Mental Wellbeing at Work: Perspectives of Software Engineers

Mental wellbeing has various definitions and is a concept that has been studied in multiple fields, including psychology [94], philosophy [116], and HCI [112]. Although there is no single canonical conceptualization of mental wellbeing, one characteristic that spans multiple conceptualizations is that mental wellbeing is not binary but a continuum. Specifically how this continuum is labeled varies; some have used “flourishing” to “languishing” [64], and others have used “excelling” to “in crisis” [25].

Mental wellbeing is multi-faceted and consists of different factors. For example, in the hedonic (i.e., pleasure) and eudaimonic (i.e., purposes) conceptualizations [24], affective, cognitive, and behavioral factors all contribute to one's mental wellbeing. Other conceptualizations similarly included a diverse sets of factors, particularly social factors [38, 65, 114]. Because these different factors can change over time, mental wellbeing [34] therefore is not static but dynamic.

Within HCI, studies have explored mental wellbeing in different contexts and with different populations, including healthcare [113], social media [95, 96], and teens and college students [10, 81]. A sizeable amount of HCI research on mental wellbeing on sensing, tracking, and delivering interventions via gamification and mobile technologies [55, 63, 99, 100, 114]. Studying mental wellbeing in the context of work has been an especially important and ongoing conversation within HCI [19]. Our study takes the opportunity to further explore the lived experiences of individuals’ mental wellbeing from an integrated perspective that acknowledges the complexity of mental wellbeing.

Workers have experienced declining mental wellbeing in recent years [88, 124] that has led to costly consequences for both the employer and employees. 52 percent of surveyed workers in the U.S.A indicated they experienced more burnouts in 2021, with 67 percent across all age groups believing that they experienced worsening burnout since 2020 [115]. Similarly, a comparison of survey data from 2019 and 2021 shows an increase from 59 percent to 76 percent of U.S.A workers who have experienced signs of poor mental wellbeing [88]. For the employer, some of the impacts are reduced employee productivity, lower job performance [43, 72, 108, 123], and difficulties in workforce retention [45, 106, 109]; for the employee, poor mental wellbeing has led to worsened physical health [117] and spillover effect into personal life [8, 47].

One framework useful for looking at mental wellbeing in the workplace is the Job Demand Control Support (JDCS) model. The JDCS model states that wellbeing at work is influenced by job demands, job control, and social support [60]. Job demands include any tasks that require physical or cognitive effort; job control includes employee autonomy at work and skill contribution; and social support refers to the relationships between employees and their colleagues and supervisors [32, 60]. Based on a low-to-high spectrum for all three components of the model, different job profiles have been created, such as low strain, active, passive, and high strain [60]. Past studies in HCI have used this model to develop contextual personas [126] to better understand employee needs. Other studies that have identified specific contributing factors to poor mental wellbeing at work, such as a lack of social support [6] and a lack of autonomy over work responsibilities [48, 51], also align with the JDCS model. Other common factors include failure to integrate work and personal life [6], poor company culture around work and mental health [45, 51], and interpersonal conflicts with coworkers [48, 51, 73]. Individual differences such as personality traits[75, 87], emotional control and self-motivation [20], and social structures [73] also influence the extent to which an individual experiences the impact of these work factors on their mental wellbeing. Within software engineering, additional factors include interruptions [57, 68, 125], interpersonal conflicts during engineering tasks (e.g., paired code review) [40], unmet job needs [68], and working overtime [85]. Socio-cultural factors such as gender and race are also contributing factors. For example, female software engineers are more likely to experience micromanagement and imposter syndrome, leading to poor mental wellbeing at work [118]. The JDCS model is therefore useful to help us understand the various factors that affect software engineers.

To identify the various factors that affect individuals, HCI researchers have used sensing and tracking approaches to study indicators, such as job strain and stress [53, 66, 105, 120], job satisfaction [56], and happiness [42]. While there has been growing interesting in understanding mental wellbeing from the lived experiences of participants [20, 30, 67], little has focused on the work context. A broader approach to understanding mental wellbeing may reveal additional ways in which mental wellbeing manifests itself through employees’ work practices and technology usage.

