In this study, the online smart research platform Credamo¹ and its data collection services was used to design and conduct the experiment. The employees of enterprises and institutions across China were grouped by random sampling. Each group presented only one condition. For a pilot test, a total of 39 questionnaires were collected, and 11 non-knowledge-workers and those who failed to pass the attention-check questions were excluded. As a result, 28 data were obtained. After modifying the screening questions for knowledge workers based on the pilot results, a formal experiment was conducted. A total of 378 responses were collected. After excluding 83 non-voice responses, 295 valid interview data were obtained. By screening out 135 non-knowledge workers and those who failed to pass the attention-check questions, 243 valid responses for the experiment data were obtained.

Among them, 146 women accounted for 60% of the total sample, and 97 men accounted for 40% of the total sample. There were 34 subjects aged from 18 to 24, 101 subjects aged 25–30, 89 subjects aged from 31 to 40, 13 subjects aged 41–50, and 6 subjects aged 51–60, accounting for 14, 42, 37, 5, and 2% of the total sample, respectively. There were 2 subjects with junior high school education or below, 2 subjects in high school or technical secondary school, 16 associate-degree students, 178 college undergraduates, 45 subjects who were postgraduates or above, accounting for 1, 1, 6, 73, and 19% of the total sample, respectively.

There were 48 participants from state-owned enterprises, 21 participants from public institutions, 6 participants from state organizations, 130 participants from private enterprises, 13 participants from foreign-funded enterprises, 24 participants working as professionals, and one participant identified as self-employed, accounting for 20, 9, 2, 53, 5, 10, and 1% of the total sample, respectively. There were 72 participants whose positions are ordinary employees/non-managers, 87 participants at the basic-level management/junior technician/section level, 66 participants at the middle management/intermediate technician/department level, and 18 participants at the senior management/senior technicians/office/bureau level, accounting for 30, 36, 27, and 7% of the total sample, respectively.

The instruments used in this study are as follows:

The measurement of perceived supervisor support was based on the scale developed by Hayton et al. (2012). We selected the first four items according to the size of factor load value, and then modified them to better suit the participant ethics and culture. For psychological capital we adopted the Psychological Capital Scale (PCQ-24) compiled by Luthans et al. (2008), which includes four dimensions of self-efficacy, optimism, hope and resilience, with six items for each dimension. For employee wellbeing we adopted Zheng et al. (2015)'s integrated perspective that employee wellbeing is not only the cognition and perception of employees' satisfaction at the work and life levels, but also the psychological experience and satisfaction state of emotions displayed at the work and non-work levels, and thus the employee wellbeing questionnaire mainly includes the three dimensions of employee life wellbeing, employee work wellbeing, and employee psychological wellbeing.

In terms of experimental materials, the research took emotional management as the theoretical basis, through consulting hot events and relevant reports, and using action learning facilitation and brainstorming methods to guide the thinking process of three psychology graduate students and instructors, and finally summarized the following situations.

This study employed a proof-of-concept experiment by controlling and manipulating a single independent variable (the application of IVA) and observing the corresponding changes of the dependent variables. The method of interactive survey experiment with different priming materials and IVA content were used for two purposes: (1) to measure the relationship between supervisor support, psychological capital and employee wellbeing; (2) to explore the impact of the application of IVA on the above variables through situational simulation experiments.

The study adopted a 2 (IVA voice gender: male/female) × 2 (IVA voice trait: machine voice/anthropomorphic voice) two-factor between-subject design—corresponding to Group 1−4 in the script, and two control groups were designed: the plain text group and no interaction group (plain text means that the feedback of the IVA was presented in the form of text—corresponding to Group 5 in the script; no interaction means that the subjects filled in the scale directly after reading the background materials, and did not participate in the interaction of the IVA—corresponding to Group 6 in the script). The process of the experiment was that the subjects first read the situational materials after the end of the day's work, then completed the scale after voice interaction with IVA under prompts, and understood the needs of knowledge workers for voice assistant services through voice message interviews. See Figure 1 for a diagram display of the procedure and see Supplementary material for the fully downloaded scripts for the experiment.

