“I Kind of Bounce off It”: Translating Mental Health Principles into Real Life Through Story-Based Text Messages

Findings:

Participants expressed a strong desire for messages that convey “authenticity” — by which they meant the feeling that a message is from a real person. Participants like FP7 and FP37 mentioned that knowing a story is written by a real person would increase the sense of authenticity, allowing them to connect to story characters and find greater motivation for change. They anticipated that such stories would also give them a sense of comfort in knowing that someone else was able to overcome their difficult situation, which would boost their confidence in dealing with their own struggle. FP9 said,

“I like hearing other people’s stories and what they did, and it kind of helps me feel a little better. And I kind of bounce off it and do what they did and try these new things that they’re doing.”

Some participants also shared negative experiences reading what they believed to be manufactured stories that were fabricated to make them feel better. These individuals preferred stories that came from real people since they felt that manufactured stories were disingenuous and had little credibility. When we asked our participants how they distinguish manufactured stories from real ones, they stated that they often relied on “gut feelings”. However, FP7 speculated that stories with less specificity were more likely to be manufactured: “If the story sounds like this could happen to anybody, my mind makes me think, ‘Oh, then it didn’t happen to anybody.’”

Implications for Our Deployment:

We engendered authenticity in our stories using three different approaches. First, we created stories as a research team by drawing from our own real-world experiences rather than fabricating scenarios. The stories were written by members of our research team (see Section 5.1 for details about story generation). To make the origin of the stories clear, each one started with a sentence that explicitly indicated that they were true stories from real people. Second, we wrote stories in the first-person perspective with the intention to give the story character a voice, which would in turn help form a connection with the reader [24]. Third, we included enough details about the characters and situations to paint a concrete image in the readers’ minds. For example, rather than just referring to college students in general, we included details like their academic majors or their relationship to other characters in the story (e.g., an economics student sharing their problems with a close friend). These details are not always necessary for understanding a story, yet messages with more details are often perceived as more credible since they are more likely to trigger people’s memory of related experiences [100]. These nuances align with practices that generate authenticity while creating personas (i.e., concrete representations of target users) in user-centered design [3].

Findings:

Participants perceived that existing online resources and digital intervention services for mental health only provide superficial and generic suggestions like “do not overthink” and “try out a fun activity”. Despite sometimes finding these suggestions useful, participants felt that they have little impact because people often do not have the time or energy to think through how to apply them in their own lives. Hence, participants advocated for messages that give more precise guidance and suggestions. Participants recommended that messages should be able to explain why a certain practice would be beneficial to their wellbeing. FP12 explained that even a brief justification would be beneficial to them:

“I think one sentence would be nice. So, for me, I didn’t ever take breathing exercises seriously until someone said to me, ‘Oh, but the reason why it helps is because it brings you back to the present and helps you concentrate or prevent anxiety.’ So, that makes me take it more seriously.”

Implications for Our Deployment:

Inspired by the structure of Aesop’s Fables [39, 103], we explicitly provided a concrete takeaway message for the reader at the end of each story. This practice aligns with Barthes’ three-layer narrative structure [8], which consists of layers for actions, characters, and morals. The action and character layers describe the physical setting of the world and the characters involved, while the moral layer pinpoints the goal of the story [29]. For the action and character layers, our stories involved real people in real situations. For the moral layer, we introduced a psychological concept (e.g., “This person experienced a thinking trap of ‘overgeneralization’.”) and expanded on how the concept is tied to the story. Thus, we prompted readers to reflect on the purpose of the story and what it illustrates rather than on particular details or characters. In our case, a team member trained in clinical psychology designed the concrete takeaway messages to highlight how the stories illustrate common cognitive distortions.

Findings:

A frequent suggestion from our interviews was that the users of a story-based text messaging service should be involved as active participants rather than as passive readers. Our participants suspected that if all of the communication is one-way (i.e., from the system to the user), users might not spend as much time reflecting on the stories. FP1 and FP3 posited that we could increase engagement by following the stories with prompts for people to share their own experiences. When this suggestion was mentioned to other participants, many agreed that there is value to composing their own messages, but concerns were also raised. Some participants expected that people would be hesitant about sharing their personal details through a messaging platform, while others speculated that people will not always have the time or energy to write more than a couple of sentences.