Several models are popularly used when assessing a mental wellbeing solutions, including Cooper's primary-secondary-tertiary model [22] and the Institute of Medicine's promotion-prevention-intervention approach to mental wellbeing management [119]. These models categorize the types of solutions available for mental wellbeing. Kinmay [66] found that organizations and employees typically use secondary and tertiary solutions, which closely aligns with the promotion and intervention stages of the Institute of Medicine's approach. A primary or prevention solution addresses sources of poor mental wellbeing before it manifests; a secondary or promotion solution aims to increase awareness and one's understanding of mental wellbeing; and a tertiary or intervention solution focuses on reducing symptoms of poor mental wellbeing.

In recent years, addressing employee mental wellbeing has become a core initiative for many organizations [124]. They have tried to address this issue through improved mental wellbeing related benefits, both as prevention and intervention, such as an increase in paid time off (PTOs) and vacation [27, 69, 109], financial support for wellbeing related activities [109], and enhanced insurance plans with employee assistance programs (EAPs) and mental health services (e.g., counseling) [37, 71, 124]. Furthermore, as many organizations evolve to permanent remote or hybrid working, they are also addressing factors that specifically impact remote workers such as providing flexible work schedules and organizing social events to counteract a lack of social interactions [54, 80]. Larger technology organizations are moving beyond mere policy changes by addressing the nature of work itself. For example, Microsoft urged their leadership teams and other organizations to rethink productivity metrics to include employee wellbeing, aiming to lessen the productivity pressure felt by employees [36].

Employees are also using technologies to manage mental wellbeing at work. These solutions typically help employees in one of two ways: (1) stress-sensing and (2) delivering mental wellbeing intervention and coping mechanisms. For example, smartwatches and other wearable technologies [5, 39, 98] are now equipped with heart rate sensors that allow workers to assess their stress levels and initiate action to alleviate stress [17]. Mindfulness activities and cognitive behavioral therapy (CBT) interventions are typically delivered via standalone mobile applications (e.g., Headspace [50] and Calm [16]) or embedded within the tools commonly used by employees (e.g., Microsoft Viva Insights [79]). Through the use of cameras and other body movement sensors, recent HCI research started to explore identifying the right moment at work to delivery stress-reduction solutions [2, 53, 105]. Some organizations further encourage their employees to use mental wellbeing mobile applications by reimbursing their purchases of those applications [89].

Despite these organizational efforts to improve mental wellbeing, many employees feel these efforts are insufficient and often not useful [37, 84]. Therefore, the use of organizational policies and initiatives remains low [44]. This is in contrast with the positive viewpoint that organizations have on the mental wellbeing policies they provide [37]. Additionally, user engagement and adherence continue to be a challenge for mental wellbeing mobile applications [13]. Although there is growing interest in using persuasive designs to increase user engagement, a recent study also found that it can hinder the positive impact of mental wellbeing mobile applications. Specifically for applications that focus on using mindfulness to reduce poor mental wellbeing symptoms, attempts at persuasive designs such as daily streaks turned the need-based intervention into a daily mundane task [61].

In early 2022, the Great Resignation prompted workers in two major industries – healthcare and technology - to resign due to increased job demands and burnout rates [21]. Within the software engineering field, 20 percent of software engineers across the globe were actively looking for new opportunities, prioritizing their experience at work [86]. Compared to many other professions, software engineers in the U.S.A have a high employee suicide rate [18], and 83 percent of the surveyed software engineers in the United Kingdom constantly experienced burnout at work [3]. Researchers have found that some software engineers use substances such as cannabis to increase concentration and induce comfort [26, 29], which can lead to a variety of problems when when overused. Additionally, software engineers’ poor mental wellbeing sometimes leads to low code quality and process adherence [42] that can directly affect products used in high-risk industries such as healthcare.

