Fogg et al. defined persuasive technology as “interactive computing systems designed to change people's attitudes and behaviors” (60). Similarly, Oinas-Kukkonen et al. defined persuasive systems as “computerized software or information systems designed to reinforce, change or shape attitudes or behaviors or both without using coercion or deception” (61). Both Fogg and Oinas-Kukkonen proposed persuasion frameworks that are useful in analyzing and designing persuasive technologies—the Fogg Behavior Model (FBM) (62) and Persuasive Systems Design (PSD) framework (58), respectively. The FBM posits that a person must meet three conditions to perform a target behavior: (i) be sufficiently motivated, (ii) possess the ability to perform the behavior, (iii) receive the trigger to perform the behavior. Hence, behavior is a combination of three factors that must occur at the same time: motivation, ability, and triggers. The PSD framework builds upon the FBM to describe how to inject persuasive features that motivate behavior change into systems during the design process. The PSD framework comprises 28 persuasive strategies grouped into four categories—primary task support, dialogue support, system credibility support, and social support. The primary task support category comprises persuasive strategies that support users in carrying out their primary task. In contrast, the dialogue support category includes persuasive strategies that provide some degree of system feedback to users. The system credibility support category comprises persuasive strategies that support designing a system that it is more credible and thus more persuasive. The social support category includes persuasive strategies that enable system users to interact or socialize with others. Table 1 shows the four categories and their respective persuasive strategies, as well as a description of each strategy according to (58).

Additional frameworks have emerged over the years, such as Cialdini's Principles of Persuasion (CPP) (59). CPP comprises six (6) persuasive strategies described below:

Although the PSD framework is comprehensive and widely used to extract persuasive strategies implemented in mobile apps, the framework did not capture strategies such as reciprocity, consistency/commitment, and scarcity which are commonly implemented in apps for behavior change (57). In addition, social proof, liking, and authority strategies in CPP are similar to social facilitation, liking, and authority strategies in PSD. Hence, in this study, we focused on reciprocity, consistency/commitment, and scarcity from CPP in addition to the PSD strategies.

Previous research has investigated the persuasiveness of apps targeting various domains, including health, using the PSD framework only or in conjunction with other persuasion frameworks.

In the general area of health and wellness, Langrial et al. (63) reviewed 12 healthcare apps to identify the PSD framework's persuasive strategies. Their findings revealed that strategies in the primary task support category (especially self-monitoring, reduction, and personalization) were relatively broadly utilized, while dialogue support strategies were largely underutilized. In addition, the implementation of social support strategies was found to be shallow. Similarly, Oyebode et al. (57) conducted a comparative analysis of 80 health and wellness apps to evaluate the persuasive features implemented using the PSD and CPP frameworks. Their findings revealed that personalization, self-monitoring, suggestion, trustworthiness, surface credibility, and liking are the dominant persuasive strategies under the PSD framework, while consistency/commitment is the dominant strategy under CPP. In addition, persuasive strategies under the social support category were generally underutilized. To investigate the effectiveness of persuasive technologies for health and wellness, Orji and Moffatt (64) conducted a literature search and found that persuasive technologies effectively promote healthy behaviors, with 92% of reviewed articles reporting positive outcomes. Similarly, Hamari et al. (65) conducted empirical research on persuasive technologies (95 articles), and the results show that 92.6% of the reviewed articles reported positive outcomes.

In the area of mental health, Alqahtani et al. (55) analyzed 103 mental health apps to determine their persuasiveness using the PSD framework and Behavior Change Techniques (BCTs). Their results revealed that self-monitoring is the most frequently employed persuasive strategy, followed by personalization, reminder, and rewards. Similarly, Chang et al. (66) analyzed 12 mental health apps using the PSD framework and Goal-setting theory. Their findings showed that persuasive strategies in the primary task support categories, including goal setting, are the most employed, while social support strategies were the least employed.

