decision-making, potentially undermining civil servants\u0026rsquo; creativity and adaptability (Abioye et al., \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2021\u003c/span\u003e). Additionally, the technological shift caused by AI threatens to displace repetitive, basic, and routine roles within the public sector, heightening concerns over job security and professional anxiety among civil servants (Sari, \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eIn sum, despite ongoing debates regarding the impact of AI on the civil service, it is undeniable that the emergence of AI has introduced profound changes to both public sector operations and the work performance of civil servants. Accordingly, academic inquiry has increasingly focused on how and through which mechanisms AI affects civil servants\u0026rsquo; work performance. Among the various explanatory factors, self-efficacy and innovative ability have received widespread attention due to their critical role in accounting for individual performance differences. Some scholars argued that the promotion of AI competencies can enhance civil servants\u0026rsquo; self-efficacy, thereby enabling them to approach their tasks with greater confidence and competence (Ramadian et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Meanwhile, innovation capability, as a key dimension of individual resource integration and problem-solving, plays a vital role in advancing organizational performance. Meanwhile, innovation ability, as a key dimension of individual resource integration and problem-solving, is also a crucial factor in enhancing organizational performance. Civil servants\u0026rsquo; proficiency in AI technologies is positively associated with their innovative capabilities, and those with stronger innovation ability tend to perform better in terms of task execution and responsiveness to public needs.\u003c/p\u003e\u003cp\u003eTaking into account the intricate interconnections among these variables and the identified gaps in existing literature, this study aims to reveal the complex connections among AI application, self-efficacy, innovation ability, work performance, and willingness to change. To be more exact, this study sought to explore how AI application affect civil servants\u0026rsquo; work performance, the role self-efficacy plays in this process, the extent to which innovation ability mediates this relationship, and how willingness to change contributes to the enhancement of innovation ability through AI application and self-efficacy. To achieve this goal, this study collected 296 completed questionnaires from MPA(Master of Public Administration) students who are employed in various public sector organizations. Based on this, a series of empirical analyses were conducted, including a structural equation model (SEM), chain mediation model, moderation model, and bootstrap method. These analyses were designed to identify the logical relationships and influence mechanisms among the five core latent variables. The research results of this study are beneficial for examining the chain-mediating pathway and the moderating role of willingness to change, which provide theoretical insights and empirical evidence for empowering AI technologies to enhance the work performance of civil servants.\u003c/p\u003e"},{"header":"2 Research hypotheses","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1 AI application and work performance\u003c/h2\u003e\u003cp\u003eAccording to the Technology Acceptance Model (TAM), individuals\u0026rsquo; technology usage behavior is primarily influenced by their Perceived Usefulness and Perceived Ease of Use of the technology, both of which subsequently exert a direct influence on performance outcomes (Marangunić \u0026amp; Granić, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2015\u003c/span\u003e). In the context of accelerating digital transformation within the public sector, artificial intelligence technologies\u0026mdash;particularly generative AI(AIGC)\u0026mdash;have been widely adopted in routine tasks such as document drafting, information retrieval, data processing, and decision support. These technologies provide important and substantial assistance to civil servants by improving task completion speed, reducing administrative burden, and enhancing information processing efficiency (Casalino et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2020a\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003eb\u003c/span\u003e). Existing studies have shown that effective use of AI can optimize task workflows, facilitate knowledge acquisition, and improve information management, thereby contributing to both individual productivity and overall organizational performance (Rusho \u0026amp; Chan, \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2023\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eIn China\u0026rsquo;s digital government initiatives, civil servants are often required to process large volumes of structured and unstructured data. The application of AI technologies helps alleviate these cognitive and labor-intensive demands, thereby enhancing their capacity to handle complex tasks (Shahzad et al., \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2023\u003c/span\u003e). In summary, the application of AI by civil servants can significantly enhance their capabilities in information acquisition, document processing, and assisted decision-making. This, in turn, reduces routine workload and contributes to higher execution efficiency and job performance in complex administrative environments. Therefore, this paper proposes the following hypothesis:\u003c/p\u003e\u003cp\u003e\u003cem\u003eHypothesis H1: AI application is positively associated with work performance.\u003c/em\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2 The mediating role of self-efficacy\u003c/h2\u003e\u003cp\u003eSelf-efficacy refers to individual subjective judgment regarding their ability to accomplish specific tasks, directly influencing behavioral motivation, task engagement, and performance levels (Bandura, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e1982\u003c/span\u003e). The TAM also emphasizes that technology use is not solely determined by the attributes of the technology itself, but also affects work performance by shaping individuals\u0026rsquo; cognitive and psychological states (Chen et al., \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). In the context of AI integration into public governance practices, the application of AI by civil servants contributes to their perception of task feasibility and convenience, which in turn enhances their confidence in their abilities\u0026mdash;that is, their self-efficacy. For instance, when civil servants use generative AI to assist in drafting official documents, retrieving information, or processing data, they may experience a tangible sense of improved efficiency. This positive experiential feedback enhances their perception of their work capabilities, which in turn fosters greater motivation and initiative in task execution.\u003c/p\u003e\u003cp\u003eMoreover, existing research has confirmed that self-efficacy significantly enhances work performance by strengthening an individual\u0026rsquo;s goal orientation and increasing their willingness to undertake challenging tasks, which in turn improves the quality and efficiency of task completion (Bandura, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e1982\u003c/span\u003e). In addition, self-efficacy has been found to enhance psychological resilience in the face of pressure or obstacles, reduce feelings of burnout, and help maintain high levels of performance (Shoji et al., \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). Taken together, technology use facilitates the development and maintenance of self-efficacy, which itself is a critical psychological factor influencing individual work performance (Shank \u0026amp; Cotten, \u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Therefore, self-efficacy serves as a psychological mechanism that links AI application with work performance. Based on this, the following hypotheses are proposed:\u003c/p\u003e\u003cp\u003e\u003cem\u003eHypothesis H2a: The wide application of AI would increase civil servants\u0026rsquo; self-efficacy.\u003c/em\u003e\u003c/p\u003e\u003cp\u003e\u003cem\u003eHypothesis H3: The increase in self-efficacy due to the use of AI would mediate the improvement of work performance.\u003c/em\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e2.3 The mediating role of innovation ability\u003c/h2\u003e\u003cp\u003eInnovation ability refers to the individual capacity to identify problems, generate new ideas, integrate resources, and develop solutions within specific contexts. It is a critical factor influencing one\u0026rsquo;s creative contributions and performance within an organization (Prajogo \u0026amp; Ahmed, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2006\u003c/span\u003e). In the era of rapidly evolving information technologies, artificial intelligence offers civil servants novel tools for knowledge acquisition, task generation, and cognitive expansion. Its capabilities in multimodal content creation, language structuring, and data analysis provide algorithmic support for stimulating innovative thinking. The application of AI disrupts traditional path-dependent approaches to task execution by offering more flexible and efficient methods of working (Zhao \u0026amp; Cao, \u003cspan citationid=\"CR52\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). This encourages civil servants to adopt more creative and proactive approaches to problem identification, strategic decision-making, and resource allocation. Studies have shown that technology adoption can enhance innovation awareness and behavior by improving information processing capabilities and reducing the cost of trial-and-error in innovation (Benson, \u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2019\u003c/span\u003e; Lemke, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2010\u003c/span\u003e). Meanwhile, innovation ability\u0026mdash;considered a high-level cognitive and practical competence\u0026mdash;influences work performance not only by transforming ways of thinking but also through direct contributions to task effectiveness. Civil servants with strong innovation capabilities are better equipped to develop novel solutions and overcome existing limitations when dealing with complex or unstructured problems, thus improving task quality and organizational adaptability (Laviolette et al., \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). Besides, innovation ability also contributes to optimizing service processes and public resource allocation, ultimately enhancing governance efficiency and citizen satisfaction. Empirical research has also confirmed a significant positive correlation between innovation capacity and both individual and organizational performance (Prajogo \u0026amp; Ahmed, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2006\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eAccordingly, innovation ability may serve as both a cognitive and skill-based outcome of AI utilization, as well as a key mediating variable that links AI use with improved work performance among civil servants. Based on this, the following hypotheses are proposed:\u003c/p\u003e\u003cp\u003e\u003cem\u003eHypothesis H2b: The wide application of AI would increase civil servants\u0026rsquo; innovation ability.\u003c/em\u003e\u003c/p\u003e\u003cp\u003e\u003cem\u003eHypothesis H4: The increase in innovation ability due to the use of AI would mediate the improvement of work performance.