Implications for Our Deployment:

We encouraged active and reciprocal participation in our deployment by asking users a reflective question at the end of each story, typically along the lines of “Have you had a situation where you had to deal with a similar challenge?”. To mitigate some of the concerns raised about people’s willingness to share personal information, participants in our deployment were made aware that their responses would not be shared beyond the research team. We anticipated that by answering such questions, users would be able to reflect on their own experiences and how the stories’ messages could be applied to their situation. This approach is similar to dialogic inquiry between parents and children during storytelling, in which the parent periodically pauses to ask the child to relate the story to their own experience [110].

Findings:

Participants in our studies suspected that reading too much about people’s struggles could sometimes induce sadness or decrease motivation. They instead suggested that we should focus on positive stories that inspire hope. That being said, participants also mentioned that presenting too rosy a picture of living with mental health symptoms would appear disingenuous or set unrealistic expectations for people. Therefore, participants suggested finding a balance between positivity and a realistic depiction of people’s struggles. FP8 explained that stories should show how a person overcame their struggles without trivializing the time and effort involved in making change:

“I like the idea of a successful story. But I would also like to be like, ‘Oh, people were going through something really bad, and now they’re getting help. Even though they’re not 100%, they are seeing how that’s helping and how it’s making everything better’. So not really just skipping directly to the end, just being able to say, ‘Hey, I’m still struggling, but I’m getting better.’ It’s just kind of a reminder, like, ‘Hey, it’s not going to fix itself overnight.’”

Implications for Our Deployment:

Although we avoided content around experiences that have potential to trigger secondary trauma (e.g., sexual assault, family deaths) [79], we sought to relay the serious issues young adults face. These topics included, but were not limited to, struggles precipitated by academic stress, relationships with parents, or failures in school and careers. We ended each of our stories on a positive note wherein an individual achieves some level of improvement in their situation or a new outlook on their problems, allowing the reader to see the value of making change. We sought to make these conclusions as realistic as possible; instead of claiming that characters had made tremendous progress or completely resolved their problems, our stories showed that small steps can be made to improve one’s situation.

Findings:

Participants provided a range of opinions regarding the appropriate length of stories. Four interview participants preferred longer stories that provided more context, detailed explanations, and clear examples. FP6 said,

“When you receive just a short like one sentence, it’s very easy just to like read it and be like, ‘Okay, I’m done.’ But if you force someone to read five sentences, then they’re much more likely to like take those like extra 10 seconds to process what they’re reading.”

Participants in the focus groups also felt that people would be able to more easily connect with messages that had more detail, expecting that shorter messages would be “vague” and “unclear”. In contrast to these views, FP2 and FP5 worried that longer stories would be broken across a long series of consecutive text messages, reflecting that most text messaging services limit the number of characters per message. Breaking stories across too many messages was predicted to lead to annoyance, particularly since many people set their phones to vibrate or ring whenever they receive a notification. FP5 speculated that seeing and hearing many notifications in quick succession would be overwhelming, especially for people who are preoccupied with other tasks. Another point that was raised by participants was that it would be nice to have a mechanism for customizing the kinds of stories that users would see. FP7 expressed that the feeling of control would create a sense of ownership over the system, which would keep them engaged.

Implications for Our Deployment:

To investigate the importance of message length, we created short and long versions of each story with either 3 or 5 messages, respectively. Most of the consolidation for the short versions came at the beginning of the stories; while the long stories used a couple of messages to set up the context and describe the characters and their problems, the short stories only used one message for that purpose. The short versions were also condensed when the character reaches a resolution to their problem, but both the short and long versions had similar amounts of text at the end dedicated to explaining how the story illustrates a common cognitive distortion.

