The human fear of uncertainty, preference for order and structure, and desire for predictability are instinctive tendencies (Webster and Kruglanski, 1994). Therefore, whenever significant breakthroughs occur in science and technology, voices of anxiety and fear often rise to a fever pitch. On one hand, advancements in AI technology instantly enhance human efficiency, while the remarkable self-learning capabilities of machines in the AI 2.0 era generate considerable excitement. For instance, AI systems like ChatGPT rapidly amassed 123 million monthly active users within just 3 months of their launch (Nature, 2023). On the other hand, as artificial intelligence technology continues to penetrate deeper into human life, issues such as personal data leaks, low-quality information output, fraudulent misuse, and information overload have begun to trigger information anxiety (You, 2023; Zhu et al., 2023; Wang and Zhang, 2023). The development of technology is meant to serve humanity, but technological alienation stands in opposition to humans. Technological alienation is one of the important fulcrums for analyzing the modernity crisis in human society within the field of social economy according to Marx’s theory of alienation. Meanwhile, the employment anxiety of workers is considered one of the key points in discussing technological alienation in the era of artificial intelligence (Xu, 2022). The anxiety stemming from technological alienation is essentially rooted in the personal existential crisis of workers. Although artificial intelligence has not yet widely replaced human jobs, research suggests that the formation and expansion of the private ownership of digital means of production continue to perpetuate and entrench the vulnerable position of workers in the job market under private ownership conditions, becoming the premise and basis for the realization of workers’ employment anxiety in the context of AIGC technology alienation.” (Wang and Su, 2024).

However, in previous studies, researchers have generally exhibited a “Pro-Change Bias” regarding the diffusion of technological innovations, assuming that people are mostly willing to accept change and try innovative products. This assumption serves as a foundation for exploring the extent and reasons for the adoption of innovations (Talke and Heidenreich, 2014). Therefore, classic theories, including the Diffusion of Innovations, Technology Acceptance Model (TAM), and Unified Theory of Acceptance and Use of Technology (UTAUT), adopt a pro-change “positive bias.” These theories focus on the psychological, attitudinal, and behavioral aspects of users’ acceptance of innovative technologies and products, while to some extent overlooking users’ instinctive aversion to the uncertainty, disorder, and ambiguity brought about by innovation.

In reality, for most people, there is a greater tendency to maintain the status quo and consistency rather than pursuing change and innovation. Some studies summarize this inclination as users’ aversion to change and their satisfaction with the current situation (Talke and Heidenreich, 2014). Some researchers have identified six reasons for people’s instinctive rejection of technological innovation: fear of losing control, dogmatism, cognitive rigidity, limited ability to cope with change, intolerance during the adaptation period, weak curiosity, and habitual use (Xiang, 2024).

For contemporary Chinese society, artificial intelligence has not yet fully matured; therefore, not all workers have interacted with or utilized AI, nor do they possess accurate understandings of it. However, due to the “ignorance” stemming from information asymmetry and the “fear” amplified by media portrayals, the pseudo-issues that have not yet been encountered by users during the diffusion of technological innovation may be more complex than the real issues arising from current technologies. This can lead to longer resistance periods in the diffusion of technological innovation (Xiang, 2024).

Therefore, we propose Research Hypothesis 1: Current artificial intelligence technology has already generated workplace anxiety in the service industry.

Since the emergence of social media, information technology has begun to have a profound impact on people’s social relationships, subsequently affecting their life satisfaction (Pang, 2022). Some studies suggest that this impact may be negative (Kross et al., 2013). Especially for developing countries, the negative impact of these relationship-oriented media may even be manifested in the slowdown of continuous population growth (Hossain and Munam, 2022). The anthropomorphic features of artificial intelligence lead people to easily perceive it as a peer or even a friend during use, fostering interactions based on para-social relationships (Joerling et al., 2019); For example, research results regarding ChatGPT indicate that, in addition to efficiently providing relevant content services, the relevant responses offered by ChatGPT can also meet users’ emotional needs, which are positively correlated with happiness (Aburumman et al., 2024), thereby enhancing their life satisfaction. Research data from the United States has indicated that artificial intelligence is likely to reduce the demand for skilled labor and their wage levels in the short term, further exacerbating wage disparities. This may lead to job polarization, resulting in a decrease in the number of medium-income, medium-skill demand positions (Frey and Osborne, 2017). According to survey data from Frey, there are 702 occupations in the future United States that could potentially be replaced by computers, and 47% of American workers are at risk of employment disruption due to artificial intelligence (Frey and Osborne, 2017).