\u003c/em\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\u003ch2\u003e2.4 The chain-mediating role of self-efficacy and innovation ability\u003c/h2\u003e\u003cp\u003eBeyond independent mediating mechanisms, the application of AI may also exert an indirect impact on civil servants\u0026rsquo; work performance through a sequential mediation pathway involving self-efficacy and innovation ability. According to self-efficacy theory, individuals\u0026rsquo; beliefs in their capabilities to accomplish specific tasks significantly enhance their initiative and willingness to explore when confronted with challenging situations (Marangunić \u0026amp; Granić, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Zhao et al., \u003cspan citationid=\"CR51\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). They tend to exhibit stronger capabilities in problem identification, strategy generation, and resource integration, thereby exhibiting a higher level of innovation ability (Bandura, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). Furthermore, the TAM suggests that perceived usefulness and perceived ease of use not only shape individual willingness to adopt technology but also indirectly affect performance by influencing cognitive evaluations and psychological responses (Davis et al., \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2024\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eWithin an administrative environment where AI-assisted work is increasingly widespread, civil servants who receive positive feedback during AI use\u0026mdash;such as improved task efficiency and more effective information processing\u0026mdash;are likely to experience enhanced self-efficacy. This strengthened self-efficacy, in turn, improves their ability to adapt to technological change and motivates them to proactively apply AI tools in problem-solving, thereby enhancing their innovation ability. The enhancement of innovation ability, in return, lays the groundwork for improved service delivery and work performance.\u003c/p\u003e\u003cp\u003eTherefore, the application of AI may influence work performance not only through isolated psychological mechanisms but also through a chain mediation pathway involving self-efficacy and innovation ability, reflecting a deeper and more complex route of indirect influence. Based on this theoretical reasoning, the following hypothesis is proposed:\u003c/p\u003e\u003cp\u003e\u003cem\u003eHypothesis H5: self-efficacy and innovation ability play a chain-mediating role in the relationship between AI application and work performance.\u003c/em\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec7\" class=\"Section2\"\u003e\u003ch2\u003e2.5 The moderating role of willingness to change\u003c/h2\u003e\u003cp\u003eIn technology-driven organizational change, individuals\u0026rsquo; attitudes toward change and their willingness to accept it are critical factors influencing the effectiveness of technology implementation and the development of individual innovation capabilities (Grant et al., \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2020\u003c/span\u003e; Jiao \u0026amp; Zhao, \u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Willingness to change refers to an individual proactive disposition and behavioral tendency in response to environmental shifts and organizational technological reform. It is a key psychological precondition for converting personal resources into adaptive actions (Ballet \u0026amp; Kelchtermans, \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2008\u003c/span\u003e).\u003c/p\u003e\u003cp\u003eExisting studies suggest that, beyond individual motivations, organizational factors\u0026mdash;such as the climate for change, leadership orientation, and institutional support\u0026mdash;also play a crucial role in shaping employees\u0026rsquo; change-related behaviors (Herold et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2008\u003c/span\u003e). When organizational culture actively encourages technology adoption and institutional innovation, or when supervisors demonstrate clear support for change, individuals are more likely to perceive technological reform as legitimate and necessary. This recognition helps cultivate a positive psychological attitude toward change, thereby strengthening their willingness to embrace new technologies and adapt proactively.\u003c/p\u003e\u003cp\u003eThus, in an environment characterized by a strong change-oriented culture, civil servants with a high willingness to change are more inclined to perceive AI as an opportunity for innovation and process optimization. They are likely to engage in active learning and exploration during AI application, which in turn promotes the development of their innovation ability. In contrast, those with a low willingness to change may exhibit passive or resistant attitudes during AI implementation, lacking the motivation and intention to transform technological potential into innovative practice.\u003c/p\u003e\u003cp\u003eMoreover, while self-efficacy enhances individuals\u0026rsquo; confidence in task performance, whether this confidence translates into creative behavior often depends on the level of environmental support and the individual\u0026rsquo;s psychological readiness (Marangunić \u0026amp; Granić, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2015\u003c/span\u003e; Zhang, \u003cspan citationid=\"CR50\" class=\"CitationRef\"\u003e2024\u003c/span\u003e). For civil servants with a high willingness to change, the motivation and confidence derived from self-efficacy are more likely to manifest as proactive change-seeking and innovation attempts, thereby strengthening their innovation capabilities. Conversely, in contexts where change motivation is weak or organizational support is lacking, even individuals with high self-efficacy may struggle to translate internal confidence into actual creative outcomes due to the absence of motivational triggers for action (Choi, \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2011\u003c/span\u003e). Based on the above analysis, the following hypotheses are proposed:\u003c/p\u003e\u003cp\u003e\u003cem\u003eHypothesis H6a: Willingness to change exerts a positive moderating influence on the relationship between AI application and innovation ability.\u003c/em\u003e\u003c/p\u003e\u003cp\u003e\u003cem\u003eHypothesis H6b: Willingness to change exerts a positive moderating influence on the relationship between self-efficacy and innovation ability.\u003c/em\u003e\u003c/p\u003e\u003cp\u003eThe hypothesized research model, which visually represents these proposed relationships, is presented in Fig.\u0026nbsp;1.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"No\" id=\"Taba\" border=\"1\"\u003e\u003ccolgroup cols=\"1\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\"\u003e\u003cp\u003e\u003c/p\u003e\u003cp\u003eFIGURE 1\u003c/p\u003e\u003cp\u003eResearch model map\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"},{"header":"3 Methodology","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\u003ch2\u003e3.1 Participant and recruitment\u003c/h2\u003e\u003cp\u003e This study was approved by the Academic Committee of the South China University of Technology and was conducted in strict accordance with the guidelines of the Declaration of Helsinki, with all participants provided informed consent before their participation. A total of 327 MPA students were recruited through an online survey. These participants were employed in various public sector organizations, including government and public institutions. To ensure data quality, responses from participants who failed the attention check or displayed inconsistent answers were excluded. After data cleaning, 296 valid questionnaires were retained, resulting in an effective response rate of 90.52%. The final sample covered 134 prefecture-level cities across 29 provinces (including autonomous regions and municipalities) in China. Among the respondents, 46.6% were male and 53.4% were female. Participants' ages ranged from 22 to 64 years, with an average age of 32.2. Notably, the proportion of those aged 22 to 50 reached 94.6%. In terms of educational background, respondents ranged from high school to Ph.D, with the majority holding a bachelor\u0026rsquo;s degree. 70.9% of the participants were members of the Communist Party of China. The administrative level of the respondents ranged from staff-level(gu-ji) to department-level(chu-ji), with 98.6% being staff-level(gu-ji) and section-level (ke-ji) officers, indicating that the vast majority of respondents are grass-root civil servants. Monthly income levels varied from 1,800 yuan to 20,000 yuan, with an average monthly salary of 7,452 yuan.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\u003ch2\u003e3.2 Measures\u003c/h2\u003e\u003cp\u003eAI application: The scale used to measure AI application was adapted from recent research (Obenza et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2024\u003c/span\u003e; Wirtz et al., \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2019\u003c/span\u003e), with appropriate revisions made to fit the context of this study. The scale consists of six items, including sample statements such as \u0026ldquo;I am willing to discuss recent developments and practical applications of AI with people around me\u0026rdquo; and \u0026ldquo;AI has changed the way I live and work.\u0026rdquo; In the current study, Cronbach\u0026rsquo;s alpha for scores of AI application was 0.6719.\u003c/p\u003e\u003cp\u003eSelf-efficacy: The measurement of self-efficacy among civil servants was adapted from the existing study (Xu \u0026amp; Xue, \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). The scale includes three items, such as \u0026ldquo;I believe I am capable of performing my job well.\u0026rdquo; In the current study, Cronbach\u0026rsquo;s alpha for scores of self-efficacy was 0.7298.\u003c/p\u003e\u003cp\u003eInnovation ability: Based on existing studies (Doroodian et al., \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). The scale consists of four items, including statements such as \u0026ldquo;My writing ability for text-based tasks (official documents, reports, proposals) has improved\u0026rdquo; and \u0026ldquo;I have become more creative and have more innovative ideas.\u0026rdquo; In the current study, Cronbach\u0026rsquo;s alpha for scores of innovation ability was 0.8510.\u003c/p\u003e\u003cp\u003eWork performance: The measurement of work performance was adapted from (Iaffaldano \u0026amp; Muchinsky, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e1985\u003c/span\u003e), with item modifications made to reflect the specific context of this study. The scale comprises five items, including statements such as \u0026ldquo;The average time I need to complete the same tasks has significantly decreased\u0026rdquo; and \u0026ldquo;The quality of output in my work has noticeably improved.\u0026rdquo; In the current study, Cronbach\u0026rsquo;s alpha for scores of work performance was 0.7877.\u003c/p\u003e\u003cp\u003eWillingness to change: To measure this construct, we adapted items from some research (Metselaar, \u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e1997\u003c/span\u003e; Miller et al., \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e1994\u003c/span\u003e). The scale includes four items, such as \u0026ldquo;While completing my tasks with high quality, I also strive for innovation and breakthroughs\u0026rdquo; and \u0026ldquo;My department encourages innovation and change in daily work.\u0026rdquo; In the current study, Cronbach\u0026rsquo;s alpha for scores of willingness to change was 0.7528.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e\u003ch2\u003e3.3 Statistical processing\u003c/h2\u003e\u003cp\u003eTo achieve robust evidence, a chain mediation model test was necessary. Thus, SEM analysis and statistical tests were conducted utilizing STATA 16. We subsequently adopted a mediation testing procedure to evaluate the mediation effects of self-efficacy and innovation ability (Baron \u0026amp; Kenny, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e1986\u003c/span\u003e). Further, we construct a chain mediation model to examine the chain-mediating role of self-efficacy and innovation ability. Additionally, to enhance the robustness of our mediation analysis, we applied the bootstrapping method (Bolin, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2014\u003c/span\u003e). Finally, we conducted moderation effect tests for willingness to change through interaction item construction and simple slope analysis.\u003c/p\u003e\u003c/div\u003e"},{"header":"4 Results","content":"\u003cdiv id=\"Sec13\" class=\"Section2\"\u003e\n\u003ch2\u003e4.1 Common method variance test\u003c/h2\u003e\n\u003cp\u003eThe five key core latent variables used in this paper, which are AI application, self-efficacy, innovation ability, work performance, and willingness to change, come from the same source. To prevent the potential interference of common method variance on the reliability of our findings, Harman\u0026rsquo;s single factor test was used in this study to test the common method variance. The results shown in Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e revealed relatively low factor loadings of a single factor on these variables (mostly ranging between 0.3 and 0.6) accompanied by generally high uniqueness (mostly above 0.7), indicating that a single factor does not dominate the variance explanation for these items. Therefore, it can be concluded that no significant common method variance exists, ensuring a robust foundation in reliability and validity for subsequent structural equation modeling analysis.