The protocols of the current study were approved by the Ethics Review Committee of School of Psychology, Fujian Normal University. The participants provided their written informed consent to participate in this study.

Qualitative results were acquired by asking the subjects the following two questions: (1) “Please recall the problems you have encountered with the supervisors at work recently and what events need intelligent voice help?” (2) “What services do you think the intelligent voice assistant can provide to enable you to obtain emotional support?”. Through coding and analysis, it is found that the services that knowledge workers want voice assistants to provide mainly include: information transmission and communication with leaders, work tips, answering questions and solving doubts, information search, providing encouragement and comfort, relieving stress, providing emotional support for leaders, feeling emotions, life tips, helping to make calls, sending messages and holiday condolences, which can be summarized into three categories: work support, emotional support and life support (see Table 1 for the coding schema).

In this study, we found that there were positive correlations between perceived supervisor support, employee wellbeing and psychological capital, and psychological capital played a completely intermediary role between supervisor support and employee wellbeing. When employees have a higher perceived supervisor support from their supervisors, their psychological capital is stronger and their wellbeing is higher.

As a part of organizational support, supervisor support has two main effects: the first direct effect, including the satisfaction of personal needs and the proportion of health and wellbeing, will directly reduce the impact of work stress. Second, as the agent of the organization, immediate supervisors' positive attitudes toward their subordinates are regarded by them as conferring support of the organization, and employees may not bring out their best work output until their internal psychology are warmly impacted (such as attaining emotional commitment). At the same time, supervisor support also positively affects subjective wellbeing, which reduces turnover intention (Gordon et al., 2018). If the level of supervisor support were low, employees could not feel the care from the organization. This is because the employees' own emotional input may not receive corresponding feedback and response, which would weaken the direct emotional connection between employees and the organization (Eisenberger et al., 2002; Dawley et al., 2008; Khan et al., 2015).

The subjective wellbeing of employees is influenced by leadership style as well as by perceived supervisor support (Celik et al., 2020). Having open and frank communication and reliable supervisors can enhance employees' sense of emotional support, enhance employees' wellbeing (Aycan and Eskin, 2005), promoting employees' self-esteem and sense of identity with the organization, and can lead to more pleasant feelings (Herrbach and Mignonac, 2004). Therefore, strengthening supervisor support can help employees experience positive emotions in their work and alleviate the pressure and occupational fatigue caused by negative emotions (Gilbreath and Benson, 2004; Nabawanuka and Ekmekcioglu, 2022). The conclusions of this study are consistent with these findings.

Previous studies showed that psychological capital was positively associated with job satisfaction and positive affect, and negatively associated with negative affect and psychological and physical distress (Wardani et al., 2020; Ho and Chan, 2022). In order to enhance employees' psychological capital, recently emotion management has been paid more and more attention in human resource management. Specifically, it is a management method that managers use sincere words to strengthen the emotional connection between managers and employees, to meet the psychological needs of employees, and to form a harmonious working atmosphere (Bolton, 2005; Sloan, 2008; Kaplan et al., 2014). When employees feel their supervisors are more supportive, they are more satisfied with their job and may have enhanced performance. In future organizational communication, we should pay more attention to the role of emotion management. Managers should show love and care for employees, establish harmonious interpersonal relationships, and ultimately achieve the emotional sublimation of loyalty to the organization. The enrichment of psychological capital in a nurturing and supportive environment contributes to employees' psychosocial functioning in the long run (Ho and Chan, 2022).