In China, it is predicted that over 50% of jobs will be impacted by artificial intelligence in the next 20 years. However, research also suggests that due to differences in educational levels, labor skills, and the regions of the workforce (Liu, 2023), the emergence of artificial intelligence may serve as either a complementary or a substitutive force for different occupations (Chen H. Y. et al., 2018; Wei et al., 2021). If some workers perceive AI as a substitutive force that displaces their roles, this could negatively affect their evaluations of job and life satisfaction, as well as reduce their willingness to participate in the labor market (Acemoglu and Restrepo, 2016).

In the service industry, artificial intelligence provides more meaningful interactions in human-machine communication than other forms of robots (Kim et al., 2019). Artificial intelligence can predict user preferences through intelligent algorithms, significantly enhancing the efficiency and accuracy of services (Chung et al., 2016). Artificial intelligence’s cognition, behavior, and emotional expression can more effectively meet the needs of consumers with higher demands for personalization, thereby enhancing users’ well-being and satisfaction (Henkens et al., 2020). For users with high levels of social anxiety, interacting with human service providers may lead to persistent and significant feelings of discomfort, tension, and even fear due to concerns about being scrutinized, among other negative manifestations (Li et al., 2017). In contrast, artificial intelligence service providers can mitigate these negative manifestations. When interacting with robots, users experience a greater sense of self-control, which empowers consumers with high social anxiety to engage in goal-directed motivations and regulate their behaviors (service robots) (Liu et al., 2024).

Although robotic service providers cannot guarantee the emergence of positive psychological outcomes for users, concerns about potential privacy risks associated with robot services, for example, can lead consumers to develop negative usage attitudes (Hsu and Lin, 2016). However, the emergence of artificial intelligence is still seen as a significant challenge to human service industry workers, the more low-skill jobs that do not require special skills or complex thinking, the more susceptible they are to the panic of “machine replacement.”(Brynjolfsson et al., 2024). The social perception, self-evaluation, and coping strategies of workers in these industries may be negatively impacted. For instance, employees facing unemployment anxiety are more likely to focus on the negative aspects of society, prioritize negative information, and overlook positive information. They may lose confidence and maintain low self-esteem, leading them to adopt negative coping strategies such as avoidance and resistance when confronting the challenges posed by artificial intelligence in their work and life, which can affect their mental health. Therefore, we propose Research Hypothesis 2 and 3. Research Hypothesis 2: AI workplace anxiety negatively predicts the life satisfaction of service industry workers. Research Hypothesis 3: Negative emotions mediate the relationship between AI workplace anxiety and life satisfaction.

The definition of social support can be viewed from both subjective and objective perspectives. From the subjective perspective, social support is considered a form of information that allows individuals to feel that they are cared for, loved, and respected (Cobb, 1976). From the objective perspective, social support is defined as an objective existence or the social interaction relationships that individuals perceive (Sarason et al., 1983). Social support can encompass material support as well as emotional support, which includes psychological comfort. It can also be subdivided into normative support provided by laws and institutions, known as formal support, and non-normative support obtained through social relationships such as marriage, kinship, and community ties (Mu, 2002; Li, 2012). It can be also categorized into three research perspectives: perceived social support, actual social support, and online social support. Actual social support refers to objective and tangible support such as financial aid, instrumental help, or formal institutional support; Online social support refers to interpersonal support exchanged through social media platforms or online communities. Perceived social support is defined as an individual’s perception of the level of social support received. In situations where individuals experience difficulties or stress, the perception of social support can mitigate the negative impact of adverse stimuli on their physical and mental well-being, enhancing positive emotional experiences (Zeng and Lü, 2024; Zeng et al., 2023). Research has also focused on the impact of social support on individuals’ work status. Studies involving various professions, such as nurses, telecommunications employees, and open-pit coal miners, have shown that higher levels of social support are associated with lower negative emotions in the workplace (Xu et al., 2024; Li et al., 2023), This is because social support can positively predict individuals’ self-efficacy, enhancing their positive self-perception in the workplace and leading employees to believe that the organization values their contributions (Ortan et al., 2021; Holzgang et al., 2023). This, in turn, reduces anxiety and depressive emotions experienced at work. It is worth noting that some studies suggest that, compared to Western societies that emphasize individualism, in Chinese society, which advocates collectivism, people tend to place a greater emphasis on group harmony and have a higher need for mutual assistance and reliance, which means that social support has a more significant impact on individuals in collectivist cultures (Xiong et al., 2021). Therefore, we propose Hypothesis 4: Social support moderates the relationship between AI workplace anxiety and negative emotions among workers in the service industry in China.