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec14\" class=\"Section2\"\u003e\n\u003ch2\u003e4.2 Confirmatory factor analysis(CFA)\u003c/h2\u003e\n\u003cp\u003eTo further verify the discriminant validity among variables, we performed a CFA using Stata 16.0 software. There are five core latent variables concerned in this study: AI application, self-efficacy, innovation ability, willingness to change, and work performance. The analytical results are shown in Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e. According to the results of data analysis, the single-factor model showed an inferior fit (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{}^{2}\\)\u003c/span\u003e\u003c/span\u003e/df\u0026thinsp;=\u0026thinsp;3.155\u0026thinsp;\u0026gt;\u0026thinsp;3, CFI\u0026thinsp;=\u0026thinsp;0.792\u0026thinsp;\u0026lt;\u0026thinsp;0.8, RMSEA\u0026thinsp;=\u0026thinsp;0.085\u0026thinsp;\u0026gt;\u0026thinsp;0.08) compared to the five-factor model (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{}^{2}\\)\u003c/span\u003e\u003c/span\u003e/df\u0026thinsp;=\u0026thinsp;1.896\u0026thinsp;\u0026lt;\u0026thinsp;3, CFI\u0026thinsp;=\u0026thinsp;0.918\u0026thinsp;\u0026gt;\u0026thinsp;0.8, RMSEA\u0026thinsp;=\u0026thinsp;0.055\u0026thinsp;\u0026lt;\u0026thinsp;0.08). To further prevent the single-factor comparison bias, we constructed three-factor (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{}^{2}\\)\u003c/span\u003e\u003c/span\u003e/df\u0026thinsp;=\u0026thinsp;2.557, CFI\u0026thinsp;=\u0026thinsp;0.852, RMSEA\u0026thinsp;=\u0026thinsp;0.073) and four-factor models (\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{}^{2}\\)\u003c/span\u003e\u003c/span\u003e/df\u0026thinsp;=\u0026thinsp;2.171, CFI\u0026thinsp;=\u0026thinsp;0.890, RMSEA\u0026thinsp;=\u0026thinsp;0.063; \u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{}^{2}\\)\u003c/span\u003e\u003c/span\u003e/df\u0026thinsp;=\u0026thinsp;2.361, CFI\u0026thinsp;=\u0026thinsp;0.872, RMSEA\u0026thinsp;=\u0026thinsp;0.068) sequentially, which did not significantly enhance fit indices. These analytical outcomes demonstrate satisfactory discriminant validity among variables, justifying further analyses.\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n\u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n\u003ctable id=\"Tab1\" border=\"1\"\u003e\u003ccaption\u003e\n\u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e\n\u003cdiv class=\"CaptionContent\"\u003e\n\u003cp\u003eCommon method deviation test(N\u0026thinsp;=\u0026thinsp;296)\u003c/p\u003e\n\u003c/div\u003e\n\u003c/caption\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eBernbach\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eVariable\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eFactor1\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eUniqueness\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eCronbach\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eVariable\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eFactor1\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eUniqueness\u003c/p\u003e\n\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd rowspan=\"6\" align=\"left\"\u003e\n\u003cp\u003e\u0026alpha;\u0026thinsp;=\u0026thinsp;0.6719\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAIP1\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.4555\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.7925\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"3\" align=\"char\" char=\".\"\u003e\n\u003cp\u003e\u0026alpha;\u0026thinsp;=\u0026thinsp;0.7298\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eWE1\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.3255\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.8941\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAIP2\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.2774\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.9231\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eWE2\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.4417\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.8049\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAIP3\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.4213\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.8225\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eWE3\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.5185\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.7311\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAIP4\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.5103\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.7396\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"4\" align=\"char\" char=\".\"\u003e\n\u003cp\u003e\u0026alpha;\u0026thinsp;=\u0026thinsp;0.7528\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eIR1\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.6233\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.6115\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAIP5\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.6427\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.5870\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eIR2\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.5152\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.7346\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAIP6\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.3711\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.8623\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eIR3\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.5427\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.7055\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd rowspan=\"5\" align=\"left\"\u003e\n\u003cp\u003e\u0026alpha;\u0026thinsp;=\u0026thinsp;0.7877\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eWP1\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.5031\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.7469\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eIR4\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.6061\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.6326\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eWP2\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.5094\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.7405\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"4\" align=\"char\" char=\".\"\u003e\n\u003cp\u003e\u0026alpha;\u0026thinsp;=\u0026thinsp;0.8510\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eIA1\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.5392\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.7092\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eWP3\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.6698\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.5514\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eIA2\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.7395\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.4532\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eWP4\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.5630\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.6831\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eIA3\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.7184\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.4840\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eWP5\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.6334\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.5988\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eIA4\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.7350\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.4598\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\n\u003c/div\u003e\n\u003cdiv class=\"gridtable\"\u003e\n\u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n\u003cdiv class=\"colspec\" align=\"char\"\u003e\u0026nbsp;\u003c/div\u003e\n\u003ctable id=\"Tab2\" border=\"1\"\u003e\u003ccaption\u003e\n\u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e\n\u003cdiv class=\"CaptionContent\"\u003e\n\u003cp\u003eConfirmatory factor analysis(N\u0026thinsp;=\u0026thinsp;296)\u003c/p\u003e\n\u003c/div\u003e\n\u003c/caption\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eModel\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eModel\u003c/p\u003e\n\u003cp\u003eDescription\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{}^{2}\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003edf\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003e\u003cspan class=\"InlineEquation\"\u003e\u003cspan class=\"mathinline\"\u003e\\(\\:{\\varDelta\\:}^{2}(\\varDelta\\:\\varvec{d}\\varvec{f})\\)\u003c/span\u003e\u003c/span\u003e\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eCFI\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eTLI\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eRMSEA\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eSRMR\u003c/p\u003e\n\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e1\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eFive-factor\u003c/p\u003e\n\u003cp\u003eModel\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e377.222\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e199\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.918\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.905\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.055\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.050\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e2\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eFour-factor\u003c/p\u003e\n\u003cp\u003eModel\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e440.797\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e203\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e63.575\u003c/p\u003e\n\u003cp\u003e(4)***\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.890\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.875\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.063\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.062\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e3\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eFour-factor\u003c/p\u003e\n\u003cp\u003eModel\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e479.352\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e203\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e102.13\u003c/p\u003e\n\u003cp\u003e(4)***\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.872\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.855\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.068\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.057\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e4\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eThree-factor\u003c/p\u003e\n\u003cp\u003eModel\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e526.702\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e206\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e149.48\u003c/p\u003e\n\u003cp\u003e(7)***\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.852\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.834\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.073\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.059\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e5\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eSingle-factor\u003c/p\u003e\n\u003cp\u003eModel\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e659.374\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e209\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e282.152\u003c/p\u003e\n\u003cp\u003e(10)***\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.792\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.770\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.085\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.066\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\n\u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv class=\"Section2\"\u003e\u0026nbsp;\u003c/div\u003e\n\u003cdiv class=\"Section2\"\u003e1. It is a hypothetical model. 2. Self-efficacy and innovation ability are combined as one factor. 3. Willingness to change and work performance are combined as one factor. 4. Self-efficacy, innovation ability, and willingness to change are combined as one factor. 5. All variables are combined as one factor. ∆\u0026chi;2 test is relative to Model 6.