In the domain of disease management, Geuens et al. (67) reviewed 28 mobile apps for managing chronic arthritis using the PSD framework and BCTs. Their results revealed that persuasive strategies in the system credibility support are the most employed, followed by strategies in the primary task support, dialogue support, and social support categories in decreasing order of utility. Similarly, Fadhil and Wang (68) investigated the persuasiveness of 19 diabetes management apps and coded them using the PSD framework and BCTs. Based on their findings, persuasive strategies in the primary task support category (especially personalization, self-monitoring, and tailoring) and system credibility support (especially expertise and trustworthiness) are the most employed. In contrast, dialogue support and social support strategies are the least employed.

Furthermore, Win et al. (69) examined the persuasive features employed in medication management apps retrieved from 13 research articles using the PSD framework. Based on their findings, persuasive strategies in the primary task support and dialogue support categories (such as tailoring, self-monitoring, and reminders) were widely employed in the apps, while those in the social support category were the least utilized. Furthermore, Lehto et al. applied the PSD framework to identify and analyze persuasive strategies in web-based alcohol and smoking interventions retrieved from 23 articles. Their findings showed that primary task support strategies (especially reduction, self-monitoring, simulation, and personalization) were relatively widely utilized, while persuasive strategies in the social support category were moderately utilized. In addition, dialogue support strategies were generally found to be underutilized.

In the physical activity domain, Matthews et al. (70) reviewed fitness apps from 20 research articles using the PSD framework. Their findings revealed that self-monitoring is the most widely represented strategy in the primary task support category, followed by personalization and tunneling, while suggestion is the most common dialogue support persuasive strategy. Almutari and Orji (71) focused on identifying social support strategies employed by physical activity apps in 32 studies using the PSD framework. They found that majority of the apps employed the competition persuasive strategy to promote social interaction, followed by the cooperation and social comparison strategies.

In summary, the literature review above shows that the dominance of the strategy utilization in persuasive and behavior change app design varies; however, there is no analysis targeting the stress management domain and how the results compare with other areas of mental health. Therefore, this paper utilized the PSD and CPP frameworks to deconstruct the persuasive strategies implemented in 150 stress management apps. We conducted a comparative analysis to understand the persuasive strategies employed in the apps and uncover new insights that drive future research.

Apps were extracted from Google Play and Apple's App Store automatically using a script that we developed using JavaScript and Node.js runtime. Search keywords were formulated by combining the term “stress” with each of the following terms: management, monitoring, support, detection, mindfulness, meditation, breathing, relief, relaxation, curbing, prevention, and treatment. The search was conducted in February 2020 and produced 4,375 apps in total (3,308 apps from App store and 1,067 from Google Play). For analysis, the inclusion criteria were: (1) apps must have more than four user comments, (2) apps must be related to stress, and (3) apps must be free or free with in-app purchases (free/free^*). The apps that did not meet all of these criteria were excluded. To avoid duplication, any app that appeared on both platforms were considered as one. Information like app name, platform (i.e., iOS, Android, or both), developer name, free/free^* were collected. We randomly selected 150 out of the 318 eligible apps for review (see Figure 1). One hundred fifty is a good number for generalizable results.

The purpose of coding the apps in our research is to evaluate the number and type of persuasive strategies employed in mental health apps specifically related to stress. We collected 150 apps and deconstructed the persuasive strategies employed using the PSD and CPP frameworks. We combined both frameworks to boost the coverage of our coding process. Two persuasive technology researchers coded the apps after installing them and performing various tasks using the apps' features. For apps with in-app purchases, researchers accepted the free trial to enable examination of all persuasive strategies employed in the apps. The interrater agreement score for each strategy was computed after coding. Finally, a third expert reviewer was involved in resolving any disagreement for strategies having agreement score <100%. Figure 2 shows the steps of coding the apps. Appendices A, B present the summary of the apps evaluated and the persuasive strategies employed in the apps.

