“I finally felt I had the tools to control these urges”: Empowering Students to Achieve Their Device Use Goals With the Reduce Digital Distraction Workshop

Most research on DSCTs has focused on use of a single device, typically the smartphone (see Roffarello and De Russis [95] for a review). In practice, people’s challenges with self-regulating use often involve multi-device contexts [18, 81, 96], where unwanted use curbed by a DSCT on one device might ‘spill over’ to another [56]. Some researchers have started to explore ways to take multi-device use into account [47, 56, 82, 93]. This work suggests that DSCTs implemented on one device can indeed — when assessed on screen time metrics — help people make progress towards their goals without causing negative second-order effects [56]. However, which evaluation measures to use remains an open question [8].

Many DSCTs are explicitly designed to limit time spent [e.g., 48, 51, 52], and the most common outcome measure in efficacy studies is time spent overall or in specific apps [95]. This to some extent aligns with users’ goals, as surveys of smartphone use, as well as individual platforms such as YouTube, suggest that goals to reduce time spent are common [38, 52, 68]. However, time spent is often an unreliable indicator of whether a DSCT is helpful [39, 65, 88, 89]. Lock-out mechanisms, a common feature in DSCTs, illustrate this conundrum: whereas locking users out of devices after a time limit may seem effective when assessed by reducing time spent, it can lead to strongly negative responses when users find themselves in ‘out-of-routine’ situations where they need to use their devices [48].

The most frequently used subjective outcome measures in DSCT research are questionnaires related to ‘addiction’ (e.g., the ‘Smartphone Addiction Scale’ [59, 95]). This, too, to some extent aligns with users’ goals: analyses of publicly available user reviews suggest that the most frequent purpose of use for DSCTs is to help oneself stay focused on demanding or boring tasks amidst readily available digital distractions. This capacity is captured to some extent by addiction scales [72]. However, the ‘addiction’ framing has been criticised as imprecise, and for pathologising everyday self-control struggles [5, 14, 41, 60, 87, 110]. One recently proposed alternative (which we adopt in the present paper, see Section 3.2.2) is to adapt measures from basic psychology research on everyday self-regulation to the context of digital device use [71]. This might both help capture global ability to self-regulate behaviour in a more precise manner than the scales related to behavioural addiction, and help fill the ‘theoretical gap’ in DSCT research, by grounding measures in relevant psychological research [37, 73, 95].

A key consideration, as researchers explore measures of digital wellbeing to evaluate DSCTs against [8], is to ensure that those measures accurately capture specific populations’ usage goals and preferences for support [79]. Since most recent empirical investigations of the goals that DSCTs should serve have focused only on smartphone use, it would be useful to collect additional data on these goals and preferences in cross-device contexts. Moreover, open data sharing is rare in DSCT research, but might accelerate the development of empirically grounded measures, and allow for easier comparison between populations [69, 114].

Hence, our first use of data from the workshop focuses on the following question (RQ1): in the context of their personal ecosystem of digital devices, how do students want to change their use?

People vary substantially in the way they use digital devices, both compared to others and to themselves over time [16]. To empower people to influence their own behaviour using DSCTs, we therefore need effective ways to elicit individualised goals and apply DSCTs in bespoke ways to support those goals [86].

Work on ‘digital self-nudging’ suggests that empowering people to co-create their own DSCTs (e.g., using the Shortcuts automation app on iOS) can support a sense of agency, accomplishment, and perceived usefulness [92]. Active involvement is likely to be particularly important for DSCTs, as they often involve interventions that restrict the user. This can be an effective way to influence behaviour, but also cause psychological ‘reactance’, in which the user is motivated to circumvent or otherwise rebel against the intervention [74, 106]. This is likely to be reduced if the user freely commits themselves to a DSCT because it supports their personal goals and values, as opposed to having it assigned by experimental condition or algorithmic allocation [66].

The research to date, however, has almost invariably focused on a single intervention on a single device or service, pre-assigned to research participants [95]. Few studies have investigated the practical challenge for which stakeholders such as university counselling services urgently need advice: how can we help people identify and apply the right combination of DSCT(s) for their personal needs (see also [80])?

