Overview

Following a convergent parallel mixed–method study design we analyzed quantitative and qualitative data separately and chose a joint display in the results section for a comprehensive overview over our major findings. Since qualitative data allows for new themes to emerge and existing themes to be confirmed or discarded through the participants, we used the 4 major themes identified in qualitative research and complemented findings with quantitative data (see Table 1 for an overview).

Table 1 Overview over mixing of resultsSample characteristicsExpert interviews and focus groups

Two focus groups included a total of 8 participants aged 15 to 20 years (Mage = 18.6 years) with 4 females and 4 males (dropouts: N = 0). A total of 5 expert interviews (2 psychologists, 2 pedagogues, and 1 representatives of the digital industry) were conducted. Five out of 6 steering committee members agreed to participate at this time (see Additional file 4). With one expert, we could not arrange an interview appointment within a reasonable timeframe.

Survey sample

In our representative online survey, 666 individuals, aged between 16 and 25 years (Mage 21.3) with a representative gender distribution were included (Table 2). A more detailed description of socio-demographics of the sample can be found elsewhere [13]. From 685 youths that completed the online survey, 19 individuals were excluded after completion of quality control checks (eg, implausible response time or pattern).

Table 2 Demographic data including age, gender, educational level and ethnic background of participants (N = 666)Concepts of mental healthYoung peoples’ understanding of mental health

From young people’s perspectives, mental health mostly meant successful coping with challenges (8/8). Thus, being “mentally healthy”” implied being able to cope with challenges or crises in a successful manner:

"It [mental wellbeing; CG] doesn't necessarily mean that you're always in a good mood and always just happy, but that if things aren't going so well, you don't crash completely, because somehow you’re just strong enough to get out of there again." [P3, FG1]

Young people in focus groups also highlighted conflicts with themselves, parents or friends, as well as work—or school—related issues as the most challenging circumstances for their mental health (4/8). One female participant underlined technology and media consumption to be a source of stress for her mental health, especially the constant distraction by messages in social media [P1, FG2]. For half of the young people, “mental health” also meant awareness of one’s physical and mental wellbeing, including a better understanding of the functioning of one’s mind. One male participant stated: “I think, very often you are depressed or sad when you just can't figure out what makes you sad.“ [P4, FG1].

Focus group participants described a wide range of activities to promote and sustain their mental health. Most common were sporting activities (8/8), alone time (4/8), meditation or meditative activities such as drawing or writing (3/8), and social contacts, for example with friends or family (4/8). Participants also underlined paying attention to sleep and waking times or keeping track of their nutrition. With regard to distancing oneself from technology, one female participant emphasized to regularly "refrain from using cell phones and screens and such things" (P1, FG2). At this point, none of the participants had explicitly addressed the use of mHealth apps to promote their mental health.

Experts’ understanding of mental health in youth

In general, experts had a broad understanding of mental health in young people (eg, confidence, self–regulation of feelings or resilience). One expert emphasized the importance of behaviors that complement mental health, such as enough sleep, eating regularly, avoiding too much screen time, social contacts, exercise, and/or leisure activities (EXP1).

Concerning mental health in the planned app, two experts (EXP4, EXP2) raised concerns about using the terms “mental health”, “mental health promotion” or “prevention”, given their semantic proximity to “illness”. In order to avoid reinforcing the binarity of being healthy vs. ill, they recommended using the term “fitness”:

“You have a kind of continuum from what extends from health to non–health. And not necessarily a hard transition. And that's why I think it might be helpful not to put this label of health or illness on such an application, but rather [to focus; CG] on fitness.”[EXP4]

All experts underlined the high future relevance of digital interventions for mental health promotion and prevention due to a low threshold access to for young people, affinity of youths to digital media, and the inevitability of future developments in technology. Therefore, non–commercial providers should engage in app development as well as to:

“…pick up on this trend, the power that is in there and use it productively and maybe leave out a few things that you would do if you were to use it as a business idea, which should then be highly profitable.”[EXP3]

However, two mHealth experts reported few to no mentions of apps in communication with colleagues or young people (2/2). Yet the current COVID–19 pandemic was seen as an opportunity for change [EXP4, EXP5].