Participants also wanted the ability to control the content of the stories that they read. We also experimented with a couple of different mechanisms for allowing participants to control the content of the stories that they read:

• Up-Front Choice format: With this approach, participants were given a single sentence or question to gauge their interest in a particular story (e.g., “Do you ever feel like everything is going wrong at once?”). If the participant expressed interest in that topic by responding “yes”, they were sent the rest of the story; if not, they were sent another prompt and the process repeated until a suitable match was identified.

• Start and Switch-Out format: With this approach, users received the first 3–4 messages of the story and then an additional message asking them if they wanted to continue hearing the rest of the story. Users who requested more of the same story would see the rest of the story that they had already started, while users who requested a different story were given a new story on a different psychological principle.

Our rationale behind both of these formats was that seeing a short teaser or segment of a story would be less overwhelming and would also give users enough information to decide if they want to engage with it. Although the Up-Front Choice format arguably allows the greatest control in selecting a relevant story, it also demands more action from the user at the outset. Therefore, we were interested in how users would respond to both strategies. Figure 1 illustrates the key interactions in these two approaches. Note that the Start and Switch-Out format was only applied with longer stories since there is no need to extend shorter ones.

6.2.1. Authenticity.

Participants believed that the stories were authentic and actually came from a real person. When we asked them why they believed this to be true, some noted the level of concrete details within the stories. A few reported that they found the stories authentic because the scenarios they described mimicked what was happening in their own lives. For example, P27 coincidentally took an economics class just like the central character in the overgeneralization story and was thus able to imagine themselves in that character’s situation:

“As the person was describing how they were trying to look for ways to snap themselves out of it, I saw those as coping mechanisms that anybody could use with the things that they were trying to do.”

Others expressed that although they had not gone through similar experiences to the ones presented in the stories, they could still connect to the feelings being portrayed. After reading the story on all-or-nothing thinking, P3 and P35 resonated with the feeling of being unproductive due to procrastination, despite being involved in different occupations and fields. Nevertheless, P11 acknowledged that it is impossible to come up with a small set of stories that resonate with people across all demographics, occupations, cultures, and circumstances; even so, they argued that stories with specificity could still seem relatable since most people have felt helpless or anxious at some point of their lives.

Contrary to these perspectives, our stories did not resonate with P37, who advocated for greater personalization:

“If you guys already had basic information about me, like I am out of school or working or whatever, and then gave me stories that were already tailored to what my life might be, then I think that would be really helpful and it would just resonate with me easier.”

On the whole, we observed that participants were more likely to regard stories as authentic if they described experiences that they had gone through themselves, either in terms of characters’ specific circumstances in life or their feelings and mental health struggles. In other words, participants often correlated relatability with authenticity. One suggestion we received for making stories more relatable was to leverage literary tropes since they are generally familiar to many audiences. An example of a broadly recognizable literary trope would be an underdog story in which a character is able to overcome significant adversity relative to their peers to achieve a successful outcome.

6.2.2. Concrete Takeaways.

Most of our participants appreciated the fact that each story was explicitly connected to a concept from psychological literature. For many, the connection transformed reading the text messages from being a casual leisure activity into being a productive learning opportunity with a concrete benefit: the lesson that was learned at the end of the story. P30 mentioned that “putting a name [psychological term] to the story” was helpful because it gave them the knowledge and terminology they needed to conduct further online research. Other participants felt that without the concrete takeaway message at the end, they may not have known how to interpret the stories or make use of them. P28 stated,

“I like that stories were tied to a principle because if this was used with a younger crowd, like teenagers, and if the teenagers I know read this, they’ll be like ‘Why did you make us read this?’. So you have to connect it back to something so that people understand why it’s being told.”

On the other hand, P3 felt that the explicit connections to psychological concepts detracted from the emotional experience of reading the stories and made them sound too prescriptive. They went on to say that stories should have room for interpretation so that people can decide for themselves how to best apply the lessons to their own lives. P13 felt similarly and said that stories should leave room for interpretation since “psychology does not have a definitive solution for everyone’s problems”. These findings suggest that more work is needed to understand how explicit messaging should be when connecting stories to the principles they are meant to illustrate.

