Abstract
While permanent connectivity has made media use ubiquitous and apps have become engagement optimized, little research has focused on the termination of (mobile) media sessions. This study addresses this shortcoming by exploring how internal, environmental, and media context cues situationally prompt users to stop using an app. In an event-based experience sampling study, 118 participants reported on disengagement from TikTok or Instagram (T = 1,893 sessions). We identified five disengagement types through multilevel latent class analysis: non-self-determined disengagement, self-determined disengagement, disengagement through priority shifts, temporary disengagement, and “mindless” disengagement. Across the types, competing activities, goal achievement, and push notifications emerged as the most prevalent cues and disengagement mostly occurred as an effortless solution to activity conflicts. In one type, it was accompanied by negative emotions. This typology contributes to a nuanced understanding of disengagement, suggesting that while worrisome forms exist, they do not dominate disengagement in day-to-day life.
Introduction
In mobile environments, media can be used nearly anytime and anywhere. Content is increasingly immersive and seemingly endlessly arranged (Anderson & Wood, 2020). Hence, built-in boundaries indicating when to stop mobile media use almost disappear. This leads to the question: Given that we can access mobile media anytime and without constraints, what prompts us to discontinue a mobile media use session?
An encompassing understanding of disengagement (i.e., terminating a use session) is largely absent from communication science to date. Nonetheless, it is as important as media selection for a holistic view of media use processes. While research on digital disconnection is growing, it primarily focuses on intentional reductions in media use (see Nassen et al., 2023 for an overview). How and why media use is stopped in a specific situation has rarely been addressed theoretically (Reinecke & Meier, 2021) and empirically (for exceptions, see Baumgartner & Kühne, 2024; Gilbert et al., 2024; Rixen et al., 2023; Tran et al., 2019). Since media use in general and mobile media use specifically strongly vary situationally and intraindividually (Schnauber-Stockmann, Scharkow, et al., 2025), we assume that there are also different types of disengagement in day-to-day life. These disengagement types present typical cue patterns based on the co-occurrence of environmental context cues, internal cues, and media context cues (Schnauber-Stockmann, Bayer, et al., 2025), indicating that disengagement is an option in a given situation (Logan & Cowan, 1984; Weidacker et al., 2021; Wonneberger et al., 2009).
Hereby, we look at disengagement at the level of mobile media apps, which allows us to bundle and explore all three cue categories that lead individuals to stop their current media use activity. We use the apps TikTok and Instagram as examples because they are particularly known for immersive, visual feeds from which withdrawing seems particularly difficult (Roberts & David, 2023).
The present study makes three key contributions to disengagement research and theorizing. First, it proposes broadly applicable theoretical premises on the scarcely studied but important phenomenon of disengaging from media. Second, it collects, systematizes, and empirically tests disengagement cues from various research traditions. Third, it identifies types of disengagement within mobile media environments. By identifying typically co-occurring cue patterns, we provide a comprehensive foundation for investigating different styles of disengagement. Thus, not merely focusing on the relative frequency of cue occurrence but also accounting for cue combinations, we aim to delineate and explore the variety and complexity of real-world disengagement behavior. Such a descriptive and naturalistic approach not only establishes a fundamental basis for theory development and research on disengagement (Scheel et al., 2021) but also offers vital insights into how media use is stopped in situ.
What Makes Individuals Stop Using Media in a Specific Situation?
Whereas there is a large body of research in psychology and some accounts in communication science about how behavior in general and media use specifically is inhibited before it is started (see e.g., Hofmann et al., 2012), surprisingly little is known about what makes individuals stop an ongoing (media use) behavior. In psychology, the existing approaches either mainly focus on forced stopping in lab experiments (e.g., the stop-signal paradigm, Verbruggen & Logan, 2008) or refer to control over behaviors that run against current goals or are seen as (situationally) problematic (e.g., Hofmann et al., 2012). In communication science, early research on TV channel switching identified factors that make individuals terminate watching one channel or show and tune in to another (e.g., Fahr & Böcking, 2009; Heeter, 1985; Webster & Wakshlag, 1983; Wonneberger et al., 2009). More recent research has focused on the avoidance of specific content in digital environments (i.e., news avoidance or selective avoidance; Andersen et al., 2024) or disconnection from digital devices as preventive strategies to reduce (digital) media use (Klingelhoefer et al., 2024; Matthes et al., 2022; see Nassen et al., 2023 for an overview). However, given the assumptions that (a) individuals are not “lost and helpless” once they have started media use but possess the agency to disengage and (b) many day-to-day (media use) behaviors are mundane and unobjectionable, thus not necessarily problematic, it is important to complement selective avoidance and disconnection research and look at how users disengage from a media use episode.