Current use of mHealth apps in young people

When asked for the use of mHealth apps in our focus groups (with a broad definition of mHealth apps, also including mental health promoting physical activity), the majority of young people (7/8) reported first–hand experiences. Participants reported the usage of sports and fitness apps (eg, “Adidas Sport App”, “Freeletics”; 3/8), meditation and mindfulness apps ("Calm", "Waking Up," “Headspace", etc.; 3/8), apps for sleep optimization ("Sleep Cycle" and others; 2/8), and nutrition apps (“My–Fitness–Call”; 1/8). This is in line with the survey results where more than two–thirds of young people reported to have experience with such apps (only 30% never used one).

Among survey participants who already had experiences with mHealth apps, 60% make regular use (at least once a week) of 1–2 apps, however 29% indicated to use none of their installed mHealth apps regularly. We also presented a selection of common app content or exercises to those who reported to regularly use mHealth apps (n = 473). Surprisingly, among these users, “never utilized” was selected most often (on average in 27% of cases) compared to the other answer options. Nevertheless, apps that promote physical activity or apps that track physiological data (eg, one’s heart rate) were rather common among existing app users. Less popular were apps with a diary options, apps that support stress management or relaxation/mediation apps. More details are shown in Fig. 1.

Fig. 1

Distribution of app types used among survey participants (N = 473). The response categories “often” and “very often” were combined to “(very) often”. The response categories ‘rarely’ and ‘very rarely’ were combined to “(very) rarely”

Young people in focus groups reported that they had used an mHealth app at least once (7/8) or more frequently over a short period of time (6/8), but had stopped the use of most apps. For 2 participants, this happened due to reaching their “goal”. Half of the focus group participants highlighted high costs or payment models as reasons, and for 2 participants their use ended with the free testing period of the app. In other cases, either the app was assessed to be unsuitable to reach the goal because it was too unreliable (2/8). For example, the alarm of a sleep cycle app did not go off, or the use was too demanding (3/8). For the latter, one participant stated concerning a nutrition app:

“I don't use it very often, I only used it once or twice just to see how I eat during the day ...” [P3, FG2]

“But you did stop at some point?” [Facilitator, CG]

P3, FG2: “Yes, exactly, because it is very exhausting to go through this every day …to keep listing the products there.”[P3, FG2]

In stressful periods (eg, final exams) 3 out of 8 focus group participants made use of mHealth apps. Two young people used mHealth apps for habit tracking (eg, to control eating habits or to self–organize).

AI use in the context of mHealth appsYoung people on AI in everyday life

When asking for young peoples’ perspectives on AI in general, most participants of our focus groups were able to name central mechanisms of AI (7 of 8). In particular, the terms “learning” (5 of 8), “machine learning" [P2, FG1], and “AI as a learning algorithm” [P4, FG1] were mentioned. Participants agreed upon the principle that more data leads to better outcomes of AI, although they reported a lack of understanding of the underlying technical details:

“I've heard of it, most definitely. Of course I don't know the technical background, but what I imagine is that if you give an input, the system learns“[P1, FG2]

AI usage was linked to application areas such as social media (2/8), shopping (2/8), and personalized advertising (2/8). Young people also mentioned online translators (eg, google translate) or the internet of things (eg, temperature control).

Most participants had a positive attitude towards AI use in general (6/8). They reported a pragmatic approach towards AI and data usage, which implied sharing data and use mHealth apps as long as the app was useful to them. One participant elaborated:

“I think there are two sides to this. I find it...scary, if you do not yet know exactly what is happening with your data…. On the other hand, I think it's great when I go to...Spotify ‘your mix of the week’ and I kind of like all the songs.... [As] long as I somehow benefit from it and know that my data is safe…, then I actually think that's good and I'm positive about it.”[P4, FG2].

Uneasiness about the opacity of machine learning processes was mentioned rather casually (6/8) and was often associated with popular media, especially science fiction movies (eg, “Terminator” [P2, FG1]). However, both focus groups came to the conclusion that regulations and control mechanisms for AI are necessary (8/8).