Another way to consider the work of software engineers and how it relates to mental wellbeing is through the JDCS model as described in Section 2.2. According to the model, a high strain job has high demand, low control, and low social support, which leads to increased risk of experiencing poor mental wellbeing [46, 49, 60]. Although we were unable to find past studies that directly assessed the work of software engineers along the JDCS model's spectrum, we can infer that software engineering is high strain based on prior studies conducted on the everyday work of software engineers [45, 77, 78]. It is high demand because it is tool-intensive and technology-reliant [6, 101], especially because software engineers face constant changes in product requirements and user needs [23], and because most of their work is performed via technology (e.g., coding, project management, code review). Their work can also be considered low control because software engineers often are interrupted in their individual focus time and must juggle getting their own work done with frequent meetings with collaborators [78, 125], “resulting in highly fragmented work” [78]. While the recent increase in hybrid and remote work [102] allows for work schedule flexibility, which increased job control [35], it also increased job demand by pressuring employees to be responsive over instant messages. At the same time, it reduced social interactions with colleagues [35]. Thus, their work also has low social support. By better understanding the work of software engineers and their mental wellbeing management challenges, we can be informed on future technology designs that support not only software engineers’ mental wellbeing but also other professions that are considered high strain.

While a growing body of research exists to examine mental wellbeing in the work context, a number of open questions remain about the challenges and barriers employees face in addressing their mental wellbeing. Consequently, this study has three primary research questions:

1. How does mental wellbeing manifest itself at work for software engineers?
   
2. What challenges and barriers do software engineers face with current company policies and digital solutions for mental wellness?
   
3. What do software engineers expect from technology for managing mental wellbeing?
   

Participants were primarily recruited through LinkedIn. Using convenience sampling [33]. The research team posted the study on their professional network to recruit the initial participants. We then asked the initial participants whether they would be willing to refer additional participants to our study (i.e., snowball sampling [33]). Participants were eligible for this study if they were: (1) self-identified as a software engineering professional, (2) U.S.A based, (3) English speaking, and (4) over the age of 18. We recruited a total of 17 participants, but 3 of them either did not meet the eligibility criteria or were unavailable to participate. In the end, we conducted semi-structured interviews with 14 participants, of which 5 were female, and 9 were male. All participants worked for different companies and ranged from entry-level developers to managers. Table 1 further summarizes the demographics of our participants.

All interviews lasted approximately 45 to 60 minutes and were conducted over Zoom by the first author. Participants were asked a series of open-ended questions regarding their experience with mental wellbeing at work, strategies used to manage mental wellbeing and coping mechanisms for poor mental wellbeing, the impact of poor mental wellbeing on work, and company policies that aimed to support employee mental wellbeing. With the participants’ consents, all interviews were recorded in video and audio. All audio was transcribed by Zoom using the live caption feature during the session. Both the text file of the live caption and the audio recording were then saved only locally on a password-protected device. The transcripts were then anonymized, reviewed, and further edited by the interviewer to ensure accuracy by comparing them to the audio recordings. All participation was voluntary and the participants did not receive monetary compensation for their participation.

To analyze the interview data, we used inductive open coding [97]. The first two authors were the primary coders and frequently consulted with the rest of the research team for feedback, following the coding process for multiple coders as suggested by Hill et al [52]. We first independently open-coded two transcripts and compared our codes to develop a codebook with codes such as “definition of poor mental wellbeing”, “coworker's communication”, and “repeated occurrences of poor mental wellbeing”. After three iterations, we finalized our initial codebook. Subsequently, each researcher coded three different transcripts individually and these were jointly reviewed and discussed to gain consensus on the codes used, as well as any new codes or modifications to the codebook. The two researchers then repeated the process of coding three transcripts and reviewing their findings weekly until all transcripts were coded. We also collected memos relevant to the research questions, and discussed and resolved disagreements along the way through open dialogue. After all transcripts were coded, the two researchers grouped relevant codes to identify themes based on the research questions. These include mental wellbeing at work, contributing factors to poor mental wellbeing, impact of poor mental wellbeing, and unsuccessful strategies. The preliminary themes were then discussed with the other members of the research team and a final set of themes were developed through that discussion.