*p\u0026thinsp;\u0026lt;\u0026thinsp;0.05. **p\u0026thinsp;\u0026lt;\u0026thinsp;0.01. ***p\u0026thinsp;\u0026lt;\u0026thinsp;0.001.\u003c/div\u003e\n\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\n\u003ch2\u003e4.3 Descriptive statistics and correlation analysis\u003c/h2\u003e\n\u003cp\u003eMeans, standard deviations, and correlations between the research variables are presented in Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e. It can be seen from Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e that the means for AI application, self-efficacy, and innovation ability are 3.992, 4.279, and 3.814. These results indicate that civil servants show high levels of self-efficacy and creative cognition in the context of the wide application of AI technology. Further calculation based on item count shows approximate scoring rates of 79.84%, 85.59%, and 76.28% for these three primary latent variables, confirming most respondents\u0026rsquo; acknowledgment of the positive impacts of AI applications on work performance. The correlation analysis further indicates significant positive correlations between AI application and self-efficacy (r\u0026thinsp;=\u0026thinsp;0.383, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), innovation ability (r\u0026thinsp;=\u0026thinsp;0.566, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), and work performance (r\u0026thinsp;=\u0026thinsp;0.576, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), preliminarily supporting Hypotheses H1, H2a, and H2b. Additionally, significant positive correlations were found between self-efficacy and innovation ability (r\u0026thinsp;=\u0026thinsp;0.428, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Besides, these two variables strongly correlated with work performance (r\u0026thinsp;=\u0026thinsp;0.399, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001\u003csup\u003e1\u003c/sup\u003e; r\u0026thinsp;=\u0026thinsp;0.590, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001\u003csup\u003e2\u003c/sup\u003e), preliminarily supporting Hypotheses H3, H4, and H5. Furthermore, significant correlations were observed between willingness to change and AI application (r\u0026thinsp;=\u0026thinsp;0.515, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), self-efficacy (r\u0026thinsp;=\u0026thinsp;0.495, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), and innovation ability (r\u0026thinsp;=\u0026thinsp;0.554, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). These findings indicate the likelihood that willingness to change significantly influences various dimensions, including AI application, self-efficacy development, and individual innovation enhancement.\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n\u003ctable id=\"Tab3\" border=\"1\"\u003e\u003ccaption\u003e\n\u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e\n\u003cdiv class=\"CaptionContent\"\u003e\n\u003cp\u003eDescriptive statistics and correlation matrix of variables (N\u0026thinsp;=\u0026thinsp;296)\u003c/p\u003e\n\u003c/div\u003e\n\u003c/caption\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eM\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003eSD\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003e1\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003e2\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003e3\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003e4\u003c/p\u003e\n\u003c/th\u003e\n\u003cth align=\"left\"\u003e\n\u003cp\u003e5\u003c/p\u003e\n\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e1. AI application\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e3.992\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.475\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.000\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e2. Self-efficacy\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e4.279\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.446\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.3837***\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.000\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e3. Innovation ability\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e3.814\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.737\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.5659***\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.4283***\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.000\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e4. Willingness to change\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e3.922\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.651\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.5146***\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.4954***\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.5543***\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.000\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\u0026nbsp;\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e5. Work performance\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e4.039\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.540\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.5762***\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.3988***\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.5899***\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.5138***\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e1.000\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\n\u003c/div\u003e\n\u003c/div\u003e\n\u003cp\u003e1. M and SD are used to represent mean and standard deviation, respectively.\u003c/p\u003e\n\u003cp\u003e2. *p\u0026thinsp;\u0026lt;\u0026thinsp;0.05. **p\u0026thinsp;\u0026lt;\u0026thinsp;0.01. ***p\u0026thinsp;\u0026lt;\u0026thinsp;0.001. (dup: 6 ?)\u003c/p\u003e\n\u003cdiv id=\"Sec19\" class=\"Section2\"\u003e\n\u003ch2\u003e4.4 Chain Mediation Model Test\u003c/h2\u003e\n\u003cp\u003eStata 16 was used in this study to verify Hypotheses H1, H2a, H2b, H3, and H4 by multiple linear regression analyses. As Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e and Fig.\u0026nbsp;2 illustrate, after incorporating all control variables, AI application showed a significant positive correlation with civil servants' work performance (Model 1, \u0026beta;\u0026thinsp;=\u0026thinsp;0.6473, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), supporting Hypothesis H1. This result indicates that, despite the learning cost and crowding-out effects of AI technology, its extensive application significantly enhances individual performance and output among civil servants.\u003c/p\u003e\n\u003cp\u003eThen, following Baron and Kenny\u0026rsquo;s (\u003cspan class=\"CitationRef\"\u003e1986\u003c/span\u003e) mediation test procedure, we evaluated the mediating roles of self-efficacy and Innovation Ability, the results of which are shown in Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e. After incorporating all control variables, we found AI application positively correlated with individual Self-efficacy (Model 2a, \u0026beta;\u0026thinsp;=\u0026thinsp;0.3453, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). With control applied to AI application, self-efficacy showed a significant positive correlation with work performance (Model 2b, \u0026beta;\u0026thinsp;=\u0026thinsp;0.2550, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), thus validating Hypotheses H2a and H3. Moreover, as Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e4\u003c/span\u003e further indicates, AI application is positively associated with innovation ability (Model 3a, \u0026beta;\u0026thinsp;=\u0026thinsp;0.8344, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). With AI application controlled, innovation ability correlates positively with work performance (Model 3b, \u0026beta;\u0026thinsp;=\u0026thinsp;0.2845, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), thereby validating Hypotheses H2b and H4. This result also confirms the view of the Stimulus-Organism-Response, which suggests that external stimulus induces a behavioral response by influencing an individual\u0026rsquo;s internal mental state. AI application significantly enhance civil servants\u0026rsquo; self-efficacy and innovation ability, which in turn enable them to perform more effectively in their work, thereby leading to better work performance.\u003c/p\u003e\n\u003cp\u003eSubsequently, we tested the chain mediation model of AI application \u0026rarr; self-efficacy \u0026rarr; innovation ability \u0026rarr; work performance, using Stata 16 as well. After controlling AI application, self-efficacy exhibited a significant association with innovation ability (Model 4a, \u0026beta;\u0026thinsp;=\u0026thinsp;0.3991, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Furthermore, when all three core latent variables\u0026mdash;AI application, self-efficacy, and innovation ability\u0026mdash;were simultaneously included in the regression model, each demonstrated a significant positive relationship with work performance (Model 4b, \u0026beta;\u0026thinsp;=\u0026thinsp;0.3812, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; \u0026beta;\u0026thinsp;=\u0026thinsp;0.1530, p\u0026thinsp;=\u0026thinsp;0.011; \u0026beta;\u0026thinsp;=\u0026thinsp;0.2556, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), thereby validating Hypothesis H5.\u003c/p\u003e\n\u003cp\u003eFinally, following the approach proposed by Hayes (2017), we conducted a multi-step mediation analysis using the bootstrap method. As shown in Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e5\u003c/span\u003e and Fig.\u0026nbsp;2, the indirect effect of AI application on work performance through self-efficacy is 0.0881, and 95%CI = [0.03585, 0.14029], excluding 0. This result indicates that civil servants can significantly enhance their self-efficacy through extensive application of AI technologies, which in turn improves their work performance. Therefore, Hypothesis H3 is supported. Moreover, the indirect effect of AI application on work performance via innovation ability is 0.2374, and 95%CI = [0.15035, 0.32441], excluding zero. This result confirms a significant mediating role of innovation ability, thereby providing further support for Hypothesis H4. To be further, the indirect effect of AI application on work performance through self-efficacy and innovation ability is 0.0352, and 95%CI = [0.01186, 0.05858], excluding zero. This result confirms a significant chain mediation effect among the four core latent variables, offering robust support for Hypothesis H5. This finding demonstrates that the application of AI technologies significantly enhances civil servants\u0026rsquo; self-efficacy, which in turn promotes their innovation ability, thereby ultimately improving their work performance.