Opportunities and risks seen by young people

Concerning the use of AI in the mHealth app we are currently developing in the living lab AI4U, as presented in focus groups, participants showed general interest and openness and expressed few to no concerns about the usage of AI (7/8). Similarly, in the online survey, only a minority (17%) reported to feel negative about the application of AI in general, and 19% negative about the embedding of AI in mHealth apps.

In focus group 1, the combination of a great variety of interventions, and the possibility to tailor the app to one’s personal needs to find “your own path” was especially desirable for one participant [P2]. Similarly, other young persons (4/8) highlighted the advantage of personalization due to the AI elements:

“Yes, I think that in the area of mental health or health in general it could work pretty well to use such artificial intelligence, because…everyone is super individual and of course it works better if it is well adjusted to each person.“ [FG1_S].

At this point, 5 out of 8 young people stated that they have few concerns in sharing information with the app. Participants justified broad data usage, especially if the app is useful to solve user–defined problems. Two young people argued that their smartphone already gathers many personal data, thus sharing more information was no major concern to them:

“When I think about it, my cell phone probably already knows so much about me anyway, because it is somewhere nearby all the time...that I would maybe give more information than others.” [P2, FG1]

Asking which information participants were willing to provide in order to improve tailoring and functionality of AI (given all guidelines being followed), youth in the survey mostly agreed to share information like steps taken per day, physiological data (eg, heart rate), or sleep behavior, for example day/night rhythm or duration of sleep (see Fig. 2). At the same time, a majority of young persons disagreed to share real–life conversations or the recording of surrounding or ambient noises, as well as to the analysis of text or text length of messages that have been sent.

Fig. 2

Attitudes of survey participants towards data sharing across different types of data (N = 666). The response categories “agreement” and “strong agreement” were combined to “(strong) agreement”. The response categories “strong disagreement” and “disagreement” were combined to “(strong) disagreement”

During focus groups, young people also expressed reservations about AI's use of “personal” data, such as diagnoses (“app should not replace the doctor” [P2, FG2]) or one’s deeper feelings and thoughts. One male participant stated that he would share the level of stress he had but not the reason for the distress itself, like something he would put in a diary [P2, FG1]. Doubts were also expressed about the app´s functionality and ability to comprise more profound concepts (eg, melancholic feelings, thoughts about meaning in life, problems with friends or family). One male participant was skeptical about the app’s ability to grasp deeply entangled emotional problems and to give adequate answers [P4, FG1]. Reluctance to share personal information was linked to trustworthiness of the app producers (4/8), where non–commercial providers were trusted more, compared to providers with financial interests. This is in line with survey results, where only 20% deemed it very important to know who developed the app, while 23% were indifferent and 15% deemed it unimportant. Two–thirds of young people reported to have greatest trust towards independent research institutions (e.g., universities or hospitals) when compared to health insurance, state institutions, or companies such as “Google”. The latter and other companies with a potential commercial interest where trusted the least (see Fig. 3).

Fig. 3

Trust in potential app developers among survey participants (N = 666). The response categories “trust” and “great trust” were combined to “(great) trust”. The response categories “no trust at all” and “low trust” were combined to “no or low trust”

Young people in our online survey reported barriers for the use of mHealth apps including lacking motivation and costs for the app itself, while a majority assessed feeling worse than before, difficulties in handling the app or losing personal contact with real–life friends as ‘no barrier to usage’ (see Fig. 4).

Fig. 4

Reasons for non-usage of mHealth apps among survey participants (N = 666)

Opportunities and risks seen by experts

Experts expressed positive views on the use of AI in mHealth apps (5/5) by emphasizing the low–threshold access to young people (2/5), especially for reaching potential risk groups early on (eg, young persons with few social contacts). Similarly positive was the experts’ view on the potential for better–informed recommendations (2/5), resulting in a high applicability of the app in counseling and school centers (3/5). However, 2 experts stressed that the app should be connected to mental health counseling facilities in order to inform users about nearby professional services [EXP1, EXP5]. Especially in cases of suicidality, the app should include further measures for risk reduction and enable access to further help for those in need (2/5). For some experts the app may serve as a supplement to face–to–face interventions or counseling services (2/5). By empowering young people to take care of their mental health themselves, the app has the potential to close existing treatment gaps in mental health promotion and prevention for youths with minor mental health problems.