This research was approved by the Institution Review Board of the research team's university. This included gaining verbal informed consent from each study participant at the commencement of the interview. All recordings were stored securely on an encrypted device accessible only to the researcher conducting the interviews. All personal data (i.e., names, company names) was anonymized before the transcripts were uploaded onto a secure University drive accessible only to the researchers involved in this study. Each recording was deleted once it had been transcribed.

Mental wellbeing can often be a difficult subject to discuss and participants may experience discomfort disclosing certain experiences. When conducting our interviews, we were mindful of aspects that build rapport including validation, reflection, and other active listening techniques. For example, we acknowledged that it could be difficult to disclose mental wellbeing challenges at work and we thanked our participants for sharing their experiences with us throughout the interview. We also ensured that participants had control and autonomy to decline to answer any question, while feeling comfortable asking the researchers to remove certain responses from the record if desired.

Our research team consists of experts in software engineering, clinical psychology, and HCI. Both the first and second authors have experience working with software engineers in the industry, with the second author has worked extensively as an associate software engineer and a manager. The multidisciplinary nature of the team allowed us to create an interview protocol that was tailored to software engineers and considered terminology that was unique to their work. It also helped establishing rapport with our participants more easily. For example, we asked about “work practice” and “the way you write code” instead of using the term “process”. Because the first author worked in the technology industry, they were able to connect with the participants in the interview sessions by having a shared understanding of jargon such as “standups” and “agile”. In the data analysis phase, the second author leveraged their background as a software engineer to provide additional insight. Their familiarity with the rhythms and stressors of software development activities helped in the coding, resulting in a meaningful codebook with its resulting set of themes. Other authors who were not software engineers also consulted on data analysis, which helped reduce potential bias of the first two authors and allowed the team to understand the data from different perspectives.

In this section, we describe indicators of and contributing factors to poor mental wellbeing at work and strategies used to manage mental wellbeing based on our participants’ lived experiences.

While participants were aware of and engaged in trying to manage their mental wellbeing, many of them still continued to experience challenges. In this section, we highlight some of the organizational and individual challenges that they identified.

In our interviews, we asked participants whether they had any potential solutions to address mental wellbeing at work. Despite wanting to be away from technology, some participants felt that technology could nonetheless help. For example, they believed that automation and passive sensing could be implemented to support employee mental wellbeing, such as technology that can analyze and predict an employee's mental wellbeing based on employee work data:

“I read about research even the tone of voice, there's, there are pretty good, machine learning models that basically can predict if someone is depressed or even susceptible to suicidal thoughts by just analyzing the voice.” (P6)

Similarly, personalized assistants “like a Siri or something that they could, you know, help you under–understand, you know, why your heart rate is elevated or something” (P11). The participant further elaborated on how a digital assistant at work could also “give you a little pep talk” (P11) so they would “have that kind of [supportive] person that's always in your corner” (P11). Another participant raised the idea that a software application can also “block you [from doing work]…maybe at noon it's like okay, you're supposed to take a break” (P9) and would not allow one to resume work until a break has been achieved. It is important to note that participants also emphasized the importance of privacy rules and user autonomy, especially ensuring that the application “wouldn't force you to do something” (P9). These ideas help offload the responsibility to recognize poor mental wellbeing to technology.