\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n\u003ctable id=\"Tab4\" border=\"1\"\u003e\u003ccaption\u003e\n\u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e\n\u003cdiv class=\"CaptionContent\"\u003e\n\u003cp\u003eRegression analysis of AI application, self-efficacy, innovation ability, and work performance(N\u0026thinsp;=\u0026thinsp;296)\u003c/p\u003e\n\u003c/div\u003e\n\u003c/caption\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n\u003cth colspan=\"2\" align=\"left\"\u003e\n\u003cp\u003eVariables\u003c/p\u003e\n\u003c/th\u003e\n\u003cth colspan=\"3\" align=\"left\"\u003e\n\u003cp\u003eRegression coefficients\u003c/p\u003e\n\u003c/th\u003e\n\u003cth colspan=\"3\" align=\"left\"\u003e\n\u003cp\u003eFits the index\u003c/p\u003e\n\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eModel\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eDV\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eIV\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u0026beta;\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003et\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003ep\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eR2\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAR2\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eF\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eModel 1\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eWork performance\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAI application\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.6473\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e11.67\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e0.3477\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e0.3295\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e19.12***\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eModel 2a\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eSelf-efficacy\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAI application\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.3453\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e6.66\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e0.1632\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e0.1399\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e7.00***\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd rowspan=\"2\" align=\"left\"\u003e\n\u003cp\u003eModel 2b\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"2\" align=\"left\"\u003e\n\u003cp\u003eWork performance\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAI application\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.5592\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e9.65\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"2\" align=\"left\"\u003e\n\u003cp\u003e0.3848\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"2\" align=\"left\"\u003e\n\u003cp\u003e0.3654\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"2\" align=\"left\"\u003e\n\u003cp\u003e19.88***\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eSelf-efficacy\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.2550\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e4.15\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eModel 3a\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eInnovation ability\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAI application\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.8344\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e11.07\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e0.3525\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e0.3344\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e19.53***\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd rowspan=\"2\" align=\"left\"\u003e\n\u003cp\u003eModel 3b\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"2\" align=\"left\"\u003e\n\u003cp\u003eWork performance\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAI application\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.4099\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e6.70\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"2\" align=\"left\"\u003e\n\u003cp\u003e0.4452\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"2\" align=\"left\"\u003e\n\u003cp\u003e0.4277\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"2\" align=\"left\"\u003e\n\u003cp\u003e25.50***\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eInnovation Ability\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.2845\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e7.09\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd rowspan=\"2\" align=\"left\"\u003e\n\u003cp\u003eModel 4a\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"2\" align=\"left\"\u003e\n\u003cp\u003eInnovation ability\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAI application\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.6966\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e8.93\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"2\" align=\"left\"\u003e\n\u003cp\u003e0.4013\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"2\" align=\"left\"\u003e\n\u003cp\u003e0.3825\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"2\" align=\"left\"\u003e\n\u003cp\u003e21.30***\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eSelf-efficacy\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.3991\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e4.83\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd rowspan=\"3\" align=\"left\"\u003e\n\u003cp\u003eModel 4b\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"3\" align=\"left\"\u003e\n\u003cp\u003eWork performance\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAI application\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.3812\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e6.18\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"3\" align=\"left\"\u003e\n\u003cp\u003e0.4576\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"3\" align=\"left\"\u003e\n\u003cp\u003e0.4385\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"3\" align=\"left\"\u003e\n\u003cp\u003e24.04***\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eSelf-efficacy\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.1530\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e2.55\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e0.011\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eInnovation Ability\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.2556\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e6.18\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\n1. Control variables were included in all regression models. DV, Dependent Variable; IV, Independent Variable.\u003c/div\u003e\n\u003c/div\u003e\n\u003cp\u003e2. *p\u0026thinsp;\u0026lt;\u0026thinsp;0.05. **p\u0026thinsp;\u0026lt;\u0026thinsp;0.01. ***p\u0026thinsp;\u0026lt;\u0026thinsp;0.001. (dup: 8 ?)\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n\u003ctable id=\"Tab5\" border=\"1\"\u003e\u003ccaption\u003e\n\u003cdiv class=\"CaptionNumber\"\u003eTable 5\u003c/div\u003e\n\u003cdiv class=\"CaptionContent\"\u003e\n\u003cp\u003e\u003cstrong\u003eMediation effect test(N\u0026thinsp;=\u0026thinsp;296)\u003c/strong\u003e Control variables were included in all regression models.\u003c/p\u003e\n\u003c/div\u003e\n\u003c/caption\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd width=\"110\"\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd colspan=\"4\" width=\"441\"\u003e\n\u003cp\u003e95% CI\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd width=\"110\"\u003e\n\u003cp\u003eEffect\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"110\"\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003e\u0026beta;\u003c/em\u003e\u003c/strong\u003e (standardized\u0026nbsp; path coefficient)\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"110\"\u003e\n\u003cp\u003e\u003cem\u003eSE\u003c/em\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"110\"\u003e\n\u003cp\u003e\u003cem\u003eLL\u003c/em\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"110\"\u003e\n\u003cp\u003e\u003cem\u003eUL\u003c/em\u003e\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd width=\"110\"\u003e\n\u003cp\u003eX\u0026rarr;M1\u0026rarr;Y\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"110\"\u003e\n\u003cp\u003e0.0881\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"110\"\u003e\n\u003cp\u003e0.0266\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"110\"\u003e\n\u003cp\u003e0.0359\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"110\"\u003e\n\u003cp\u003e0.1402\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd width=\"110\"\u003e\n\u003cp\u003eX\u0026rarr;M2\u0026rarr;Y\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"110\"\u003e\n\u003cp\u003e0.2374\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"110\"\u003e\n\u003cp\u003e0.0444\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"110\"\u003e\n\u003cp\u003e0.1504\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"110\"\u003e\n\u003cp\u003e0.3244\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd width=\"110\"\u003e\n\u003cp\u003eX\u0026rarr;M1\u0026rarr;M2\u0026rarr;Y\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"110\"\u003e\n\u003cp\u003e0.0352\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"110\"\u003e\n\u003cp\u003e0.0119\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"110\"\u003e\n\u003cp\u003e0.0119\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd width=\"110\"\u003e\n\u003cp\u003e0.0586\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\n\u003c/div\u003e\n\u003cdiv class=\"gridtable\"\u003e\n\u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec22\" class=\"Section2\"\u003e\n\u003ch2\u003e4.5 Moderation Effect Test\u003c/h2\u003e\n\u003cp\u003eTable\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003e delineates that the interplay between willingness to change and AI application robustly fosters innovation ability(Model 1, \u0026beta;\u0026thinsp;=\u0026thinsp;0.316, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Referencing Fig.\u0026nbsp;3, the interaction effects and simple slopes suggest that at subdued levels of willingness to change, AI application imparts a significant and positive influence on innovation ability(simple slope\u0026thinsp;=\u0026thinsp;1.111, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Conversely, at elevated willingness to change levels, this influence markedly intensifies (simple slope\u0026thinsp;=\u0026thinsp;0.419, p\u0026thinsp;=\u0026thinsp;0.012), corroborating that willingness to change acts as an effective moderator within the nexus of AI application and innovation ability, thereby validating Hypothesis H6a.\u003c/p\u003e\n\u003cp\u003eSimilarly, the results shown in Table\u0026nbsp;\u003cspan class=\"InternalRef\"\u003e6\u003c/span\u003e also demonstrate that the synergy between willingness to change and self-efficacy significantly propels innovation ability(Model 2, \u0026beta;\u0026thinsp;=\u0026thinsp;0.269, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). Furthermore, a simple slope analysis is conducted on the above moderating effects, and the results are shown in Fig.\u0026nbsp;4. Under a lower level of willingness to change, self-efficacy has a mild promoting effect on the improvement of innovation ability (simple slope\u0026thinsp;=\u0026thinsp;0.420, p\u0026thinsp;=\u0026thinsp;0.019). Whereas, with the increase of willingness to change, the promoting effect of self-efficacy on the innovation ability is significantly increased(simple slope\u0026thinsp;=\u0026thinsp;0.899, p\u0026thinsp;=\u0026thinsp;0.01), confirming that willingness to change constructively regulates the interrelation between self-efficacy and innovation ability, hence affirming Hypothesis H6b.\u0026nbsp;\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n\u003ctable id=\"Tab6\" border=\"1\"\u003e\u003ccaption\u003e\n\u003cdiv class=\"CaptionNumber\"\u003eTable 6\u003c/div\u003e\n\u003cdiv class=\"CaptionContent\"\u003e\n\u003cp\u003eModeration effects test(N\u0026thinsp;=\u0026thinsp;296)\u003c/p\u003e\n\u003c/div\u003e\n\u003c/caption\u003e\n\u003cthead\u003e\n\u003ctr\u003e\n\u003cth align=\"left\"\u003e\u0026nbsp;\u003c/th\u003e\n\u003cth colspan=\"2\" align=\"left\"\u003e\n\u003cp\u003eVariables\u003c/p\u003e\n\u003c/th\u003e\n\u003cth colspan=\"3\" align=\"left\"\u003e\n\u003cp\u003eRegression coefficients\u003c/p\u003e\n\u003c/th\u003e\n\u003cth colspan=\"3\" align=\"left\"\u003e\n\u003cp\u003eFits the index\u003c/p\u003e\n\u003c/th\u003e\n\u003c/tr\u003e\n\u003c/thead\u003e\n\u003ctbody\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eModel\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eDV\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eIV\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u0026beta;\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003et\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003ep\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eR2\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAR2\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eF\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd rowspan=\"3\" align=\"left\"\u003e\n\u003cp\u003eModel 1\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"3\" align=\"left\"\u003e\n\u003cp\u003eInnovation ability\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAI application\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.618\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e7.73\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"3\" align=\"left\"\u003e\n\u003cp\u003e0.4344\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"3\" align=\"left\"\u003e\n\u003cp\u003e0.4286\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"3\" align=\"left\"\u003e\n\u003cp\u003e74.76***\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eWillingness to change\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.453\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e7.55\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eAI application\u003c/p\u003e\n\u003cp\u003e\u0026times;willingness to change\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.316\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e3.22\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd rowspan=\"3\" align=\"left\"\u003e\n\u003cp\u003eModel 2\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"3\" align=\"left\"\u003e\n\u003cp\u003eInnovation ability\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eSelf-efficacy\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.443\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e4.51\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"3\" align=\"left\"\u003e\n\u003cp\u003e0.3538\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"3\" align=\"left\"\u003e\n\u003cp\u003e0.3471\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd rowspan=\"3\" align=\"left\"\u003e\n\u003cp\u003e53.29***\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eWillingness to change\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.536\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e8.65\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003ctr\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003eSelf-efficacy\u003c/p\u003e\n\u003cp\u003e\u0026times;willingness to change\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"char\" char=\".