From experts’ perspectives, interventions of the planned app should provide “knowledge transfer”, for example to empower young people to realize how actions, feelings and thoughts are intertwined (2/5). Further recommendations included the promotion of “self–acceptance” / “self–worth” (2/5), “physical activity” (2/5), “social skills” (2/5), and “emotion regulation”, for example getting access to one’s feelings (2/5). Young peoples’ motivation to use the mHealth app might be increased by “giving stories / everyday examples” (3/5), tailored tips and ideas for specific situations (3/5), the initiation of self–reflection (2/5) or interventions to experience self–efficacy (2/5). One expert emphasized that several youth studies underlined that communication with peers was essential for young people in internet activities, thus the option to communicate with other young people should be included in the app to ensure a continuous use [EXP3].

Experts’ concerns about the app were mostly related to not being able to reach or motivate young people for the apps’ training (3/5). Furthermore, experts underlined that various barriers for the implementation of the app need to be taken into account, such as structural obstacles, for example prejudices of school officials against mHealth apps (2/5) or persisting lack of technical equipment in schools (2/5). They reported possible negative influences in enforcing digital media overuse [EXP5] and, given the continuous pressure to self–optimize, empowering potentially problematic “perfect” self–images [EXP3]. Based on the recommendation of one psychologist, the app should therefore not arouse false expectations in young people:

“I would see a risk if the app claimed: ‘if you go through these ten steps...then you are a different person’ (laughs). ...A good app would be characterized by the fact that the user does not internalize a problem centered perspective, but that he...gets the feeling: ‘I am okay’.” [EXP5]

Recommendations for AI–informed mHealth apps with EMA/EMIsYoung peoples’ recommendations in focus groupsApp includes a variety of exercises and gives orientation

Young people underlined that the planned app should not only provide tailored exercises (to one’s needs and goals) from a broad variety of exercises, but to also give users an overview and orientation over possible “routes” (eg, training sequences), in order to support users in finding and defining “their own path” [P2, FG1]. Or as one participant stated:

“...you have a lot of apps that can promote health, but you have to find the one app that suits you best ..., it’s actually exactly the other way around [with the planned app; CG], so that you have an app,...that can indicate the various routes to you....That's pretty handy.” [P2, FG1]

Personalized personalization–App includes users in AI decision making process

All young people demanded control over the degree of personalization through AI (8/8). One participant therefore proposed the concept of “personalized personalization” [P1, FG2], which means to be able to self–determine the time and degree of further personalization, ideally at the beginning of the app use. Also, young persons demanded flexibility and the possibility to change those settings retrospectively (2/8), for example to be able to receive suggestions for exercises once again, even after dismissing them initially (allowing for development or a change of mind).

App gives recommendations when “it’s smart enough”

Participants had concerns about imprecise or bad suggestions of the AI especially at the beginning of app use, due to a lack of data and therefore insufficient tailoring of the app. Thus, 2 participants proposed that AI should come into use only when sufficient personalization was reached to avoid imprecise suggestions (2/8):

“…when I'm still at the beginning, the app doesn't know much about me yet, but it still gives me suggestions .... And then I tend to distance myself from the app…because it simply suggests things that I don't need at all. So, I think...that it would be really cool if this AI only came in at a certain point in time, when it is smart, or when it’s smart enough to step in.”[P3, FG2]

App suggests solutions for difficult situations

Besides having a set of exercises at hand to volitionally practice certain aspects of mHealth, participants underlined the importance of tailored solutions (eg, a help button) in difficult moments of not feeling well (4/8).