In addition to creating innovative tools at work to help manage mental wellbeing, participants also suggested changing their current work tools to be more social. As described earlier, social support was an effective way to cope for our participants, even when participants were merely present in the same space. However, remote work settings removed many social experiences that were once a part of in-person collaboration and participants longed for updating technologies at work to bring back togetherness. For example, P3 remembered how they “would go up to your coworker and say hey, Let's open this up and let's review it” (P3) and wished there was a tool for “more interactive pull requests” (P3) where they and their coworkers would be able to work on it simultaneously. Another participant expressed wanting better brainstorming technology that could provide the “same experience for a bunch of people to just get into a room and kind of just spitball with a whiteboard” (P4) as they found existing solutions lacking. Additionally, participants expressed the importance of talking about mental wellbeing at work with each other. They suggested leveraging existing technology to foster such communication. For example, having a “never ending Google Docs” (P3) helped P3 to self-reflect and bring up mental wellbeing challenges with their manager. Similarly, P14 wanted a way to track their personal goals and contributions to share with their manager and the higher management: “I do a lot of emotional labor, and that doesn't it doesn't count for anything on my team as far as like um, career growth” (P14). As a result, work technologies can play a role in facilitating team-based conversations about employee mental wellbeing.

Increasingly, both organizations and employees have become more aware of the importance of discussing and managing mental wellbeing at work. However, we are still in the early stages of understanding how poor mental wellbeing manifests itself at work. For example, our study suggests that while current research has focused on stress and its management [53, 66, 105, 120], software engineers described additional indicators of poor mental wellbeing that include affective feelings such as frustration, anger, dissatisfaction, and unhappiness, and physiological changes such as headaches and muscle tension. Stone [109] also suggested that for software engineers, poor mental wellbeing can manifest itself as a feeling of emptiness, or employees questioning their purposes of work. Similarly, Smith [104] studied the work of nurses and found that poor mental wellbeing manifested as both stress and physical fatigue. Additionally, different professions may have varying levels of expectations of the thresholds for what constitutes poor mental wellbeing. For example, the JDCS model [60] suggests that job roles with high demand, low control and low social support typically lead to worsened mental wellbeing; however, our study reveals that while software engineering fits the JDCS high strain profile, some software engineers appreciated and enjoyed the challenges that came with the job and understood that their work came with a considerable amount of stress. Raizada [91] compared this idea to physical injuries, where he stated that academic scholars may expect anxiety and depression to be part of their work much like how athletes expect injury as a result of their work. However, as Raizada [91] pointed out, it is important to identify when it is too much and have preventive measures in place for that tipping point. Consequently, we need to identify factors that lead to poor mental wellbeing based on the context of the particular profession.

Additionally, it is common for employees to experience a spillover effect of poor mental wellbeing at work, where emotions and behaviors experienced at work are carried over into daily life outside of work [8, 47]. Similarly, the opposite is true when mental wellbeing challenges from personal life are brought into work [8, 47]. Prior to the adaption of remote work, our participants mentioned that software engineers were responsible for being on-call occasionally, which required them to be available all day by phone, being vigilant of any emergencies that could arise and resolving them as soon as they could. This means “work” includes not only the work done at the employees’ work offices, but also at their homes. The increased implementation of remote and hybrid work further blurred the lines between work and personal life. Because this has been a known concern, organizations have leveraged technologies to mitigate productivity and collaboration challenges. However, these technologies have become a part of their work and we argue that organizations and the developers of those technologies have not considered the impact these technologies have on employees’ mental wellbeing. For example, Slack and Zoom can be used to increase socialization and team collaboration. But at the same time, our participants felt that being on Slack meant they felt pressured to always be available and responsive to their teammates’ requests and vice versa because of the instantaneous nature of Slack messaging. This shows that it not only adds to the blurred line between work and personal time, but also increased interruptions at work and constant reminders and pressure of deadlines. As new virtual workspaces such as Horizon Workrooms [90] gain traction among organizations with remote employees, it will be crucial for the HCI community to be sensitive to what mental wellbeing means in these new work environments and technologies.