6.2.3. Active Participation.

We added the reflection prompt to encourage participation from users and help them to reflect on their life experiences similar to the stories. In the interviews, participants like P11 noted that the writing activity was a “fun break” from their other work and school obligations. Other participants expressed that the action of writing responses to the reflection prompt made the stories feel more interactive and helped them better understand a persistent source of troubles in their lives. Some participants went as far as suggesting that the reflective prompt should explicitly request users to share stories that they think would help other people, creating a sense of community with the broader userbase.

People who did not respond to the reflection prompt noted that they did not always have the time or energy to write about their life, although some of them clarified that they did not necessarily dislike the prompts. In some cases, people told themselves that they would find a better time to respond to the prompt, only to forget later in the day. Another common complaint was that some people had difficulty writing long messages on their mobile phones. P7 said,

“It’s kind of hard to text or message a lot on the phone. The reason why I had done the writing is that I could do it on my computer since my phone and computer are linked, so it was easier to do it that way. I wouldn’t have done it if it was just on my phone, because there’s just too much typing and if you make edits, it’s just too much.”

6.2.4. Balance between Positivity and Realistic Struggles.

Participants generally desired stories that had the reasonable balance between portraying the struggle of the character and inspiring hope in the end. P20 commented that although having a positive ending results in a psychologically healthier mindset, reaching an improbable resolution to complicated issues (i.e., a fairytale ending) could be dissatisfying. In this regard, P39 appreciated acknowledgements of people’s struggles, saying:

“I found the stories helpful because they weren’t trying to spread toxic positivity, like ‘ignore bad thoughts’ or ‘concentrate on all the good’. It’s about acknowledging the bad as well.”

Some participants felt that our stories should have also covered more sensitive topics like child abuse or sexual harassment. For example, P27 posited that our text messaging service would have significant benefit for them if our stories acknowledged traumatic experiences like child abuse and illustrated how people have overcome such experiences. Nevertheless, participants suggested that stories with sensitive topics should be designed with caution so that they validate the readers’ struggles without inducing secondary trauma.

6.3.1. Scaffolding Feelings of Connection.

Participants often reported that they felt connected with the story characters, with the degree of perceived authenticity playing an important role. The authenticity of our stories led our participants to recognize that there are other people in the world who are going through similar experiences. Moreover, the feeling of connectedness gave them the confidence to overcome their struggle. In the words of P34,

“Remembering that other people experienced the same thing is encouraging. It reminded me, ‘This is life, I can deal with it.’”

Finding connection with others also gave participants a new perspective on their own problems. P36 reported that due to the COVID-19 pandemic they could not reach out to friends to talk about how they were struggling academically, but they perceived that reading the stories and answering the reflection prompt served as a sort of replacement for supportive discussions. On a different note, P20 felt that the story characters were experiencing even more severe problems than them, which gave them confidence to tackle their relatively less difficult life issues.

6.3.2. Adapting Stories to Personal Contexts.

The reflection prompt that we added after the story encouraged people to identify and elaborate on similarities between the characters’ situation and their own. Many participants voluntarily shared their own life experiences after this prompt, writing in detail about their struggles and their negative thought processes. For example, P1 connected the cognitive distortion of fortune-telling to their relationship with their roommates, finding that they had made assumptions about what their roommates were thinking and feeling without ever testing these assumptions:

“I can easily tell when my roommates use some of my things that are not theirs (e.g., olive oil, body wash, etc.) and I often think that me confronting them about it will make them dislike me and use my stuff even more out of spite — because in my head, I think they’ll say I’m being uptight since they only use a little bit at a time. ”

Participants informed us that the stories gave them a template into which they could insert their own experiences to better visualize how certain behaviors affected themselves and others. In P12’s words,

“I want to simulate people in my brain. I prefer to know their initial state. I want to know what their input and output was, given such a problem exists. I want the information to be complete so I can do a simulation more properly. It helps me figure out how I can apply those situation in my own life.”

In some cases, placing one’s situation into the story format helped people identify faults in their behavior for the first time. For example, P14 was not familiar with the concept of fortune-telling before reading the story on that topic, nor did they realize that they had the habit of expecting bad things from any situation; the story made them aware of their own pattern of thinking despite the example being in a different context.