An online survey was conducted using the professional research platform WJX (Questionnaire Star). The survey targeted service industry employees aged 18 to 58 in Chongqing and Shenzhen, the selection of Chongqing and Shenzhen for this study was strategic for several reasons. Shenzhen, recognized as a Special Economic Zone and a hub for innovation, boasts a rapidly expanding service industry that mirrors the trends of modern economics of China. On the other hand, Chongqing, a pivotal economic and transportation nexus in Western China, presents a varied service sector. Together, these cities represent a diverse range of service industry workers, enhancing the study’s conclusions’ applicability across different contexts. The survey included positions such as supermarket staff, delivery workers, marketing personnel, and restaurant workers. A total of 600 questionnaires were distributed, with 549 valid responses collected, resulting in a response rate of 91.5%. Among the participants, there were 230 males (41.9%) and 319 females (58.1%); 273 individuals (49.7%) were from urban areas, while 276 (50.3%) were from rural areas; 255 participants were married (46.4%), 275 were single (50.1%), and 19 were divorced (3.5%); 312 had no children (56.8%) and 237 had children (43.2%); 291 participants had an education level of college or below (53.0%), while 258 had a bachelor’s degree or higher (47.0%); 292 individuals had a monthly income of less than 5,000 CNY (Chinese Yuan)(53.2%), 193 had an income between 5,000 and 10,000 CNY (25.3%), and 118 had an income between 10,001 and 20,000 CNY (21.5%). The average age was 29.35 ± 8.49 years. All participants signed an informed consent form before participating in the survey. This study was approved by the Ethics Committee of Chongqing University of Arts and Sciences (Approval No. CQWLDF0026).

The China Service Trade Association contacted service industry enterprises in Chongqing and Shenzhen to conduct the survey among their employees; Company management provided employees with the link to the online questionnaire and requested them to complete it. This study employs a cross-sectional survey design, aiming to examine the predictive relationship between AI job anxiety and life satisfaction, as well as the mediating and moderating effects involved A questionnaire survey is used, which consists of three parts. The first part is the “Informed Consent Form, “which informs participants about the purpose, content, anonymity, and confidentiality of the survey. Participants are asked to voluntarily sign the informed consent form, and upon signing, they will proceed to the formal questionnaire. The second part covers demographic variables, including gender, household registration, education level, whether the participants have children, marital status, and monthly income. The third part includes the survey instruments: Satisfaction with Life Scale, Artificial Intelligence Anxiety Scale (AIAS), the Depression Anxiety Stress Scales (DASS) and the Multidimensional Scale of Perceived Social Support (MSPSS).

Data entry and analysis were performed using SPSS 23.0. Independent samples t-tests and analysis of variance (ANOVA) were conducted to examine demographic differences in life satisfaction among service industry employees. Regression analysis was employed to investigate the predictive relationship between AI workplace anxiety and life satisfaction. The mediation effect of negative emotions and the moderating effect of social support were assessed using Model 4 and Model 7 from the Process macro. To evaluate potential common method bias among the variables, Harman’s single-factor test was applied. The results revealed that the first common factor accounted for 26.95% of the variance, which is below the critical threshold of 40%, indicating that the data in this study are not affected by bias due to similar research methods. A significance level of p < 0.05 was set to determine statistical significance.