However, we also found that technologies used for work tasks can also help manage mental wellbeing. For example, shared Outlook calendars allowed participants to schedule meetings with each other and be aware of each other's focus time, helping participants gain control over their work schedule. This in turn reduce the chances of poor mental wellbeing manifesting as a result of coordination challenges and frequent context switching. As new tools and technologies are added to employees’ work routine [6], it will be important to examine how individual's interactions with technologies affect their mental wellbeing.

A variety of different factors contribute to an individual's sense of mental wellbeing. HCI scholars and social scientists have studied mental wellbeing through the lens of the social ecology model [82, 110], which highlights the importance of considering interactions between different ecosystems of an individual, including their microsystem, mesosystem, exosystem, macrosystem, and chronosystem [14]. These ecosystems represent the individual, their immediate social circles, communities, the society, social and cultural values and beliefs, and changes that happen over time. As shown in Figure 1, we found a similar structure for mental wellbeing at work in this study; participants did not look at mental wellbeing at work as purely an individual issue but instead saw it at multiple levels – individual, team, and organization. Furthermore, employee mental wellbeing is affected by the interaction of multiple factors [48]; therefore, as Ng and Fischer state, "wellbeing should also be understood in multiple levels simultaneously" [83]. While our study found similar contributing factors to poor mental wellbeing as prior studies [48, 51, 73, 87], we also observed interactions among the individual, team, and organization levels (Figure 1). For example, our findings suggest that having a positive team dynamic helps counteract the negative impact of organizational factors that typically lead to poor mental wellbeing at work, such as heavy workload as a part of the job demands. Individual factors such as personalities and past experience with challenging situations affected employees’ perspectives on how problematic factors at other levels may be. Similarly, Tokdemir found that sleep quality (an individual factor) affected how much workload (an organizational factor) could influence employee mental wellbeing [117], suggesting an interaction between individual and organizational factors. Conversely, having an unsupportive manager or team members (an organizational factor) contributed to self-doubts (an individual factor). These interactions highlight that the various levels are not isolated from each other but rather are closely tied to each other. Consequently, when we are designing solutions to address mental wellbeing, we must account for these interactions.

Besides observing contributing factors across the individual, team, and organization levels, both our study as well as other research [45, 60, 66, 88] suggest that the mental wellbeing management is a shared responsibility across these three levels. This means individual employees, their colleagues, supervisors, and the senior leadership at the organization need to work with each other to promote mental wellbeing, as well as prevent and intervene poor mental wellbeing. Currently, this has not been done well based on our participants’ recollection. For example, when participants described how they did not leverage organizational resources for mental wellbeing because of a lack of awareness, not only the individual employee has the responsibility to seek out those resources but also the organization needs to improve communication and advertisement of such resources. Additionally, many of the current strategies and solutions primarily assume individual responsibility and do not support team-based or organization-based mental wellbeing management, including many mindfulness applications that require employees to take responsibility to initiate care for themselves. However, there is a growing focus on understanding these interactions. For instance, while a recent personal sensing study recommended breaks based on employees’ work schedules, the researchers also suggested a team-centric approach [62]. This could be recommending breaks at a team level based on multiple team members’ availability and work patterns, further facilitating the social support an employee needs for managing their mental wellbeing at work. In sum, any changes to the management at one level has to also be evaluated as to its impact at the other levels and account for the interactions amongst these different levels.

Mental wellbeing across individual, team, and organization levels and factors at different levels. Displayed in the form of an onion diagram, with individual in the smallest circle, team in the next one, and team in the largest circle.