Only very recently have communication researchers started exploring this process in more detail theoretically and empirically (e.g., Baumgartner & Kühne, 2024; Gilbert et al., 2024; Rixen et al., 2023; Tran et al., 2019). These accounts are focused on specific aspects of disengagement, for instance, highlighting the role of self-control or entertainment (Baumgartner & Kühne, 2024; Gilbert et al., 2024; Reinecke & Meier, 2021) or compulsive media use (Tran et al., 2019). As a pendant to media selection, disengagement can provide a more comprehensive view of why and how any ongoing media use activities are stopped at a given moment as a result of various antecedents at different levels of analysis, such as content, features, apps, or devices (see Meier & Reinecke, 2021). To obtain this comprehensive view, we integrate different lines of research to find meaningful reasons for disengagement. We thereby focus on the app level as a key unit of analysis in mobile communication. Our understanding of disengagement is built on the following premises.
Premise 1: Cues Indicate That Disengagement Is an Option
Previous research touching upon disengagement suggests that cues are indicative of disengagement, meaning that they signal disengagement as a possibility (e.g., Fahr & Böcking, 2009; Heeter, 1985; Logan & Cowan, 1984; Weidacker et al., 2021; Wonneberger et al., 2009). Cues are building blocks of a situation, as perceived by the individual. They may originate from the environmental context, internal state of the individual, or the media context (Schnauber-Stockmann, Bayer, et al., 2025). Cues serve as signals or triggers that potentially influence the interpretation of a situation and thereby subsequent behavior such as disengagement (for related arguments, see e.g., Bayer et al., 2016; Sundar et al., 2007). They can be distinct. For instance, changing location may be seen as a distinct cue for disengagement. Thresholds can similarly serve as disengagement cues (Fisher et al., 2023; Galak & Redden, 2018). Thresholds refer to neuralgic points in continuous processes. For instance, individuals trying to manage their mood via media use may feel gradual mood changes until they reach their desired level. Whereas disengagement cues signal that disengagement is a possible course of action, they can, but do not necessarily have to, be followed by disengagement behavior (consequential vs. inconsequential cues). Thus, individuals may perceive the cues but not act upon them, for instance, due to low levels of self-regulation.
Disengagement is shaped by constellations of concurrent cues rather than single cues. Environmental context, media context, and internal states are intertwined in complex ways that reflect the heterogeneity of media use episodes in general (Schnauber-Stockmann, Bayer, et al., 2025; Schnauber-Stockmann, Scharkow, et al., 2025) and, consequently, disengagement in everyday life. Variations in the co-occurrence of cues may be associated with different disengagement processes. For example, terminating a media use episode due to non-media obligations such as household chores may take different forms, depending on other present cues. A person may disengage quite easily if they already feel satisfied with their media use, or it may be difficult to stop if media use goals have not yet been met, highlighting the importance of the combination of secondary activities and goal attainment.
Premise 2: Disengagement Can Take the Form of Turning Away From a Current Media Use Behavior or Turning Toward a New (Non-)Media Behavior
The rationale behind disengagement can be to stop the (by now) unappealing media use behavior (e.g., because it fulfilled its initial purpose or because of averse content; see “Disengagement Cues”) or to start a more valued new behavior. This premise is mainly based on two lines of research. First, approach vs. avoidance motivation (Elliot, 2006; Krieglmeyer et al., 2013; Strack & Deutsch, 2004) indicates that individuals avoid unpleasant behaviors by terminating them. Thus, if a current media use behavior is perceived as negative, individuals may turn away from the behavior.
However, even if a current media use behavior is not perceived as averse, there may be reasons to terminate it. A second prominent line of research draws from goal switching (e.g., Godara et al., 2020; Kim et al., 2023) and attention allocation (e.g., Awh & Pashler, 2000; Gottlieb & Balan, 2010; Zelinsky & Bisley, 2015) and indicates that individuals organize their attentional focus on “priority maps” (Fisher et al., 2023). Individuals can distribute attention between different facets of a situation at the same time (Awh & Pashler, 2000; Fisher et al., 2023). For instance, they can pay more or less attention to the presence of disengagement cues. Which behavior is prioritized and thus pursued depends on how valued the competing options are (Fisher et al., 2023; Zelinsky & Bisley, 2015). Thus, during a media use episode, such priorities can shift and lead to the termination of the current media use behavior because the individual wants or needs to pursue a different behavior (e.g., because of an important duty that cannot be delayed anymore). At the app level, this means that individuals close the app they are currently using and subsequently start a new media (e.g., using another app; Tandoc et al., 2019) or non-media behavior (e.g., sleeping; Exelmans & Van den Bulck, 2021).