One notable exception is HabitLab [55, 57, 58]. This project deployed an in-the-wild intervention that helped more than 12,000 daily Chrome and Android users meet time limiting goals on websites and apps via a wide range of interventions, from hiding content feeds to displaying timers. However, a limitation of this work was that it measured success mainly in terms of limiting time spent, which is likely to have captured only a subset of people’s actual goals [68]. Broader HCI work on individualised strategies for self-monitoring and self-experimentation in behaviour change (specifically, ‘n-of-1’ studies for managing health conditions requiring individualised insights [42, 63]), suggests that this might be a problem: if individualised goals are not carefully elicited, people may end up with solutions that encourage behaviour that runs counter to their goals [86, 115]. Similarly, analyses of user reviews for DSCTs suggest that people are often frustrated that the tools they try do not match their goals for self-regulated use [72].

In sum, there is an opportunity to extend previous work with an approach that elicits individual goals and lets users actively choose and apply DSCTs. Hence, our second use of workshop data focuses on the following practical question (RQ2): Can an intervention that combines reflection on personal goals for change with self-selection of relevant DSCTs improve digital self-control?

Whereas DSCT research has focused on average effects of one specific intervention, many studies have also reported that people can vary substantially in their preference for, and benefits derived from, those interventions [47, 48, 76]. For example, Mark et al. [76] noticed in an exploratory study of information workers that the benefits of distraction blocking seemed to apply only to participants who at baseline scored lower on self-control. Getting a clear picture of the variance between people is highly practically relevant: should stakeholders like university counselling services direct most students towards a limited number of interventions that are universally useful, or should they support each student through a process to identify the tools that best meet their specific needs?

Psychological research on self-control more broadly suggests that the answer might be the former: applying the ‘process model’ of self-control, Duckworth et al. [26] found that strategies intervening earlier in the cycle of impulse generation (i.e., avoiding exposure to temptation as opposed to trying to overcome it only after unwanted impulses have arisen) are generally more effective [24]. This might also hold true for DSCTs. However, few studies have explicitly compared how multiple DSCT interventions vary in usefulness [95]³.

Again, one exception is HabitLab: Kovacs et al. [58]’s supplementary materials included a comparison of twelve different interventions on Facebook, finding that the most effective (auto-closing the tab after 60 seconds) was twice as effective as the second-best (blocking the site after a user-determined duration). However, the outcome measure was solely reduction in time spent. Given the limitations of this measure (according to which, the most effective solution would be to simply ban an app or website from being used altogether), interpreting their findings would be helped by triangulation with users’ subjective assessment of usefulness.

Hence, our third use of workshop data focuses on the following question (RQ3): How, if at all, do students’ preferences for different types of DSCTs vary?

We iteratively developed the workshop in collaboration with the counselling service at the University of Oxford, which works one-on-one with nearly 3,000 students each year. Between May 2019 and March 2020 we conducted five pilots of an in-person workshop. We had students engage in small-group discussions about their challenges around digital device use, and explore different types of DSCTs to address these challenges using a card sorting task [78, 105]. A website provided examples of how to apply DSCTs on specific devices. At the end of the workshop, participants verbally committed to try specific DSCTs.

After the onset of the COVID-19 pandemic, we used our experience from these pilots to develop an online version of the workshop. We conducted this version on Zoom, and used the online whiteboard tool Miro (https://miro.com) for components that required participant interaction. We conducted five pilots of the online version in November 2020. In March 2021, we made minor tweaks based on these pilots, made the final workshop version more widely available at the University of Oxford, and began collecting data for the present paper.

In terms of device types, 93% of participants said in the opening survey that they wanted to “change, or be more in control of”, their smartphone use. However, only a little more than one third (37%) were solely concerned with their smartphone use. 63% of participants wanted to change their use of other devices, mainly laptop / desktop computers (59%), with a small minority (7%) mentioning tablets (Figure 3 c).