6.3.3. Exploring Alternative Solutions.

In addition to reflecting on their experiences, participants also used the reflection prompt as a chance to devise alternative ways of approaching their problems. For example, P16 read the story about all-or-nothing thinking and subsequently reflected on their approach to online courses:

“When needing to complete school work, there are times where I’m just not in the mood to do so even though I know something is due soon and needs to get done. I usually end up spending hours with distracting myself and forget all about what needed to be done in the first place. The thing that I forget is I should always leave myself with plenty of time between assignments and I know this, it’s just that I tend to forget that fact when I get into one of these moods.”

The stories also had continuing impact after participants wrote out their thoughts. For example, P27 informed us that the overgeneralization story stayed in their mind a few days after they had seen it. The story made them realize that they should not have manage their situation alone, and prompted them to reached out to their social circle for support. P27 also noted in the interview that they planned to deal with future situations differently.

6.4.1. Message Format.

Table 3 suggests moderate to high levels of engagement for different formats, but we refrain from making claims about a preferred format due to the small sample size and potential confounds (e.g., for those in the two-week deployment, the Start and Switch-Out format was received after the Up-Front Choice format, which may cause order effects). Instead, we rely on the qualitative data from our interviews to understand the potential benefits and challenges related to different formats. In our interviews, participants appreciated that the Up-Front Choice format allowed them to choose a story topic instead of directly starting a story. Participants like P4 and P33 expressed that choosing the topic of the story gives a sense of control and helps them to find relatable contents. However, P3 felt that asking multiple questions before the actual story may make users “overwhelmed”.

Participants provided diverse opinions about the Start and Switch-Out format. P34 felt that this approach gave them the opportunity to see the actual story, with the choice to continue or to switch out a story allowing them to dismiss content they found boring or unrelatable. The people who did not appreciate the Start and Switch-Out format usually felt that the story was cut off in an awkward position when they were asked about continuing, potentially leaving the stories unresolved. Since the first few sentences of each story introduced the characters and their struggles, participants who never responded to the prompt for Start and Switch-Out format prompt only saw the negative setup for the story without any positive resolution. P19 said,

“Don’t just tease something horrible. And then, you know, if you never see more of the story, all you’re left with is something horrible.”

To address this, P12 suggested that the text messaging services could continue the story anyway after a certain time of non-response so that stories do not go unresolved.

6.4.2. Message Length.

Participants also had different reactions to the length of the stories we sent them. The qualitative data suggests that not all users were satisfied with the length of the stories they received. Some participants preferred shorter stories and felt overwhelmed when they received many consecutive messages. P39 recounted how receiving a long story made them feel like an emergency had arisen:

“I am often in classes or meetings. If I get six or seven notifications in the middle of a class, I would probably think, ‘Oh my God! What has happened?’ I would probably panic, and if I see these text messages then, I would be very annoyed.”

Another reason that some participants preferred shorter stories was because they felt that longer stories required them to “waste their brainpower” (P8) that could otherwise be applied to their work and responsibilities.

In contrast, people who enjoyed the longer stories appreciated the level of detail they included. Rather than having a single sentence to introduce the characters and set up the narrative, these stories used 2–3 sentences to convey rich information and draw the user in to read further. These participants felt that short stories were lacking context and ended too abruptly. P13 commented,

“I want to know the person’s initial mindset or his initial conditions. And how exactly the scenario leads to whatever decision he made. At least for me, I think that’s how I simulate people in my brain. I prefer to understand their initial state and then their input and output, given such a problem exists. I want the information to be complete so I can do a simulation more properly.”

Because of the perceived effort it took to read stories, particularly longer ones, some participants delayed engaging with text messages until later in the day. There were also situations when people were either not around their phone or were preoccupied with other tasks. Regardless of the reason, participants encountered two problems while engaging with stories after-the-fact: (1) the story messages were buried in a collection of other unread messages, making them difficult to find; and (2) people struggled to find the start and end of each individual story Hence, participants requested some way for them to differentiate stories from other messages. P8 came up with one solution, suggesting that each story could be wrapped in a unique symbol (e.g., =, +) as a delimiter.