Premise 3: Disengagement Can Vary by Effort
Disengagement can vary depending on how much effort it takes for individuals to “unglue” themselves from mobile media use. We propose that this is shaped by two key dimensions: the degree of automaticity and stickiness. Automaticity suggests that disengagement can range from a habitual, effortless response to cues (e.g., Bayer et al., 2022; Gardner et al., 2016; Neal et al., 2012 for habitual behavior in general) to a more intentional and effortful process, such as when goal conflicts arise (Fisher et al., 2023). Conversely, stickiness refers to the extent to which mobile media maintains attention (Brinberg et al., 2023; Siebers et al., 2024), making disengagement more difficult. Engaging content and a “flow-inducing” mobile environment promote prolonged screen time through the “screenertia” effect—where the longer one remains on the screen, the harder it becomes to turn away (Brinberg et al., 2023; Siebers et al., 2024).
To sum up, disengagement cues—as triggers for habitual processes or as information for more intentional processes—indicate whether stopping a media use behavior is an option in a given situation. In the following sections, we elaborate on such disengagement cues.
Disengagement Cues
To comprehensively identify meaningful disengagement cues, we conducted a literature review that examined a wide range of scattered research traditions. In doing so, we built on the current focus on entertainment in disengagement research and integrated additional cues that are potentially critical in mobile media use (e.g., algorithmic curation, privacy concerns, or ad avoidance). Following Schnauber-Stockmann, Bayer, et al. (2025), we distinguish between the categories of environmental context, internal psychological states, and media context as an organizing framework for disengagement cues.
Environmental Context Cues
We consider competing activities, location, time, social surroundings, and norms as potential environmental context disengagement cues (Schnauber-Stockmann, Bayer, et al., 2025). Competing activities refer to available behavioral options in a given moment. Their availability may signal a need to shift priorities or goals away from media use and in the direction of either more attractive or more important (see below, goal conflict) tasks (Baumgartner & Kühne, 2024; Gilbert et al., 2024; Reinecke & Meier, 2021; Tran et al., 2019).
Changes in location, for instance, getting into or out of public transportation, are typical cues that induce behavior change (see related arguments from habit research on the instigation of a behavior, e.g., Neal et al., 2012). They may also signal new behavioral options or merely make media use impossible (e.g., no Internet connection, interference from walking home and looking at one’s smartphone). Different times of the day may be associated with the perception that a media use behavior is more or less appropriate. For instance, media use late at night may be inappropriate because it interferes with sleep and therefore signals that disengagement is warranted (e.g., Exelmans & Van den Bulck, 2021). Social surroundings and norms are known as important factors related to which media use behaviors are appropriate in a given situation (Legros & Cislaghi, 2020; Tran et al., 2019) and may therefore signal that disengagement is necessary, turning attention to the persons present.
Internal Cues
Psychological states can change during media use, reaching thresholds and thus serve as disengagement cues. Based on media use research (foremost the uses and gratifications tradition, e.g., Katz et al., 1973, and mood management from the selective exposure paradigm, e.g., Knobloch-Westerwick, 2015), individuals may use media for specific purposes, for instance, to reach a goal or satisfy a need (e.g., entertainment or information), or to regulate and adjust affective states. All of these approaches are surprisingly silent on what happens once the goal is reached, the need fulfilled, or the mood managed. It may be assumed that in this case, that is, if the respective threshold is reached, users stop media use (Baumgartner & Kühne, 2024; Browne et al., 2007; W. Wu & Kelly, 2014). However, failure to reach a goal may also lead users to abandon a task—for example, when they no longer learn anything new or become frustrated because they cannot find what they are looking for (W. Wu & Kelly, 2014). Thus, both achievements and failures regarding initial goals, needs, or moods can prompt disengagement.
Furthermore, individuals may terminate a media use session because their mental resources are depleted. This can be due to a lack of capacity to process or fatigue from highly stimulating, varying, often audiovisual content, triggering cognitive arousal (particularly for TikTok; K. Wang & Scherr, 2022). Being a stop signal (van der Linden, 2011), resource depletion and fatigue can be reasons not only to discontinue the overall use (Fu et al., 2020) but also to terminate a (prolonged) use session in a given situation (Baumgartner & Kühne, 2024).