The final reflection prompt, after participants had written down their concerns, triggers, and what they had previously tried, was “Imagine your use of digital devices is exactly as you want it to be. What does that look like? (Be specific about context, time of day, apps...)”. We captured participants’ responses to this question with two higher-level goals, alongside three more specific goals that we interpreted as means of achieving the higher-level goals. The goals were closely related to participants’ concerns, and the triggers that drove them, a summary of which is included in the Appendix, section D.

Among participants who filled in the follow-up survey (52% of total⁵), we observed a large increase in self-reported digital self-control, compared to before the workshop (Cohen’s d = 0.93, mean increase = 7.9 points, scale ranges from 12 to 60, t(145) = 11.2, p < .00001; Figure 4 c). The effect size was larger for follow-ups after 1 month (d = 1.13, n = 36) than after 3 months (d = 0.87, n = 110).

As is the case for all open trials, the increase in digital self-control scores could be influenced by a number of factors, including response bias (students who find the workshop more useful might be more likely to respond) and regression to the mean (students might attend a workshop at the point in time where they struggle the most with digital distraction). When adjusting for response bias by assigning non-respondents the same score as in their opening survey (‘last observation carried forward’ [33]), the estimated increase in digital self-control remained medium-sized, d = 0.56 (t(279) = 9.4, p < .00001).

Moreover, the quantitative data suggested that discovering appropriate DSCTs, rather than simply regression to the mean, was related to digital self-control improvement: among participants who had found at most one ‘very useful’ strategy (n = 108), the effect size of the digital self-control increase was d = 0.84. However, the effect size was substantially larger (d = 1.21) among participants who had managed to find two or more ‘very useful’ strategies (n = 38). Number of ‘very useful’ strategies discovered was unrelated to the number of participants in a workshop (r = -0.002, p = .98), but strongly correlated to the number of strategies participants had tried (r = .53, p < .001).

Finally, as indicated in Figure 4 d, participants’ self-reported daily time spent declined on laptop computer (r = .29, p = .001⁶, median difference = -60 minutes, interquartile range of differences = -120 to 60 minutes,), smartphone (r = .29, p = .004, median difference = 0, IQR = -60 to 45 minutes), and tablet (r = .38, p = .04, median difference = -25 minutes, IQR = -60 to 0 minutes), but not on desktop computer (p = .9).

Participants’ assessment of the usefulness of the strategies is shown in Figure 5. The strategies were not created equal: the proportion of respondents who rated a strategy ‘moderately useful’ or higher ranged from 26% (Replace distractions on the web with a to-do list) to 84% (Limit notifications), and the proportion of specifically ‘very useful’ ratings ranged from 5% (Replace distractions on the web with a to-do list) to 55% (Hide distracting features on websites). Conversely, the proportion of ‘not at all useful’ ratings ranged from 0% (Reduce your device to the tools you need) to 37% (Replace distractions on the web with a to-do list).

The top 6 strategies all seemed generally useful, receiving ‘moderately’ to ‘very useful’ ratings from at least 76% of respondents, and ‘not at all useful’ from at most 10%. All but one of these strategies were from the blocking or removing distractions category. The exception was Focus in bursts with a timer (which ranked second on proportion of ‘very useful’ ratings, at 49%).

The remaining strategies saw greater amounts of variation. For example, grey scale was the second-most frequently tried strategy, but had a somewhat uniform distribution of usefulness ratings and the second-lowest retention rate (54%). Importantly, no strategy was outright rejected: even the ‘worst’ strategy, ‘Replace distractions on the web with a to-do list’, which was found ‘not at all useful’ by 37%, was still considered at least ‘moderately useful’ by 26% of those who tried it.

We also observed a high degree of variation in how participants combined strategies. Among the 78% of respondents who still used 2 or more strategies at the follow-up (n = 114), 91% of the specific combinations of strategies were unique to a single participant. Thus, even the most frequent unique combination (Block distracting websites or apps + Hide distracting features on websites) was applied by just 4 respondents.

Based on the survey and interview data from the workshop, we identified several design considerations for researchers and practitioners working in the domain of digital self-control.