The interrater reliability for the coded apps was measured using the percentage of agreement metric (6), as shown in Table 2. Agreement occurs when both researchers indicate the presence or absence of a persuasive strategy in an app. Disagreement occurs if one reviewer indicates the presence of a strategy and the second reviewer indicates an absence. Reliability values range between 85.7 and 100% agreement depending on the persuasive strategy. The strategy with the lowest interrater reliability (85.7%) is social role, while 27 out of the 31 strategies obtained perfect agreement scores. Overall, all intercoder reliability scores are within the acceptable range [i.e., > 60% (72)].

We classified apps into three categories, as shown in Figure 3, based on the purpose intended by the app. Most apps were designed to provide different sets of interventions to help users relieve and cope with stress. For example, apps provide different relaxation sessions or a list of games to help reduce stress levels. However, some apps were developed to monitor or track users' stress levels by allowing them to record their feelings either manually or by capturing their stress in real-time via sensors. Few apps combined both stress monitoring (tracking) and intervention by detecting users' stress levels and providing interventions such as meditation and relaxation sessions.

Table 3 presents a summary of the reviewed apps.

To answer our second research question (RQ2), we identified how each persuasive strategy is operationalized or implemented in all 150 stress management apps. We discuss these implementations in the following subsections.

Figure 9 shows the comparison between different categories of PSD and CPP frameworks regarding the number of persuasive strategies employed in the reviewed apps. Our findings showed that primary task support strategies of the PSD model are the most commonly employed in the reviewed apps, followed by system credibility support, dialogue support strategies, and the CPP strategies. However, we found that social support strategies are least employed overall. This answered our first research question (RQ1).

In this section, we compare our findings of persuasive strategies in stress management apps to an existing systematic review of apps in other mental health domains including anxiety, depression, fears and worries, and mood disorder (55). As shown in Figure 10, personalization is the most commonly implemented strategy followed by self-monitoring and reminders in anxiety apps. Similarly, personalization is the most frequently employed strategy, followed by reminders and self-monitoring in depression apps. Personalization and self-monitoring strategies are equally implemented in apps targeting fear and worry, followed by reminders, tunneling, and praise strategies. However, in mood disorders apps, self-monitoring is the most employed strategy, followed by reminders and personalization. Furthermore, rehearsal, authority, verifiability, and social learning are the least employed strategies in anxiety apps. Similarly, the least implemented strategies in depression apps are social role, praise, verifiability, and social learning. However, the least employed strategies in mood disorder are reminders, rewards, real-world feel, and third-party endorsements.

Interestingly, our findings shows that tailoring, simulation, reduction, similarity, trustworthiness, and surface credibility were not employed in any other mental health apps. This distinguishes stress-related apps from other mental health apps with respect to the strategies employed in their designs. Our finding confirms that personalization and self-monitoring are the most frequently implemented persuasive strategies in stress management and other mental health apps. On the other hand, our results show that simulation, trustworthiness, real-world feel, and expertise strategies are commonly employed in stress management apps, compared to other mental health apps. This answered our third research question (RQ3).

Majority of the apps that were coded (94%) employed the personalization strategy to deliver content and interventions personalized to individual users based on their profiles (such as demographics, medical history, etc.) and activities within the app (i.e., behavioral patterns). Evidence already affirms the global shift from standard healthcare to personalized healthcare to enhance patient experience and significantly improve their health and wellbeing (73–75). Hence, it is unsurprising that mobile health (mHealth) apps that offer personalized content or services were found to be effective by target users (76–79), including apps for managing stress (80).

Self-monitoring is the next commonly used persuasive strategy after personalization. Our results show that 59% of the analyzed apps employed self-monitoring to deliver features that enable users to record their activities manually or automatically and track progress toward set goals over time. This strategy is crucial for apps that support self-management of health-related issues by enhancing users' insights into the state of their health and motivating them to take charge and seek medical attention where necessary. For example, data captured through self-monitoring can be used to track the risk of mild or chronic health conditions [including mental health issues, such as mood disorder (81), depression (82), stress (44), and others (83)] in real-time, and then trigger interventions to mitigate the risk. Interestingly, research has also found personalization and self-monitoring as dominant strategies in other mental health-related apps (55) since user activity data (or behavioral data) can be used to offer personalized services or interventions that will improve users' health and wellbeing.