By incorporating mental wellbeing management as part of employees’ regular work practices, we can start to address challenges raised by our participants, such as motivation, time commitment, and the spillover effect. To address these issues, organizational and policy changes that support mental wellbeing are crucial [45]. However, their implementation can be complex and time-consuming. While organizational processes and policies can influence employees’ behaviors and attitudes toward mental wellbeing and its management, technology can also influence individuals’ or teams’ behaviors and beliefs [59]. As a result, we discuss ways to rethink technologies at work that integrates mental wellbeing into individual work practices, and team and organizational cultures. In our findings, we highlighted how employee's mental wellbeing experiences are impacted by the technologies they use to complete work tasks. Boivie et al. suggested that software development processes need to take users’ health and psychological wellbeing into consideration [11], particularly the job demands, control, and social support users may have at work. Similar to software accessibility features such as varying font sizes and ensuring keyboard ease of use, mental wellbeing should also be considered when designing and developing work tools. Furthermore, as accessibility has been brought more formally into the design and evaluation process for tools, [12], we need to consider how to also include mental wellbeing more formally in the process. Meanwhile, mental wellbeing management at work needs to be continuous, supporting not only intervention and prevention of poor mental wellbeing, but also promotion of general mental wellbeing [104].

One approach is to improve features to prevent known contributing factors from impacting employee wellbeing. This can be by reducing job demands such as the need to always be online and responsive, and increasing control and autonomy. A recent example is incorporating smart technologies that evaluate the number of recent notifications and recommend muting or continuing to mute notifications [15], so that instant messaging tools do not interrupt workers’ workflow or demand employee's constant attention for messaging requests.

The promotion of mental wellbeing and its management can also be integrated into work technologies. Promotion can include education around mental wellbeing and potential management solutions for poor mental wellbeing. It can also mean increasing discussion around mental wellbeing and reducing stigma at work. As we discussed earlier, mental wellbeing is a shared responsibility across the individual, team, and organization, so promotional efforts need to exist at all levels. For example, on a team level, Rico [93] suggested performing reflexive activities regularly for emotion regulation. Similarly, Rich's study integrated group discussions on mental wellbeing into doctors’ workflow and found it to be effective in promoting and managing mental wellbeing [92]. We argue that technology can help. For software engineering specifically, our participants mentioned that they follow the agile process, in which they and their team perform daily check-ins for their technical projects (i.e., stand-ups) and complete a reflection meeting (i.e., retrospective) after every project release focused on project improvements [7]. They often use technology (e.g., Standuply [107] and TeamRetro [111]) to guide these conversations. As a result, features such as including mental wellbeing check-in questions and guidelines can be incorporated directly into these technologies to help guide mental wellbeing conversation. Similarly, technologies used for setting performance goals and one-on-one check-in agendas can also include similar guiding questions.

As for integrating intervention into work tools, recent efforts have been made by both industry and academia. For example, Github's Good Day Project [57] and Microsoft's Viva Insights [79] both experimented with embedding mood tracking and daily reflections into the technologies software employees use on a daily basis. They also developed features that prompted employees to schedule focus time on their calendars. Specifically, daily reflection had a high completion rate among employees and showed promises in helping employees understand their mental wellbeing states and evaluate aspects of work that contributed to their poor mental wellbeing [57]. Similarly, one of our participants described their use of Google Docs, a common work tool, for daily reflection. They found relying on a tool they already used to be convenient. Additionally, chatbots have been used as personalized mental wellbeing management assistance to encourage mindfulness practices and other evidence-based techniques (e.g., CBT) [53, 122], along with other stress-sensing solutions and just-in-time interventions to provide intervention suggestions at the appropriate time [53, 70, 121]. Participants in these studies also self-reported improved mental wellbeing post-intervention.

Finally, it is important to consider various ethical concerns that come with introducing mental wellbeing data into work technologies. Currently, organizations often track employee productivity through data collected from tools they use and people have questioned the ethics with regards to employee privacy [58]. Considering the sensitivity of employee mental wellbeing data, conversations about the ethical implications of using personal sensing for wellbeing data and integrating mental wellbeing intervention into work tools also need to take place. Our participants already described concerns about data access and privacy. Other studies have also raised similar concerns over sharing personal sensing data with supervisors, especially when power dynamics and job security may be at play [1, 46]. As a result, future designs should consider potential ethical and privacy concerns and allow for user control and autonomy.