The total score of AI workplace anxiety among service industry employees was (2.75 ± 0.72). There were no statistically significant differences in AI workplace anxiety across employees with different demographic characteristics (p > 0.05). See Table 1.

The overall life satisfaction score for service industry employees is (3.10 ± 0.85). There are statistically significant differences in life satisfaction based on educational level, marital status, having children, and monthly income (p < 0.01). Employees with a bachelor’s degree or higher have higher life satisfaction than those with a college diploma or lower; married individuals have higher satisfaction than unmarried or divorced individuals; those with children report higher satisfaction than those without; and employees with a monthly income of 10,000 CNY or more have higher satisfaction than those earning below 10,000 CNY. There are no statistically significant differences in life satisfaction between different genders or household registration status (p > 0.05). See Table 2.

The results of the correlation analysis indicate that life satisfaction among service industry employees is significantly negatively correlated with AI workplace anxiety and negative emotions (r = −0.170, −0.413, p < 0.01), and significantly positively correlated with social support (r = 0.514, p < 0.01). AI workplace anxiety is significantly positively correlated with negative emotions (r = 0.369, p < 0.01) and significantly negatively correlated with social support (r = −0.174, p < 0.01). See Table 3.

Using the educational level, marital status, having children, monthly income, and AI workplace anxiety—factors that showed significant differences in life satisfaction from the univariate analysis—as independent variables, and life satisfaction as the dependent variable, a multiple linear regression analysis was conducted. It was found that monthly income, having children, and AI workplace anxiety together predicted 7.2% of the variance in life satisfaction among service industry employees. Notably, AI workplace anxiety significantly negatively predicted life satisfaction (t = −3.905, p < 0.001). See Table 4.

AI workplace anxiety, negative emotions, and life satisfaction among service industry employees are significantly correlated with each other, thereby meeting the criteria for mediating effect analysis. All variables were centered, and demographic variables demonstrating significant differences in life satisfaction—namely educational level, marital status, number of children, and monthly income—were incorporated as control variables. Employing AI workplace anxiety as the independent variable, negative emotions as the mediating variable, and life satisfaction as the dependent variable, the mediating effect of negative emotions was examined using Model 4 from the Process macro.

The results indicated that the direct effect of AI workplace anxiety on life satisfaction was not statistically significant (p > 0.001). In contrast, the direct effect of AI workplace anxiety on negative emotions was significant (p < 0.001), with an effect size of 0.301. The mediating effect of AI workplace anxiety on life satisfaction through negative emotions was estimated at −0.161, with a 95% confidence interval (CI) of (−0.219 to −0.107) that does not encompass zero, thereby suggesting that negative emotions serve as a complete mediator between AI workplace anxiety and life satisfaction. The total effect was −0.188, with the mediating effect accounting for 85.64% of the total effect. See Table 5.

Given that negative emotions fully mediate the relationship between AI workplace anxiety and life satisfaction among service industry employees, Model 7 from the Process macro was employed to examine whether social support moderates the relationship between AI workplace anxiety and negative emotions. The demographic variables of educational level, marital status, having children, and monthly income were included as control variables, with AI workplace anxiety as the independent variable, negative emotions as the mediating variable, social support as the moderating variable, and life satisfaction as the dependent variable to test the moderated mediation model.

The results indicated that the interaction term between AI workplace anxiety and social support significantly predicted negative emotions (β = −0.098, t = −3.455, p < 0.01), suggesting that social support moderates the effect of AI workplace anxiety on negative emotions. See Table 6.

To further analyze the moderating effect of social support on the relationship between AI workplace anxiety and negative emotions among service industry employees, simple slope analysis was conducted by categorizing social support into high and low groups, using one standard deviation above and below the mean. The results indicated that as the level of social support increased, the predictive effect of AI workplace anxiety on negative emotions weakened (from β = 0.338, t = 8.275, p < 0.001 to β = 0.173, t = 4.714, p < 0.001). Higher levels of social support reduced the impact of AI workplace anxiety on negative emotions. See Figure 1.