Moreover, the simulation strategy which allows users to observe the cause and effect of their behavior was found in 34% of the analyzed apps. For example, some apps use tree and flower metaphors such as petal growth in the flower or more leaves on the tree to show the influence of completing meditation sessions on health. Research has shown that simulation can yield positive outcomes in patients experiencing stress (43, 84).

Furthermore, tailoring strategy offers similar content to users who share common interests and rehearsal strategy that allows users to rehearse a target behavior found in 21% of the coded apps. However, tunneling and reduction persuasive strategies were found in only 9 and 6% of the stress management apps, which is surprising since mental health patients are usually advised to avoid stressful situations and complex activities that can increase their stress levels. In addition, users can get frustrated easily if in-app support or guidance is missing in mental apps (85, 86).

In the dialogue support category, reminders strategy is the most commonly employed strategy, followed by rewards and suggestion. Reminders strategy is the top strategy possibly because it keeps users focused on their intended tasks. For instance, users are usually reminded or notified at specific intervals to meditate, check their heart rate, or log their activities. The suggestion strategy offers personalized tips or recommendations as push notifications (or pop-up messages) that draw users closer to their goals. Majority of the apps that implemented the suggestion made it customizable such that users can enable/disable or change the timing/frequency of the notifications. This improves the users' feeling of agency and sense of control (87, 88), which can increase their motivation.

Rewards strategy is another motivating factor since users tend to intensify their actions if incentives are attached (89). However, offering tangible rewards to people who are already motivated could confuse them about the real motives (90) and benefits of adopting a healthy behavior (91). Social role is the least employed strategy based on our findings and this may reflect users' concern about privacy and trust if they share sensitive information about their mental health issues (such as stress) with conversational/virtual agents or human specialists via apps (92–94). This also mirrors research evidence that patients are generally less willing to share or engage in most personal communication while using mental health apps due to trust and privacy concerns, among other issues (95).

Based on our findings, system credibility persuasive strategies are the second most employed strategies in stress management apps. Our results showed that trustworthiness strategy (39%) is the most implemented strategy followed by real-world feel (38% of the coded apps) and expertise (26% of the coded apps) in the system credibility support category of the PSD framework. From our results, we also found that trustworthiness was operationalized using privacy statements and informing the user about the terms and conditions for handling the collected data, which is in line with the previous research (96, 97) which shows that trustworthiness is an essential factor for increasing user engagement and safety of the personal data in the mHealth apps. In addition, (86) affirms that users using mental health apps are more concerned about their data privacy, information sharing, and data handling transparency. This explains why trustworthiness is the most implemented system credibility strategy in stress management apps.

According to (86), users prefer to use apps that are aesthetically pleasing (surface credibility) with interventions that are clinically validated (expertise and third-party endorsements) for their mental health needs. Prior research also indicates that using strategies from the system credibility category is essential for improving the overall credibility of the mHealth apps (70). In addition, apps implementing trustworthiness, expertise, and authority strategies have greater chances of persuading users to complete their target behavior (98). However, surface credibility and third-party endorsements are the least implemented strategies in the system credibility category, based on our findings. This is a call to action for app designers aiming to motivate behavior change and increase engagement in the area of stress management.