Illustrate full capability range of DSCTs to empower participants to identify tailored solutions. Compared to other domains of behaviour change, DSCTs have a unique power to not simply remind or recommend actions, but also to change the digital environment (an exercise app cannot support you to take the stairs by removing the elevator at your workplace, but a DSCT can hide an unwanted newsfeed [55]). However, most people’s considerations for changing digital device use are anchored by simple all-or-nothing approaches (blocking an app entirely or not at all) [111], a notion reinforced by the limitations of pre-installed screen time tools.

Therefore, a key benefit of the workshop, which participants repeatedly stressed, was that it made them aware of the range of actionable ways in which they could take back control. This empowered them to search for ways to implement a strategy tailored to their situation, even if the specific solution they needed was not among the DSCTs included on the workshop website. For example, knowing that it is possible to use browser extensions to modify websites, P11 deleted the Instagram app off their Android phone and used an extension to remove the search area when using Instagram in the browser. This effectively helped them use Instagram for messaging without getting distracted, but was not among the tools on the website.

Demonstrating the range of what is possible is therefore key, but complexity should be carefully managed to avoid overwhelming participants. For example, P198 felt that the number of strategies included was “too many. I think you should have done eight”. Others wished they had more time to review the options before being asked to commit. Our workshop implementation attempted to reduce cognitive load by using layers of abstraction, combined with decision steps that gradually honed in on specific options. For example, we grouped DSCTs into 13 strategies, which themselves were grouped into just 4 main categories; the card sorting task supported initial reduction of options; and we limited the number of ‘commitment cards’ on the Miro board to just two for each participant, to help funnel towards action. While not perfect, this approach seemed sufficient to manage information overload for most (“the strategies part and definitely, I think, the commitment part (...) those two together sort of symbiotically were high impact for me, the commitment specifically because it forced me to narrow down my choices”, P118).

Scaffold self-experimentation in a way that is sensitive to how DSCTs differ in attrition challenges. Researchers have suggested that DSCTs face a general challenge of attrition over time [57, 58, 84]. Our interview data suggest⁷, however, that the nature of this challenge differs between tools. For tools that require active initiation, like focusing in bursts with a timer, participants described the usual challenges of sustaining use over time. However, strategies that were set-it-and-forget-it could benefit from inertia (“the list app replacing the YouTube newsfeed. That’s not something I have to go up and change again every morning, which made it very easy to stick with”, P118) and help them habituate to a desired state of the world (“it doesn’t even cross my mind anymore when I use these apps, it’s like I just don’t have a newsfeed anymore”, P86). This benefit seemed to apply specifically to the aspects of a set-it-and-forget-it intervention that involved removing distracting elements (“The motivational quotes themselves weren’t particularly helpful or inspirational, but the fact that they covered my Facebook and YouTube feeds was far more useful.”, P237)

From our participants’ descriptions of the post-workshop period⁸, we conjecture that strategies which involve active initiation should be tried one at a time to avoid feeling overwhelmed, and allow time for a new habit to form (e.g., P59 applied multiple things at once, but described in the interview that this approach had led him to not give each a proper try). We also note that whereas some DSCTs are naturally distinct in this respect, many can be used in both an actively initiated as well as set-it-and-forget-it way. For example, tools for blocking distracting websites or apps can often be used either actively to initiate a delineated blocking session, or be prospectively programmed to block distractions during particular hours and/or days of the week. Our data suggest that people in the set-up phase should be encouraged to consider set-it-and-forget options where relevant (e.g., setting up a weekly blocking schedule), to make inertia work to their advantage.

Prioritise self-report measures that capture individual differences in goals. Most participants wanted to change both their smartphone and laptop use, and had varying goals including gaining a sense of control, establishing clearer usage boundaries, and reducing the duration or frequency of use in various ways. Instead of introducing multiple distinct measures to capture this wide range of goals, which would make surveys overly long, we included a broader measure applicable across multiple goals and devices. Considering the range of goals our participants reported, the self-control scale we adapted from basic psychological research seemed a better fit than commonly used alternatives in DSCT research, which tend to focus on a single device (e.g., smartphone addiction scale [59]), a specific system (e.g., the NASA-TLX task load scale [35]), and/or a narrower goal (e.g., cognitive absorption [6]). Our study is a first use of this type of scale in DSCT research. We expect this or similar scale adaptations of standard self-regulation measures from psychology to be useful for other researchers in this field looking for alternatives to ‘addiction’ measures [71].