“...but that you have such a button...where it says, no idea, ‘help me’...and then ...get the whole thing rolling.” [P2, FG1]

For others, this could also mean to receive suggestions for activities in times of inactivity–which could be measured by passive data (eg, steps taken per day)–and for example to receive suggestions to meet up with friends (2/8).

App suggests distance to mobile phone

Crucial feature of the app for young people was the encouragement to refrain from the use of mobile phones (2/8). One young person especially stressed distance from social media, as precursor of stress and competitive behavior [P3, FG1].

„ I think if an app like this gets you to turn off your cell phone and do breathing exercises, or some sport...and takes you away from it a bit, then I think it's very helpful, even though it's ‘just another app’." [P1, FG1]

Mutual trainings or sharing trainings with others rarely desired

7 out of 8 participants stated at some point that they exercise for themselves and not “for others". Therefore, 5 out of 8 young people preferred not to share training sessions or outcomes with acquaintances or other app users, for example to avoid competitive behavior (3/8). Especially the public posting of achievements via the app was deemed undesirable (5/8), whereas the one–on–one exchange with close friends or family was considered acceptable by some (2/8).

Reasonable time requirements for app usage

Given the rationale of proper time management, participants raised concerns about the effort of app usage and time required for the personalization of the app as well as the amount of exercised the app recommends:

„When you have to fill out super long stuff,...then it might annoy me at some point, if it is a lengthy process.“ [P3, FG2]

Important aspects of mHealth apps for survey participants

With respect to important aspects of the app (see Fig. 5), young people considered the “comprehensibility of app content” (84%), “quality and helpfulness of training” (82%), and the “possibility for personal goal setting” (79%) as important or very important. Compared to the other aspects, the chance for comparison with other users was deemed the least important aspect in an mHealth app.

Fig. 5

Important aspects in mHealth apps among survey participants (N = 666). The response categories “important” and “very important” were combined to “(very) important”. The response categories “very unimportant” and “unimportant” were combined to “(very) unimportant”

Experts’ recommendationsInclusion of young people into app development

Overall, experts recommended involving young people as early as possible in the app development process (5/5). This could mean to put special emphasis on “risk groups” (eg, socially disadvantaged young people, or young people displaying behavioral problems), which may have different needs concerning processes and structures of interventions (EXP4).

Target group–specific language

In terms of language editing, experts underlined to ensure that formal information (eg, use of data, data protection) was comprehensible. Thus not only to comply with the “clear and plain language” required by GDPR, but also “to formulate the terms and conditions in simple language, that anyone can understand” [EXP3]. Additionally, this should imply conveying “acceptance” or “affirmation” within texts (eg, that it takes time to perform an exercise, to practice at one's own pace) rather than taking a problem–centered approach [EXP3, EXP5].

Forwarding young people to counseling centers

According to experts, the app should be able to connect young people with helpful contacts in their vicinity (eg, suicide hotlines, counseling services, trust teachers) and facilitate the structured search for professional help [EXP1, EXP 5]. Thus, the app may reach risk groups at an early stage, when they are not yet likely to attend counseling or search for clinical support yet.

Low–threshold access to potentially endangered young people

Two experts highlighted the risk of triggering self–harming behavior or suicidality through app usage [EXP1, EXP 5], but emphasized that the planned app could also mean an early, low–threshold access to potentially endangered young people. One expert therefore recommended the careful phrasing of questions and exercises, but also stressed that suicidality and suicide ideation were often hard to recognize in counseling as well [EXP5]:

“It’s the same as in counseling…It’s never clear how things will proceed and if we are very careful there...we may reach someone at a point in time when they are at risk and use the app to get them to open up somehow, to turn them around.” [EXP5]

Careful framing of AI technology

Two experts raised concerns on a widespread lack of knowledge and the risk of limited acceptance of AI–informed technology. Due to false “dystopian” imaginations of AI–technology, there might be persisting reservations especially in Germany [EXP4], why an exceedingly strong focus on AI in the app could lead to reactance [EXP2]. As a counter–measure, it was stressed to lay emphasis on its benefits, for example tailored offers made possible through AI, and to guarantee as much transparency as possible [EXP2].