\"\u003e\n\u003cp\u003e0.269\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e2.62\u003c/p\u003e\n\u003c/td\u003e\n\u003ctd align=\"left\"\u003e\n\u003cp\u003e\u0026lt;\u0026thinsp;0.001\u003c/p\u003e\n\u003c/td\u003e\n\u003c/tr\u003e\n\u003c/tbody\u003e\n\u003c/table\u003e\n1. Control variables were included in all regression models.\u003c/div\u003e\n\u003cdiv class=\"gridtable\"\u003e2. *p\u0026thinsp;\u0026lt;\u0026thinsp;0.05. **p\u0026thinsp;\u0026lt;\u0026thinsp;0.01. ***p\u0026thinsp;\u0026lt;\u0026thinsp;0.001.\u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv class=\"gridtable\"\u003e\n\u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n\u003c/div\u003e"},{"header":"5 Conclusion and discussion","content":"\u003cdiv id=\"Sec26\" class=\"Section2\"\u003e\u003ch2\u003e5.1 Study conclusion\u003c/h2\u003e\u003cp\u003eOur study provides valuable insights into AI application and its impact on civil servants\u0026rsquo; work performance. Based on the structural equation model, chain mediation model, moderation model, and bootstrap method of questionnaire data from 296 civil servants, this study found strong evidence of AI\u0026rsquo;s positive influence on work performance. Firstly, AI application had a significant direct effect on work performance (β\u0026thinsp;=\u0026thinsp;0.6473, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). This result indicated that the extensive use of AI can effectively improve the ability of civil servants in information acquisition, document processing, and decision-making, thus improving their work performance. Secondly, self-efficacy plays a mediating role between AI application and work performance (β\u0026thinsp;=\u0026thinsp;0.2550, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), as well as innovation ability notably mediated the relationship between AI application and work performance (β\u0026thinsp;=\u0026thinsp;0.2845, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). This confirmed that self-efficacy and innovation ability are important transmission mechanisms of AI application, playing a notable role in bridging the gap between AI application and work performance. Thirdly, this study further found that self-efficacy and innovation ability play a chain mediating role between AI application and work performance of civil servants (β\u0026thinsp;=\u0026thinsp;0.1530, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001; β\u0026thinsp;=\u0026thinsp;0.2556, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001). This conclusion suggests that AI application elevate self-efficacy of civil servants, and then further improves their innovation ability, which in turn promotes their work performance. Fourthly, willingness to change emerged as a positive moderator, refining the relationship between the AI application and innovation ability (β\u0026thinsp;=\u0026thinsp;0.316, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), underscoring the paramount importance of acceptance willingness and action tendency of new technology in fostering effectiveness of AI application and improving innovation ability. Besides, willingness to change also amplified the promoting effect of self-efficacy on innovation ability (β\u0026thinsp;=\u0026thinsp;0.269, p\u0026thinsp;\u0026lt;\u0026thinsp;0.001), emphasizing the transformative power of change orientation on the relationship between psychological perception and innovative practice.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec27\" class=\"Section2\"\u003e\u003ch2\u003e5.2 Theoretical and practical implications\u003c/h2\u003e\u003cdiv id=\"Sec28\" class=\"Section3\"\u003e\u003ch2\u003e5.2.1 Theoretical\u003c/h2\u003e\u003cp\u003eBased on the preceding theoretical framework and empirical analysis, this study examined the relationship between AI application and civil servants\u0026rsquo; Work Performance. In particular, it explored the mediating roles of self-efficacy and innovation ability, as well as the moderating role of willingness to change. The main theoretical contributions of this study can be summarized as follows.\u003c/p\u003e\u003cp\u003eFirst, this study focuses on the actual usage behavior of AI technologies among civil servants, introducing the TAM into the research context of the public sector, thereby extending the model\u0026rsquo;s scope of applicability. Compared with the original TAM proposed, which has been primarily applied to analyze technology adoption behaviors among corporate employees (Park et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2014\u003c/span\u003e) and educators (Teo, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2009\u003c/span\u003e), this study applied the TAM framework to the public sector\u0026mdash;a context marked by high institutional rigidity and significant heterogeneity in technological adaptability\u0026mdash;namely, civil servants. This approach responds to the practical need for research on technology adoption mechanisms in public organizations under the current \u0026ldquo;digital government\u0026rdquo; transformation. By incorporating key mediating variables\u0026mdash;namely self-efficacy and innovation ability\u0026mdash;this study further uncovers the cognitive and capability transformation processes that underlie AI application. It enriches the explanatory power of the TAM framework in mapping the process of \u0026ldquo;perception-intention-behavior-performance\u0026rdquo;, broadening the model\u0026rsquo;s theoretical boundary and providing theoretical support for understanding individual technology adoption within complex organizational environments.\u003c/p\u003e\u003cp\u003eSecondly, this study constructs and empirically validates a chain mediation model that integrates Bandura\u0026rsquo;s self-efficacy theory with organizational innovation theory to examine how AI application influences civil servants\u0026rsquo; work performance through the sequential mediators of self-efficacy and innovation ability. This study agrees with some researchers that self-efficacy has a positive impact on an individual\u0026rsquo;s ability to innovate (Tierney \u0026amp; Farmer, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2011\u003c/span\u003e; Yu, \u003cspan citationid=\"CR49\" class=\"CitationRef\"\u003e2013\u003c/span\u003e). Building on this, this study further demonstrates that, in the context of digital government, the application of AI can effectively enhance civil servants\u0026rsquo; self-efficacy, which in turn promotes their innovation ability and ultimately improves their work performance. Our findings contribute a psychological perspective to understanding the behavioral mechanisms of civil servants in the context of digital transformation.\u003c/p\u003e\u003cp\u003eThirdly, this study takes an innovative approach by measuring willingness to change from both individual and environmental dimensions, introducing it as a moderating variable. Traditional research based on the TAM has primarily focused on the impact of technological attributes on individual willingness to use. Such studies assume a relatively uniform adaptive tendency among users when facing new technologies, often overlooking potential differences in individuals\u0026rsquo; cognitive readiness and psychological receptiveness (Legris et al., \u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e2003\u003c/span\u003e; Venkatesh \u0026amp; Davis, \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2000\u003c/span\u003e). By fully considering the individual heterogeneity in willingness to change, this study empirically confirms its moderating effects on two critical paths: from AI application to innovation ability, and from self-efficacy to innovation ability. These findings offer a beneficial extension of the TAM by incorporating psychological and contextual variability into the analysis of technology adoption.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec29\" class=\"Section3\"\u003e\u003ch2\u003e5.2.2 Management implications\u003c/h2\u003e\u003cp\u003eThis study reveals the multiple pathways through which AI influences civil servants\u0026rsquo; work performance, particularly highlighting its role in enhancing self-efficacy and innovation ability, as well as the moderating function of willingness to change. Based on these findings, this paper offers the following three recommendations for the future transformation of public administration:\u003c/p\u003e\u003cp\u003eFirstly, AI competency development should be integrated into the institutional training system for civil servants. With the deepening application of AI technologies in public service delivery, traditional competency frameworks for civil servants are no longer adequate to meet the demands of intelligent governance. It is recommended that digital competencies, such as AI literacy, be integrated into civil service career development frameworks. Structured training programs should be implemented to improve civil servants\u0026rsquo; technological comprehension and application skills, enabling them to competently engage in data governance, algorithmic management, and intelligent decision-making within a digital government environment. This shift will support the transformation of civil servants from \u0026ldquo;technology users\u0026rdquo; to \u0026ldquo;intelligent governance actors\u0026rdquo;.\u003c/p\u003e\u003cp\u003eSecondly, innovation-oriented organizational culture and institutional environment should be fostered. Innovation ability, as the key mediating variable connecting technology adoption and performance improvement, requires strong organizational and institutional support to be fully realized. Public organizations should transition from an \u0026ldquo;execution-oriented\u0026rdquo; to an \u0026ldquo;innovation-driven\u0026rdquo; model. Institutional design should increase tolerance for trial-and-error and promote workflow optimization to create an exploratory and supportive work atmosphere that encourages civil servants to adapt and innovate in diverse scenarios. Additionally, organizational innovation capacity should be incorporated into performance evaluation and cadre assessment systems. A structured incentive and accountability framework can help cultivate sustained innovation and shift governance paradigms from reactive adaptation to proactive innovation.\u003c/p\u003e\u003cp\u003eThirdly, greater emphasis should be placed on stimulating individual motivation for change and cultivating a supportive atmosphere for technology adoption. The findings suggest that willingness to change significantly moderates the impact of AI use on innovation ability, indicating that the effectiveness of technology adoption depends not only on the functional advantages of AI but also on users\u0026rsquo; psychological perceptions and value alignment. Therefore, activating civil servants\u0026rsquo; intrinsic motivation for change is essential for advancing intelligent governance. On the one hand, public organizations should establish robust support systems for AI adoption to boost civil servants\u0026rsquo; confidence and sense of psychological safety in the face of technological change. On the other hand, they should build inclusive and participatory organizational mechanisms that empower civil servants to take ownership in the transformation process, thereby promoting deeper integration between technological and organizational change.