This study used an interactive survey experiment to study the impact of the application of IVA on the psychology of knowledge workers. The study found that the interaction of IVA can improve the supervisor support, psychological capital and employee wellbeing of knowledge workers, but there was no significant difference in the use effect of different voice attributes in this study. During the interaction between the subjects and the IVA, when the IVA conveys the concern and encouragement of the supervisor, the employees can feel the concern and encouragement of the supervisor, thus improving the degree of confidence and hope and enhancing the sense of wellbeing. As IVA becomes more popular, users turn to IVA not only for routine tasks, but also for expressing emotions. Sharevski et al. (2021) found that the broadcast content of the intelligent voice assistant affects subjects' perceptions of and emotions. Verleysen et al. (2015) found that as the appreciative inquiry increased among employees, so as their resilience, self-efficacy, hope and optimism, which is consistent with the results of this study via the application of IVA.

In remote work, the use of IVA is particularly important. Low quality virtual remote communication between supervisors and colleagues may render their personal goals and priorities unclear (Raghuram et al., 2001). Remote workers may need more self-regulation in this scenario. For example, it could be difficult for remote workers to coordinate some of their work tasks with colleagues, synchronize and coordinate work actions, work schedules and resting time. It may also be difficult to motivate each other and help each other through timely information exchange (Raghuram et al., 2003), which results in difficulties to obtain and share information.

In the coding of interview results, 12% of total participants also reported that they hoped that leaders could encourage and care about themselves and affirm their work value through IVAs. More than 33% of total participants hope to be able to transmit all kinds of information and communicate with leaders and colleagues through IVAs. Seventeen percent of total participants reported that they hoped that the IVA could remind their work progress and arrange work plans for themselves. Twenty-three percent of total participants reported that they hoped that the IVA could relieve their pressure by playing music and telling jokes. It can be seen that the service demand of employees for IVA is no longer a simple reminder. In view of the fact that remote work may continue in the future, we need to study and explore how to promote the management in remote work, explore the needs of employees and the conditions to meet their needs, and use the research theory results to guide the design of intervention methods to develop a model that can assist knowledge workers in their daily work Information transmission and psychological counseling, and provide emotional support for personalized IVA to improve employee wellbeing and work performance.

Based on the extensive use of IVA, this study finds that the application of IVA can improve perceived supervisor support, psychological capital and employee wellbeing of knowledge workers, and also finds that there are positive correlations among the perceived supervisor support, employee wellbeing and psychological capital of knowledge workers. On one hand, a theoretical contribution of this study is to provide another piece of evidence that enhancing organizational communication can improve employee wellbeing. Employees who feel the affirmation and respect of their supervisors can have improved self-efficacy and enriched sense of hope. On the other hand, this study may prompt organizational managers to pay more attention to knowledge workers' emotional needs and provides sufficient social support to them, either directly if they were able to (via effective leadership skill training) or indirectly through the application of tools such as intelligent voice assistants. For knowledge workers, perceived supervisor support through either means can help them to have a good mood, maintain physical and mental health, and promote work performance and creativity. Aside from knowledge workers, supervisors can also have their personalized IVA, realizing efficient organizational communication and conveying timely support and affirmation to their subordinates when the subordinates feedback their work summary or encounter difficulties, so that the subordinates can feel the care and support of their supervisors. In this process, the elegantly-designed IVA can provide companionship and encouragement to improve the psychological capital and wellbeing of employees and also feedbacks important information to the supervisors for organizational decision making, which can realize the dual-development of both employees and organizations.

Nevertheless, this study is not without limitations. One limitation is that it adopted the method of cross-sectional study, and lacked more specific and dynamic structure data. Subsequent research can study the dynamic interaction of IVA in the real work situation of employees, compared with direct interactions with supervisors, and conduct longitudinal research to track the dynamic changes of employees' psychology. In addition, this study employed a unified scenario to test the effect of using IVA. Subsequent research can detect more nuanced impacts by providing personalized IVA services for the subjects.

Last but not least, many IVA products have developed different types of voice, such as child voice, adult female voice, adult male voice, all kinds of star voice and voice with different emotions. Subsequent research shall consider whether the different voice types of different IVAs have different effects on users.