To interpret the scores from this scale, we suggest triangulating with other data. In our case, we used quantitative and qualitative survey data on the usefulness of specific DSCTs, as well as self-reported time spent on devices. Although our surveys did not ask participants to provide any objective measures (e.g. a screenshot of their devices’ screen time report), including them may aid researchers in interpreting the self-report data. For instance, the decline in subjective time spent on devices at the follow up might not correspond to actual time spent [27, 28, 99].

However, we strongly suggest that objective usage measures such as screen time are mainly deployed in the service of the subjective outcome measures. One reason for this is that assessing DSCTs based on objective outcome measures such as screen time may lead to overly narrow conclusions; interventions that block devices from being used altogether may be highly effective for reducing screen time, but interfere with intended use. Moreover, there are no universally accepted guidelines as to what constitutes ‘healthy’ screen time for adults [113], so we instead recommend prioritising self-assessments of whether device use is aligned with one’s goals. Indeed, even if we did have objective standards for how adults ‘should’ be using digital devices, there would be strong reasons to keep a focus on subjective sense of control and self-efficacy within digital environments [70]: self-determination theory posits autonomy as a basic human need and common principles in HCI design guidelines encourage support of user control as a goal in its own right [22, 68, 98, 102].

HabitLab demonstrated that applying a range of relevant DSCTs cross-device (Chrome on PC + Android) can effectively reduce time spent on websites and apps that a person wishes to use less [56]. Our study supplements this work (which was based on objective screen time measures) with self-report data, and confirms the potential to support digital wellbeing via bespoke, cross-device applications of DSCTs [82].

Our findings were generated by triangulating quantitative and qualitative self-report data from a relatively large sample of students. However, as is the case for most studies of DSCTs [69, 95], our study did not have a control group. To assess the magnitude of the potential benefits, follow-up studies with, at minimum, a waitlist control group are needed [61, 107]. Such work could also use automated rather than self-reported screen time measures to obtain a more accurate indication of usage [27, 28, 99].

Moreover, as has been encountered in studies of DSCTs, our data suggested that the effect of the workshop waned over time. Bringing about longitudinal behavioural change is a general challenge for behaviour change interventions [56, 90], but long-term evaluations are rare in DSCT research [95]. The practical approach of the present work, where we attempted to address a local challenge in a student population actively seeking out the workshop, may represent an opportunity in this respect: in the interviews, participants commonly requested that the workshop be expanded with recurrent group support (with suggestions ranging from weekly to quarterly follow-ups) to help themselves stay motivated and overcome technical challenges. Deploying regular workshops and/or ‘booster’ sessions might be an effective way to undertake longitudinal studies of self-experimentation habit-formation with DSCTs [36].

Finally, we designed the workshop to holistically integrate components that previous research suggested would be effective for behaviour change [12, 30, 62, 86, 103]. The qualitative data suggested that the components worked well together: the group setting normalised digital self-control struggles, the reflection clarified participants’ goals, and the exploration and application of DSCTs provided actionable steps. However, controlled studies could assess the relative contributions of different elements and facilitate iterative development of variant interventions [49]. For example, group support is often a crucial factor for supporting behaviour change, and in our study helped normalise digital self-control struggles. However, a self-guided version might be valuable for people who are not willing, or able, to take part in a live group setting and/or for stakeholders who lack the resources to train a facilitator. How might the workshop elements be effectively delivered in a form which does not require a live facilitator and/or group context? Investigating such questions will be highly practically relevant for stakeholders such as university counselling services to implement cost- and time-effective interventions to support digital wellbeing [80].