Although socially-driven persuasive strategies are the most frequently employed strategies in health technologies and other online support systems (64, 91, 99), we found that social support strategies are the least implemented persuasive strategies in stress management apps, as well as apps in other mental health domains (55). Most people are silent about their mental health issues possibly because they are afraid of being stigmatized if they openly discuss their health challenges (100). Consequently, they tend to withdraw from social interactions (100). Moreover, previous research has stated that it could be inappropriate to incorporate social comparison and competition in a mental health apps, particularly for stressed users since competition could induce unhealthy stress levels and discourage some people. Additionally, a positive correlation has been found between high levels of perceived competition and other mental health issues such as depression and anxiety (101). Therefore, applying social comparison and competition strategies in stress management apps could negatively impact users' stress levels.

Furthermore, literature shows that there is a relationship between cooperation and stress. According to Yuebing (102), individuals under stress worry about being exploited. Hence, they are less likely to cooperate, with tendencies to have negative expectations about their partners' behavior (102). As a result, stress may obstruct cooperation. However, incorporating social learning, social facilitation, and normative influence strategies in stress management apps could help stressed individuals to decrease the feeling of being alone, isolated, or stigmatized. Evidence shows that forums that provide peer support may foster social inclusion in unexpected and new ways and should be a core element of recovery-oriented mental health policy (103). In addition, peer support is a primary part of helping many people experiencing mental ill-health in their recovery journeys (104).

Based on our findings, reciprocity is the most implemented strategy (21 apps), followed by commitment (14 apps) and scarcity (13 apps). This aligns with the research evidence which found reciprocity as the most implemented strategy in emotional and mental health apps (57). According to (105), if people receive something valuable in advance (such as gifts) before completing a task, they are encouraged to perform the target behavior since they feel indebted. Commitment is the second most commonly implemented strategy in emotional and mental health apps based on our findings and in alignment with (57). Employing commitment in stress management apps would motivate users to commit to their daily, weekly, or monthly goals and work consistently toward achieving them.

Table 5 shows the most and least dominant persuasive strategies for app category. Every individual faces a different kind of stress and the level of stress also varies from person to person. Therefore, it is not surprising that personalization is the most dominant in both “Stress Intervention Only” and “Stress Intervention and Monitoring” categories. Self-monitoring is essential for tracking users' stress levels and helps them visualize their progress over time. Some of the apps do not track stress levels directly, but they do track user activity, which is indirectly related to stress levels (e.g., number of meditation sessions completed by the user). This explains why the self-monitoring strategy is most dominant in the “Stress Monitoring Only” category.

Trustworthiness is another important strategy for stress management apps. Users tend to use mHealth apps that clearly specifies how their data will be stored and protected in the app. On the other hand, social support strategies (e.g., social comparison, recognition, competition, co-operation, etc.) are the least dominant, as the users experiencing stress are hesitant and less likely to compete with others in order to improve their stress, and most would like to keep it private to avoid stigmatization.

Furthermore, our findings revealed six (6) strategies (i.e., tailoring, simulation, reduction, similarity, trustworthiness, and surface credibility) implemented in stress management apps but not in other mental health apps. This could be due to several reasons. First, stress management apps usually tailor educational and interactive modules to users based on established various intervention techniques, including mindfulness, meditation, or breathing exercises (106–108). Intervention modules for stress reduction could also be tailored based on user groups such as general military members or combat-injured patients (109). Second, decreasing the complexity and steps users need to perform the target behavior (reduction strategy) could reduce negative emotions that contribute to high-stress levels. Third, surface credibility is essential for stress management apps since organized, efficient, and ads-free content could contribute to stress reduction (110). Fourth, privacy and trust concerns surrounding mental health apps, including stress (111), tend to be a major factor influencing designers' implementation of trustworthiness strategy in stress management apps. Finally, simulation strategy is widely used as a technique for stress relief such that users can see the cause and effect of their behavior. For instance, recent research, informed by Laughter therapy, implemented flower blooms at a certain point based on users' laugh energy (112). Another app encourages users to practice relaxation training that would help reduce their stress when users' stress level is balanced; a tree grows up strong and healthy when users' stress level is balanced or normal (113).