7.1.1. RQ1: Applying Mental Health Principles into Real Life.

We used stories as a way of engaging people’s attention and providing concrete examples of seemingly abstract concepts, thereby reducing the cognitive burden associated with other mental wellness interventions. Our investigations enabled us to identify desirable properties of narrative interventions that would motivate people to apply mental health lessons in their lives. The authenticity engendered by the realistic portrayal of events enabled participants to connect with the story characters and gave them the required confidence to overcome their own struggle. The concrete takeaways at the end played an important role in helping people translate the stories’ lessons into their own lives. The explicit prompts for reflection went a step further by encouraging people to visualize the potential benefits of making changes in their own pattern of thinking. Together, the concrete takeaways and prompts for reflection not only reduced the cognitive effort needed to interpret the stories but also mitigated the potential for misinterpretation. That being said, some participants preferred a less prescriptive presentation of stories and wanted to interpret the stories on their own when they had the cognitive resources to do so. A potential way to resolve this tension could be to present the concrete takeaway message after first giving people the chance to contemplate the story on their own through a reflective prompt.

7.1.2. RQ2: Eliciting Benefits of Self-Reflection through Story-Based Text Messages.

Since automated text messaging interventions do not have a live conversational partner to motivate users, we took additional steps to build interactivity into our messaging program and to elicit active participation with our intervention — most notably adding a reflection prompt to encourage people to write about their own experiences. Some people used the prompt as an opportunity to draw connections between their own struggles and those of the story characters, ultimately giving them increased confidence in solving their own problems. Other people used the prompt to assess their thought patterns and discover previously unknown cognitive distortions that had been affecting their decisions. Although we did not explicitly ask participants to seek a solution to their problem, some reported that they were often motivated to make changes after such reflections, including exploring new ways to frame their experiences and alternative actions they could take. Even when participants did not write anything, interview data suggested that the prompt still served as a valuable forcing function for users to read and reflect on the story. We therefore argue that a simple question at the end of a set of story messages can encourage self-reflection that may help users immediately after reading a story, and also beyond.

7.1.3. RQ3: Balancing Message Length and Story Depth.

Our investigations also allowed us to identify several design tensions between message length and story depth that should be considered for future deployments of story-based text messages. On the one hand, participants noted a greater degree of realism when they read stories with more specific details, such as the background of the characters and the actions that led up to their challenges; this finding echoes the observation from prior work that people seek some degree of honesty when they interact with computer systems [104]. On the other hand, participants expressed that text messaging interventions should also be respectful of people’s time and energy since they may be received during a variety of activities. Longer stories take more effort to read and digest, and mobile phones can only fit so many characters on the screen at the same time. The combination of these factors led some participants to feel distracted or overwhelmed by our messages. Some participants appreciated our Start and Switch-Out format because it allowed them to first see the beginning of a story, which allowed an informed judgment about whether they should continue or not. However, since the previews often centered on the characters’ struggles, they could also be off-putting at times. Future work could explore how to generate previews that recognize characters’ struggles while also foreshadowing the success they will have in addressing them.

7.3.1. Rethinking Text Messaging Applications for Longer Exchanges.

There were some ways in which standard text messaging applications were not ideal for our envisioned interactions. Participants who received longer stories were overwhelmed by the amount of screen space and scrolling it took to read the messages [37], while those who wrote longer responses were forced to switch back-and-forth between the story and the text they were writing [115]. A dedicated smartphone app can address all of these inconveniences to better accommodate lengthier interactions. In fact, an app could even extend the benefits of narrative-based interventions by including multimedia content that makes the stories more engaging with rich multisensory experiences [127]. However, text-only interventions invite readers to use their own imagination to visualize the stories, simultaneously encouraging greater participation and providing greater flexibility as to how readers personalize the scenarios for themselves [41]. The decision of whether to use the default text messaging application or a separate dedicated app warrants further investigation and may require consideration of the target population. For example, text messaging might be one of the very few ways to reach non-smartphone users or people with limited access to smartphones [115].