Comparing how different DSCTs fare in practical settings may provide important data for crafting interventions that effectively guide people towards the strategies most likely to be useful [25]. To this end, our work provides one of the first direct comparisons of subjective usefulness of multiple DSCTs.

One takeaway is that hiding distracting features on websites is one of the most consistently useful strategies in DSCTs (highest proportion of ‘very useful’ ratings in our study). This confirms suggestions from previous work, which has encouraged researchers to study ways to provide ‘internal support’, i.e., changing problematic user interfaces directly, as opposed to simply provide ‘external support’ such as usage tracking or blocking [68, 74, 117]. Similarly, while not directly comparable, our ranking of strategies’ usefulness showed similarities to Kovacs et al. [58]’s comparison of 12 different interventions’ influence on time spent on Facebook. Here, the five most effective interventions all related to blocking or removing distractions, from force-closing the tab after 60 seconds to removing the newsfeed or comments.

When applying a theoretical lens to understand the patterns we observe, our findings broadly align with the ‘process model’ of self-control widely used in psychology [26, 31]. Thus, our overall pattern in which strategies from the ‘block or remove distractions’ category were most likely to be found useful, and strategies within the ‘goal reminders’ category least likely, mirrors the finding that self-control strategies which help people avoid exposure to temptation, rather than trying to overcome unwanted urges only after they have arisen, tend to be more effective [24]. Intervention designers might wish to keep this in mind.

However, our data also suggest that additional theoretical lenses, such as the dual systems model previously used in DSCT research [48, 73, 74], are important: focus in bursts with a timer had the second-highest proportion of ‘very useful’ ratings, even though it was not about reducing exposure to temptation. Rather, this strategy seemed to help people maintain control by breaking up their focus into smaller segments that feels manageable and easier to get started with. From a dual-systems perspective, this can be analysed as a way to support conscious System 2 control, by increasing confidence in one’s own ability to maintain focus [73].

We encourage further research to directly compare the practical usefulness of different DSCTs, to better understand their relative challenges and opportunities, and how each may fit within a unique personal device ecosystem and behaviour change process [73, 91]. In turn, this may help researchers generate recommender systems to support self-experimentation with DSCTs (e.g., ‘People who reported challenges similar to yours have found this useful’), for which we hope our open dataset can provide a useful starting point.

Overall, our participants expressed three more practical goals — keeping boundaries around time and context of use, reducing amount and frequency of use, and changing digital environments such that self-regulation is easier in the first place. Support for the two former goals is well underway: Most DSCTs focus on limiting time spent and/or frequency of use [95], and support for scheduled or geofenced notification limiting or app blocking is both possible and an active area of research [3, 46, 104].

As mentioned, however, research on DSCTs that provide ‘internal support’ and directly reorganise problematic within-service features has been more sporadic, even though it is one of the most potent strategies at our disposal [68, 74, 95, 117]. We want to draw attention to a major infrastructural challenge in this respect: the ‘hide distracting features’ strategy, with which people can directly address problematic aspects of their digital environment, was almost exclusively applied on laptop or desktop computer. This was not accidental. In principle, participants could apply this strategy on smartphone. But on mobile, people are used to accessing services such as social media via apps, browser extensions are less well supported (Chrome does not support extensions on Android), and some services (such as Snapchat) are not accessible in a web browser at all. However, no simple solution currently exists for users to adjust user interfaces in mobile apps according to their personal needs, beyond what the developer chooses to provide by design [53, 54].

Lukoff et al. [67] recently demonstrated the benefits that this might bring: changing the YouTube mobile app such that users can turn recommendations on or off increases sense of agency, satisfaction, and goal alignment. However, it requires substantial technical effort to re-implement even a limited version of the YouTube app for a single research study. Other recent work has explored technical prototypes that allow end-users to change apps by themselves, and discussed the risks and opportunities if regulators introduced a ‘right to repair’ for mobile apps (i.e., a right for end-users to exercise granular control over their user interfaces) [53]. Our findings suggest that pursuing this avenue of work could be highly valuable. We encourage researchers and regulators alike to discuss how an equivalent to browser extensions might be created for mobile apps, which could help address an infrastructural bottleneck for large-scale impact of DSCTs.