\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e\u003cdiv id=\"Sec30\" class=\"Section2\"\u003e\u003ch2\u003e5.3 Limitations and directions for future research\u003c/h2\u003e\u003cp\u003eAlthough this study offers a systematic examination of how AI affects the work performance of civil servants, it is subject to several limitations. Firstly, the sample primarily comprises grass-root civil servants, whose administrative ranks are mainly concentrated at the staff and section levels. As the operational core of China\u0026rsquo;s public administration, this group is highly involved in the execution of routine governmental functions and public affairs, thus holding a certain degree of representativeness. However, the absence of mid-level and high-level civil servants (e.g., department-level and above) constrains the study\u0026rsquo;s capacity to capture the heterogeneous characteristics of AI application among leadership in China's public sector. Future research should consider incorporating a broader range of administrative ranks to allow for hierarchical comparisons in AI adoption behaviors. Secondly, all empirical data were collected in China. Given China\u0026rsquo;s distinctive institutional settings and administrative culture, the findings may not be readily generalizable to other national contexts. Subsequent studies are encouraged to adopt a cross-national perspective by conducting comparative analyses across different countries\u003c/p\u003e\u003c/div\u003e"},{"header":"Declarations","content":"\u003cp\u003eThe data that support the findings of this study are available on request from the corresponding author.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe Ethics Committee of the South China University of Technology provided ethical approval for this study in strict accordance with the guidelines of the Declaration of Helsinki. All participants were informed of the nature and purpose of the study and provided their informed consent prior to participation.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNo applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of data and material\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets generated and analyzed during the current study are not publicly available due to participant confidentiality and institutional data protection policies but are available from the corresponding author on reasonable request.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNo applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors' contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eL.Z contributed to conceptualization, research design, data processing, empirical analysis, software implementation, figure and table preparation, writing-original draft, and writing review \u0026amp; editing. G.D contributed to survey design, data collection, writing-original draft, and writing review \u0026amp; editing. All authors ead and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNo applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors' information\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003csup\u003e1\u003c/sup\u003eZongyang Li, South China University of Technology, School of Public Administration, Guangzhou, Guangdong 510641, People\u0026rsquo;s Republic of China.\u003c/p\u003e\n\u003cp\u003e\u003csup\u003e2\u003c/sup\u003eDongming Gu, South China University of Technology, School of Public Administration, Guangzhou, Guangdong 510641, People\u0026rsquo;s Republic of China.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAbioye SO, Oyedele LO, Akanbi L, Ajayi A, Delgado JMD, Bilal M, Akinade OO, Ahmed A. Artificial intelligence in the construction industry: A review of present status, opportunities and future challenges. J Building Eng. 2021;44:103299. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1016/j.jobe.2021.103299\u003c/span\u003e\u003cspan address=\"doi:10.1016/j.jobe.2021.103299\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBallet K, Kelchtermans G. Workload and willingness to change: Disentangling the experience of intensification. J curriculum Stud. 2008;40(1):47\u0026ndash;67. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1080/00220270701516463\u003c/span\u003e\u003cspan address=\"doi:10.1080/00220270701516463\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBandura A. Self-efficacy mechanism in human agency. Am Psychol. 1982;37(2):122. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1037/0003-066x.37.2.122\u003c/span\u003e\u003cspan address=\"doi:10.1037/0003-066x.37.2.122\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBandura A. Self-efficacy: The foundation of Agency1. Control of human behavior, mental processes, and consciousness. Psychology; 2013. pp. 16\u0026ndash;30.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBaron RM, Kenny DA. The moderator\u0026ndash;mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. J Personal Soc Psychol. 1986;51(6):1173. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1037/0022-3514.51.6.1173\u003c/span\u003e\u003cspan address=\"doi:10.1037/0022-3514.51.6.1173\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBenson T. (2019). Digital innovation evaluation: user perceptions of innovation readiness, digital confidence, innovation adoption, user experience and behaviour change. BMJ health care Inf, 26(1), e000018.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBilan Y, Mishchuk H, Samoliuk N. Digital skills of civil servants: Assessing readiness for successful interaction in e-society. Acta Polytech Hungarica. 2023;20(3):155\u0026ndash;74.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eBolin JH. Introduction to mediation, moderation, and conditional process analysis: a regression-based approach. In: JSTOR; 2014.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCaianiello M. Dangerous liaisons. Potentialities and risks deriving from the interaction between artificial intelligence and preventive justice. Eur J Crime Criminal Law Criminal Justice. 2021;29(1):1\u0026ndash;23. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1163/15718174-29010001\u003c/span\u003e\u003cspan address=\"doi:10.1163/15718174-29010001\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCasalino N, Saso T, Borin B, Massella E, Lancioni F. In: Agrifoglio R, Lamboglia R, Mancini D, Ricciardi F, editors. Digital Competences for Civil Servants and Digital Ecosystems for More Effective Working Processes in Public Organizations. Digital Business Transformation Cham; 2020a.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eCasalino N, Saso T, Borin B, Massella E, Lancioni F. Digital competences for civil servants and digital ecosystems for more effective working processes in public organizations. Digital Business Transformation: Organizing, Managing and Controlling in the Information Age; 2020b.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChen S-C, Shing-Han L, Chien-Yi L. Recent related research in technology acceptance model: A literature review. Australian J Bus Manage Res. 2011;1(9):124.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eChoi M. Employees' attitudes toward organizational change: A literature review. Hum Resour Manag. 2011;50(4):479\u0026ndash;500. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1002/hrm.20434\u003c/span\u003e\u003cspan address=\"doi:10.1002/hrm.20434\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDavis FD, Granić A, Marangunić N. The technology acceptance model: 30 years of TAM. Springer; 2024.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eDoroodian M, Ab Rahman MN, Kamarulzaman Y, Muhamad N. Designing and validating a model for measuring innovation capacity construct. Adv Decis Sci. 2014. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1155/2014/576596\u003c/span\u003e\u003cspan address=\"doi:10.1155/2014/576596\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eGrant D, Jha PP, Wanjiru R, Bhalla A. Evolving willingness and ability interfaces: An innovation led transformation journey. Int J Innov Stud. 2020;4(3):69\u0026ndash;75. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1016/j.ijis.2020.06.001\u003c/span\u003e\u003cspan address=\"doi:10.1016/j.ijis.2020.06.001\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHerold DM, Fedor DB, Caldwell S, Liu Y. The effects of transformational and change leadership on employees' commitment to a change: a multilevel study. J Appl Psychol. 2008;93(2):346. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1037/0021-9010.93.2.346\u003c/span\u003e\u003cspan address=\"doi:10.1037/0021-9010.93.2.346\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHood C. The new public management in the 1980s: Variations on a theme. Acc Organ Soc. 1995;20(2\u0026ndash;3):93\u0026ndash;109. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1016/0361-3682(93)E0001-W\u003c/span\u003e\u003cspan address=\"doi:10.1016/0361-3682(93)E0001-W\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eHwang Y, Wu Y. The influence of generative artificial intelligence on creative cognition of design students: a chain mediation model of self-efficacy and anxiety. Front Psychol. 2025;15:1455015. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.3389/fpsyg.2024.1455015\u003c/span\u003e\u003cspan address=\"doi:10.3389/fpsyg.2024.1455015\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eIaffaldano MT, Muchinsky PM. Job satisfaction and job performance: A meta-analysis. Psychol Bull. 1985;97(2):251. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1037/0033-2909.97.2.251\u003c/span\u003e\u003cspan address=\"doi:10.1037/0033-2909.97.2.251\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eIvić A, Milićević A, Krstić D, Kozma N, Havzi S. (2022). The challenges and opportunities in adopting AI, IoT and blockchain technology in e-government: A systematic literature review. 2022 International Conference on Communications, Information, Electronic and Energy Systems (CIEES).\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJackson PC. Introduction to artificial intelligence. Courier Dover; 2019.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eJiao H, Zhao G. When will employees embrace managers' technological innovations? The mediating effects of employees' perceptions of fairness on their willingness to accept change and its legitimacy. J Prod Innov Manage. 2014;31(4):780\u0026ndash;98. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1111/jpim.12123\u003c/span\u003e\u003cspan address=\"doi:10.1111/jpim.12123\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLartey D, Law KM. Artificial intelligence adoption in urban planning governance: A systematic review of advancements in decision-making, and policy making. Landsc Urban Plann. 2025;258:105337. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1016/j.landurbplan.2025.105337\u003c/span\u003e\u003cspan address=\"doi:10.1016/j.landurbplan.2025.105337\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLaviolette EM, Redien-Collot R, Teglborg A-C. Open innovation from the inside: Employee-driven innovation in support of absorptive capacity for inbound open innovation. Int J Entrepreneurship Innov. 2016;17(4):228\u0026ndash;39. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1177/146575031667049\u003c/span\u003e\u003cspan address=\"doi:10.1177/146575031667049\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLegris P, Ingham J, Collerette P. Why do people use information technology? A critical review of the technology acceptance model. Inf Manag. 2003;40(3):191\u0026ndash;204. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1016/S0378-7206(01)00143-4\u003c/span\u003e\u003cspan address=\"doi:10.1016/S0378-7206(01)00143-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLemke C. (2010). Innovation through technology. 21st century skills: Rethinking how students learn, 243\u0026ndash;272.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eLexer MG, Scarcella L. Artificial intelligence and labor markets. A critical analysis of solution models from a tax law and social security law perspective. Rivista italiana di informatica e diritto. 2019;1(1):53\u0026ndash;73.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMarangunić N, Granić A. Univ Access Inf Soc. 2015;14:81\u0026ndash;95. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1007/s10209-014-0348-1\u003c/span\u003e\u003cspan address=\"doi:10.1007/s10209-014-0348-1\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. Technology acceptance model: a literature review from 1986 to 2013.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMaxwell L, Taner E, Jonathan GM. Digitalisation in the public sector: Determinant factors. Int J IT/Business Alignment Gov (IJITBAG). 2019;10(2):35\u0026ndash;52. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.4018/IJITBAG.2019070103\u003c/span\u003e\u003cspan address=\"doi:10.4018/IJITBAG.2019070103\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMehr H, Ash H, Fellow D. Artificial intelligence for citizen services and government. Ash Cent. Democr. Gov. Innov. Harvard Kennedy Sch. no August. 2017;1:12.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMetselaar EE. Assessing the willingness to change. Construction and validation of the DINAMO; 1997.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eMiller VD, Johnson JR, Grau J. (1994). Antecedents to willingness to participate in a planned organizational change.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eObenza BN, Salvahan A, Rios AN, Solo A, Alburo RA, Gabila RJ. University students' perception and use of ChatGPT: Generative artificial intelligence (AI) in higher education. Int J Hum Comput Stud. 2024;5(12):5\u0026ndash;18.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePark N, Rhoads M, Hou J, Lee KM. Understanding the acceptance of teleconferencing systems among employees: An extension of the technology acceptance model. Comput Hum Behav. 2014;39:118\u0026ndash;27. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1016/j.chb.2014.05.048\u003c/span\u003e\u003cspan address=\"doi:10.1016/j.chb.2014.05.048\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003ePrajogo DI, Ahmed PK. Relationships between innovation stimulus, innovation capacity, and innovation performance. R\u0026amp;d Manage. 2006;36(5):499\u0026ndash;515. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1111/j.1467-9310.2006.00450.x\u003c/span\u003e\u003cspan address=\"doi:10.1111/j.1467-9310.2006.00450.x\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRamadian A, Chairuddin C, Judijanto L, Rachmawati R, Sopandi E, ARTIFICIAL INTELLIGENCE COMPETENCIES, ORGANIZATIONAL SUPPORT, AND EMPLOYEE SELF-EFFICACY IN PREDICTING GOVERNMENT EMPLOYEE PERFORMANCE: A MEDIATION ANALYSIS WITH WORK ENGAGEMENT. Jurnal Manajemen dan Kewirausahaan. 2025;27(1):22\u0026ndash;32. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.9744/jmk.27.1.22-32\u003c/span\u003e\u003cspan address=\"doi:10.9744/jmk.27.1.22-32\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. THE ROLE OF.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eRusho MA, Chan MP. Analysis of Artificial Intelligence and its Impactful Implementation on Job Performance. Dinkum J Econ Managerial Innovations. 2023;2(10):571\u0026ndash;6.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSari RF. (2025). Exploring Artificial Intelligence Adoption in the Public Sector: Perceptions, Attitudes and Factors Influencing AI Adoption Among Government Public Relations Practitioners in Indonesia University of Twente].\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eSeckelmann M, Catakli D. Digital Competencies in the Civil Service. The Civil Service in Europe. Routledge; 2025. pp. 662\u0026ndash;80.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eShahzad MF, Xu S, Naveed W, Nusrat S, Zahid I. Investigating the impact of artificial intelligence on human resource functions in the health sector of China: A mediated moderation model. Heliyon. 2023;9(11). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1016/j.heliyon.2023.e21818\u003c/span\u003e\u003cspan address=\"doi:10.1016/j.heliyon.2023.e21818\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eShank DB, Cotten SR. Does technology empower urban youth? The relationship of technology use to self-efficacy. Comput Educ. 2014;70:184\u0026ndash;93. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1016/j.compedu.2013.08.018\u003c/span\u003e\u003cspan address=\"doi:10.1016/j.compedu.2013.08.018\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eShoji K, Cieslak R, Smoktunowicz E, Rogala A, Benight CC, Luszczynska A. Associations between job burnout and self-efficacy: A meta-analysis. Anxiety Stress Coping. 2016;29(4):367\u0026ndash;86. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1080/10615806.2015.1058369\u003c/span\u003e\u003cspan address=\"doi:10.1080/10615806.2015.1058369\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTeo T. Modelling technology acceptance in education: A study of pre-service teachers. Comput Educ. 2009;52(2):302\u0026ndash;12. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1016/j.compedu.2008.08.006\u003c/span\u003e\u003cspan address=\"doi:10.1016/j.compedu.2008.08.006\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eTierney P, Farmer SM. Creative self-efficacy development and creative performance over time. J Appl Psychol. 2011;96(2):277. https://doi.org/.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eVenkatesh V, Davis FD. A theoretical extension of the technology acceptance model: Four longitudinal field studies. Manage Sci. 2000;46(2):186\u0026ndash;204. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1287/mnsc.46.2.186.11926\u003c/span\u003e\u003cspan address=\"doi:10.1287/mnsc.46.2.186.11926\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eWirtz BW, Weyerer JC, Geyer C. Artificial intelligence and the public sector\u0026mdash;applications and challenges. Int J Public Adm. 2019;42(7):596\u0026ndash;615. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.1080/01900692.2018.1498103\u003c/span\u003e\u003cspan address=\"doi:10.1080/01900692.2018.1498103\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eXu G, Xue M. The relationship between proactive personality and migrant workers\u0026rsquo; perception of technical unemployment risk under the impact of artificial intelligence in China. Front Psychol. 2025;16:1474639. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/doi:10.3389/fpsyg.2025.1474639\u003c/span\u003e\u003cspan address=\"doi:10.3389/fpsyg.2025.1474639\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eYu C. The relationship between undergraduate students\u0026rsquo; creative self-efficacy, creative ability and career self-management. Int J Acad Res Progressive Educ Dev. 2013;2(2):181\u0026ndash;93.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhang T. Effects of self-regulation strategies on EFL learners\u0026rsquo; language learning motivation, willingness to communication, self-efficacy, and creativity. BMC Psychol. 2024;12(1):75. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s40359-024-01567-2\u003c/span\u003e\u003cspan address=\"10.1186/s40359-024-01567-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhao J, Li X, Wei J, Long X, Gao Z. Understanding the psychological pathways to translation technology competence: emotional intelligence, self-esteem, and innovation capability among EFL students. BMC Psychol. 2025;13(1):66. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s40359-025-02400-0\u003c/span\u003e\u003cspan address=\"10.1186/s40359-025-02400-0\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003eZhao P, Cao S. To participate or not to participate? Influence mechanism of artificial intelligence on Chinese college students\u0026rsquo; willingness to participate in online politics. BMC Psychol. 2024;12(1):525. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s40359-024-02009-9\u003c/span\u003e\u003cspan address=\"10.1186/s40359-024-02009-9\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"},{"header":"Footnotes","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003e Self-efficacy and Work Performance.\u003c/span\u003e\u003c/li\u003e\u003cli\u003e\u003cspan\u003e Innovation Ability and Work Performance.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"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":"bmc-psychology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"psyo","sideBox":"Learn more about [BMC Psychology](http://bmcpsychology.biomedcentral.com/)","snPcode":"","submissionUrl":"","title":"BMC Psychology","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"AI application, self-efficacy, innovation ability, work performance, willingness to change, civil servant","lastPublishedDoi":"10.21203/rs.3.rs-6526907/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-6526907/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cstrong\u003eBackground: \u003c/strong\u003eAs direct providers of public services and key agents of social governance, civil servants have experienced significant changes in their roles and work behaviors under the wave of digital transformation in the public sector, driven by modern information technologies such as artificial intelligence(AI). However, there remain few studies exploring how AI has influenced civil servants’ work performance and through what mechanisms these effects occur. Therefore, this study investigates the relationship between AI application and civil servants’ work performance, focusing on the mediating roles of self-efficacy and innovation ability, as well as the moderating role of willingness to change.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eMethods: \u003c/strong\u003eA quantitative research design was adopted in this study. Data were collected through an online questionnaire from 296 civil servants. The empirical analysis employs a structural equation model, chain mediation model, moderation model, and bootstrap method, with all statistical procedures conducted using Stata 16.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eResults: \u003c/strong\u003eThe findings confirmed that: AI application was positively correlated with civil servants’ work performance; self-efficacy and innovation ability mediated the relationship between AI application and work performance; A chain-mediating effect through self-efficacy and innovation ability was observed in this relationship; Willingness to change exerted a moderating effect on the relationships between AI application and innovation ability, as well as between self-efficacy and innovation ability.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConclusion: \u003c/strong\u003eThis study offers a novel perspective for understanding the behavioral motivations and underlying logic of civil servants amid the digital transformation in the public sector, by emphasizing the impact of AI application, the mediating effects of self-efficacy and innovation ability, and the moderating role of willingness to change. These findings provide beneficial insights for civil servants and policymakers, underscoring the importance of prioritizing and accelerating the learning of AI and the enforcement of AI-related competencies. Further, the study offers practical recommendations from the perspectives of fostering an innovation-oriented institutional environment and stimulating individual motivation for technology adoption.\u003c/p\u003e","manuscriptTitle":"The relationship between artificial intelligence and civil servants’ work performance: a chain mediation model of self-efficacy and innovation ability","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-07-28 13:46:08","doi":"10.21203/rs.3.rs-6526907/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"editorInvitedReview","content":"","date":"2025-08-06T10:23:36+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"30768662431024322941943750023225245184","date":"2025-08-01T06:56:10+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"97870818969384038846314787664306494493","date":"2025-07-27T09:37:35+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"88135784267849075179225592326813291634","date":"2025-07-26T14:00:34+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2025-07-25T04:06:02+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2025-07-23T11:42:37+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2025-05-09T14:44:41+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2025-05-06T15:56:00+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Psychology","date":"2025-05-06T15:54:57+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-psychology","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"psyo","sideBox":"Learn more about [BMC Psychology](http://bmcpsychology.biomedcentral.com/)","snPcode":"","submissionUrl":"","title":"BMC Psychology","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":true,"editorialSystem":"stoa","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"6a113647-2147-4435-a097-dc7ebfbf475d","owner":[],"postedDate":"July 28th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2025-07-28T13:46:08+00:00","versionOfRecord":[],"versionCreatedAt":"2025-07-28 13:46:08","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-6526907","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-6526907","identity":"rs-6526907","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}