7.3.2. Personalizing User Experience.

Past work suggests that personalization and variety can increase engagement with stories [114]. There are several promising avenues for engendering these features into our envisioned intervention. Although most participants in our deployment study were satisfied with the relatability of our stories, they also appreciated opportunities to customize stories, and participants made moderate use of two techniques for story topic selection in our deployment study (i.e., selecting a topic at the beginning of the interaction, and switching out a story for a more relevant one partway through). Future interventions could further build the extent of customizability of stories. For example, stories may allow users to “choose their own adventure” by making decisions in multiple junctures about the way the story should unfold (e.g., picking not only the topic, but also the actions taken by characters), which may allow a more game-like and engaging experience while allowing users to narrow in on the content that interests them the most. Personalization can also be applied to the structure of the stories. Some of our participants appreciated how the concrete takeaway message at the end of the stories helped them interpret the content, summarizing and reiterating points that could otherwise be lost in a thread of messages or matching the straightforwardness of most text messaging interactions. On the other hand, other participants found the takeaway message to be too prescriptive and wanted to come to their own conclusions. Although past literature has noted the benefits of indirect pathways to persuasion [80] and collaborative approaches to sense-making like collaborative empiricism [12], our findings highlight that directness has considerable value for some people. Future work might therefore examine how text messaging systems can adjust the directness of the messaging to people’s preferences. Lastly, personalization can be extended to the timing of the stories, as users are more likely to engage with their phones at certain times (e.g., weekend, non-working hours) than others [114].

7.3.3. Enabling Research on the Web.

The concrete takeaway messages at the end of our stories not only helped our participants interpret the content of the stories, but also provided them with vocabulary for situations or feelings that were previously difficult for some people to put into words. In doing so, participants felt that they were better equipped to search online for other helpful resources and examples. This opens up the possibility for stories to serve as a gateway to more expansive online resources, broadening the potential benefits of text messaging services for mental health support. However, due to the abundance of information about mental health online, people might still stumble upon unvalidated or counterproductive advice. Therefore, future work could investigate whether stories could be improved by integrating hyperlinks that lead readers to carefully chosen or crafted online resources.

7.3.4. Considerations for Large-Scale Deployments.

To sustain engagement for more than a few weeks and thereby support a broader set of cognitive interventions, there should be a mechanism for regularly generating new stories. Limiting story creation to a small group of individuals is not only inefficient but would also limit the diversity of the story content. One approach to facilitate large-scale story generation that reflects firsthand understanding of mental health problems would be to leverage users’ own content [61]. Many users in our deployment study willingly shared detailed life experiences in response to the reflection prompt, which suggests potential for intervention users to crowdsource content during their use of the system. Taking motivation from literature on guided identification of cognitive distortions [86], rich stories could potentially be elicited through structured multi-step prompts that ask users to define their problems, elaborate their experience with the problem, and identify aspects of their behavior or lifestyle they changed to address the problem [54]. Since multi-step prompts require more effort from users, it will be important to understand how users feel about providing such detailed information and sharing it with others. To reduce burden, story generation could be suggested on an infrequent basis (e.g., once every 1–2 weeks) or only to participants who are highly engaged.

As mentioned earlier, another source of stories could be dedicated crowdsourcing platforms like Amazon Mechanical Turk (MTurk). Crowdworkers could be asked not only to create stories, but also rate them according to design goals like authenticity and proper balance of positivity and negativity [48, 61]. The provided ratings could be used to both identify stories that require iteration and examples that can be shared with other story creators. Promisingly, some prior studies have found that non-experts can produce content of comparable quality to experts, while also being much more efficient [13, 78]. However, it is important to note that crowdworkers may lack understanding of psychological strategies compared to mental health treatment experts, and it may be important for clinicians to continue to play a role in curating and editing stories to ensure they accurately represent psychological concepts and strategies. Highly engaged crowdworkers could also potentially be trained to recognize high quality stories, and could play a role in creating, rating, and revising others’ stories over time [78]. Regardless of how users are prompted to provide stories, special considerations should go into maintaining users’ privacy and acquiring their explicit consent before disseminating their stories to others.