Sampling: All participants actively chose to sign up to the workshop. Hence, the distribution of challenges and goals that we observed, or of choice and assessment of DSCT usefulness, may not generalise to students, or other target populations, at large.

Follow-up response rate: 52% of participants filled in the follow-up survey. We conservatively adjusted for response bias by carrying forward the last observed digital self-control scores from non-respondents [33], but future work may consider how to increase response rate (e.g., by collecting data at recurrent in-person group support sessions, rather than solely in follow-up emails).

Subjective outcome measures: Our follow-up data included subjective measures only, collected via surveys and interviews. Self-reported screen time in particular, can correlate poorly with actual time spent [27, 28, 99]. We do believe subjective sense of time spent is valuable in its own right, not merely as a flawed proxy of objective time spent. That is, if a person has a goal to reduce time spent on their devices, it is helpful to know whether a subjective increase in digital self-control is accompanied by a subjective decrease in time spent. Nevertheless, in the best of all worlds, researchers may wish to collect both objective as well as subjective data, for example by asking survey respondents to provide numbers from Screen Time or Digital Wellbeing features on their devices, which could be triangulated with the qualitative data.

Open trial: Our study did not include a control condition, and so we were unable to quantify the influence of factors like regression to the mean or response bias. Future work could more accurately estimate the benefits of the intervention by comparing with a waitlist condition, and/or partial or modified versions.

Selection of DSCTs: Our grouping of DSCTs into 13 strategies represented our best attempt at summarising design features in current tools, in a way that was manageable for participants to consider in a card sorting exercise. The DSCTs could be grouped in other ways, and other options for specific tools than the ones we included on the workshop website could have been made. The tools we included in the present version of the workshop can be found on the Open Science Framework (osf.io/zmt78 and in Appendix C.

Device use diverts time and attention away from personally meaningful activities (95% of participants). This was most commonly expressed in relation to productivity (68% of participants), but also in relation to other valued activities such as personal hobbies or social interactions (22%). Thus, participants were concerned about the amount of time they wasted on meaningless activities on their devices (34%), because it left less time available for other things (“Waste time that could be used writing thesis (...) prevents me from investing time in learning new skills”, P117). They were also concerned that the frequency with which they, e.g., checked social media, interrupted their focus on work and studying (“Disrupting my work; making it take longer than it should do, and it being of a lesser quality”, P47), as well as on social activities (“They prevent me from connecting to the present moment / to other people (in person)”, P98).

Feeling unable to control, and dependent on, device use (52% of participants). Participants described that they found themselves using or checking their devices excessively without awareness (“Finding myself browsing social media without thinking”, P142), and/or that they felt ‘sucked in’ and unable to stop using their devices (“Looking at YouTube starts with something for work/study and then you can go down a rabbit hole because of clickbait videos”, P68, “Once I start, I can’t stop”, P61). Many felt uneasy about the dependency and pervasiveness of digital devices in their lives (14%, “I feel like I cannot leave the house without my phone”, P31), and were worried that their behaviour over time made them addicted to stimulation (7%).

Device use negatively impacts mood or mental health (43%). Participants were concerned that the content on, in particular, social media or news negatively affected their mood (“the content of the social media platforms can sometimes make me anxious or in a ‘negative space’”, P273, “Social media makes me feel inadequate”, P14). They also expressed feeling overwhelmed by the sheer amount of content/notifications/emails/messages they had to respond to (“I always feel behind on messages & guilty [...] It makes me anxious that people can contact me at any time”, P190; “Too much information that is not relevant”, P67).

Device use harms sleep or physical health (31%). Finally, participants were concerned that device use disrupted their sleep (20%, “I spend a lot of time on YouTube before bed, and that has been taking away my scheduled sleep time”, P59). Similarly, time spent on digital devices might distract them from healthier physical activities (“It ends up disrupting my exercise, food and sleep schedule”, P48, or even negatively affect their eye sight or posture (“I’m worried about my eyes because I spend so much time looking at a screen”, P25).