Everyone experiences varying stress levels at different times; hence, coping intervention can only produce the desired outcome if it fits each target user's current stress level. Hence, designers of stress management apps should leverage sensors data, in addition to self-reports where available, to detect stress levels in real-time using machine learning (ML) models and then offer personalized interventions that relieve the stress detected. Stress interventions can be personalized based on users' preferences through a pre-survey that introduces a list of recommended interventions, in which users can choose the interventions that work in reducing their stress—for example, practicing meditation or getting social support. In addition, users' past preferences can also be fed into ML models to determine the most suitable intervention. As a result, users will find the apps effective for managing their stress on an ongoing basis, thereby boosting user motivation and engagement/adoption in both the short and long term.

Most reviewed stress management apps lack in-app support and fail to simplify tasks users are expected to perform. App designers should reduce tasks into simpler steps (reduction) and offer contextual guidance (tunneling) until the tasks are successfully completed to avoid raising the stress levels of their users (114). Simplifying tasks and minimizing user efforts through providing templates for journaling which users can easily fill or change as necessary. Since feeling stressed may affect users' choices, the system should automatically arrange and present interventions based on users' needs/type (tailoring) and support their adoption of the target intervention.

Data privacy and transparency are important factors to ensure the trustworthiness of stress-based apps (96, 97) due to the sensitive nature of the data collected. Users of mental health apps usually have privacy and trust concerns that affect their decision to use such apps (111). To establish the users' trust, designers should provide the app's clear privacy policies and clear information about the data storage. Additionally, they should also provide steps for secure data transmission to third-party entities, how the user's data is stored and utilized, and steps taken to ensure users' anonymity. This information should be presented to the user in concise and plain language, and consent obtained prior to collecting any personal data (trustworthiness).

To improve system credibility, designers should use evidence-based or clinically-validated stress relief techniques that are proven to be effective for a larger group of audience in the design of their apps. These techniques can be in the form of daily or weekly psychoeducation materials (e.g., stress reduction, mindfulness skills, and well-being enhancement) developed by health professionals or taken from valid sources with proper references included or approved by top health authorities (e.g., World Health Organization, Public Health Agency of Canada, United States Department of Health and Human Services, etc.). Evidence-based contents or interventions delivered through mobile apps have been shown to be effective in reducing stress and improving mental wellbeing (108).

The usability of stress management apps is crucial for continuous engagement. Research has shown that aesthetically pleasing and stable user interfaces are more likely to be used by the users (86). Therefore, designers should make their apps visually attractive and have a consistent interface in terms of color, layout, and graphics. In addition, designers should ensure that the app's interface is simple, organized, easy to use/understand, and void of ambiguous and unstable functionalities (surface credibility). These practices will ensure that stress management apps are appealing, credible, and motivating to users.

Designers should provide a free trial upon user registration to allow them to use the premium features in the app. This can be achieved by offering a free trial for an initial period (e.g., 30 days) or linking the premium features to achievement of behavior change milestones (Scarcity). As a result, users will get the required experience to practice different stress management techniques and encourage them to continue using the app. Moreover, apps designed for stressed individuals should reduce distractions induced through in-app advertisements or ads (110).

Designers should provide users with an option to select their daily, weekly, or monthly goals for performing their target behavior, e.g., practicing breathing exercises or journaling their feeling/mood. In addition, designers should provide users with an activity dashboard to view progress and visualize how far they are from their defined goal. Moreover, the app should observe and monitor users' stress levels and recommend a set of specific goals to follow based on users' behavior or preferences. Research has shown that goal-setting, self-monitoring and feedback on users' performance are essential behavior change techniques (115).

Designers should include community, forum, or blog features to promote social or peer support as it is essential to overcoming stress (104, 116). Incorporating an online community will enable stressed individuals to get emotional support through empathy and other positive feelings from peers during stressful periods. Besides, users can get help or support from stress reduction techniques, advice, and suggestions shared on such in-app social settings. Therefore, accessing such social support will help users disconnect from stress by learning from other users' experiences and success stories (social learning).