A salient reason for disengagement may be to regain control after experiencing a loss of it. There are two reasons for a loss of control. First, users can experience a self-control failure if they use mobile media longer than they intended to or start using it despite other activities being perceived as more important (e.g., Hofmann et al., 2012; see above). Second, if individuals have experienced multiple small “losses of control” because of events during their current use episode (e.g., expectancy violations; see below), they may experience reactance, which often follows a loss of control (Ratcliff, 2021) and can prompt them to discontinue their current media use activity because a threshold of tolerated frustration was reached.
In addition, individuals can experience negative self-conscious emotions such as guilt, shame, or regret during their current media use episode, which drives them to stop (Gilbert et al., 2024; Reinecke & Meier, 2021; Rixen et al., 2023). Often, such self-conscious emotions result from users appraising a media use activity as dissonant with how they would like to use media ideally. This can be accompanied by (a) the perception that their current (entertaining) media use conflicts with other more important activities or duties, such as getting work done or sleeping (goal conflicts; Baumgartner & Kühne, 2024; Hofmann et al., 2012) or (b) internalized norms stating that time spent with mobile media use is an overall waste of time, not meaningful, or even harmful (Lee & Hancock, 2023; Tran et al., 2019). Thus, individuals appraise a media use activity as inappropriate, specifically if it has reached a certain time threshold (Tran et al., 2019), increasing the likelihood of disengagement (Turel, 2016).
Media Context Cues
Media context cues can be device- and content-related. Device-related disengagement cues are, most importantly, notifications. These notifications originate from other apps (i.e., triggering other media activities; Bayer et al., 2022) but also notifications intentionally implemented by the user to signal that disengagement is warranted, for instance, app-based time limits (e.g., Kang et al., 2024).
Content-related disengagement can be tied back to different research traditions (see initial reasoning by Fahr & Böcking, 2009): (1) states of dissonance (Festinger, 1957) and their reduction (McGrath, 2017), (2) reactance (Ratcliff, 2021), and (3) avoidance coping behavior with media content (see Elliot, 2006). Connecting with our disengagement premises, content patterns—often recommended content that users are exposed to incidentally on a feed—should be appraised constantly during use, and disengagement should become more likely if a threshold of “too much” of unappealing or “too little” of appealing content is reached. If content is moderately disliked or annoying (Hallinan et al., 2023), being repeatedly exposed to it could ultimately prompt disengagement. A single content element (e.g., a video) or other cues within the app could trigger disengagement if they violate users’ expectations and thus evoke negative emotional reactions (Min, 2019; L. Zhang et al., 2023). Given these three patterns, we propose the following content-related disengagement cues: too trivial and homogenous, too problematic and challenging, too intense persuasion attempt(s), not accurate enough, too accurate, and unsuccessful personal curation.
Content can prompt disengagement because it is or has become too trivial and homogenous for users. Trivial refers to hedonic, fun, and lightweight content. Users can feel guilt, shame, or regret over their exposure to such hedonic content because it conflicts with norms or personal standards (Reinecke & Meier, 2021). Too homogenous indicates that (recommended) content seems “recycled,” boring, or redundant to the user because it repeats itself regarding central characteristics or is not stimulating (Rixen et al., 2023; Tran et al., 2019; L. Zhang et al., 2023). From this perspective, disengagement can be explained by frequent skipping of content instead of immersing in it (Tam & Inzlicht, 2024) and the “hedonic decline,” which indicates that even if the content is pleasurable, its enjoyment fades through repetition and psychological adaptation (Baumgartner & Kühne, 2024; Galak & Redden, 2018). Since content genres, messages, or trends may repeatedly appear on the feed during an episode (see Cho et al., 2023; Reeves et al., 2021), such a hedonic decline should become likely after a particular time of mobile media use.
Furthermore, content that is perceived as too challenging or problematic can cue disengagement (Fahr & Böcking, 2009). This refers to content that evokes disgust, is particularly antisocial, strongly challenges moral values, or reminds individuals of their own negative experiences. As a consequence of seeing such content, users may experience distress (Hopwood & Schutte, 2017). Based on research on avoidance coping (Roth & Cohen, 1986; Shin & Kemps, 2020) and dissonance reduction (McGrath, 2017), we assume that exposure to such content on a feed can likely drive individuals to distract themselves from what they just saw by stopping their current media use behavior and continuing with another (non-)media activity.
As promotional content is extremely prevalent in mobile media apps (Beckert et al., 2021), too intense persuasion attempts can be a reason for disengagement. When users are exposed to influencer content or advertisements too often or for extended periods, they may recognize the persuasive intent, which can elicit reactance (Koch & Zerback, 2013). Consequently, they may stop their media use.
If algorithmically recommended content is not accurate (enough anymore), it no longer aligns with the user’s interests, attitudes, or values—indicating that the media platform’s personalization has failed (Eg et al., 2023; Zhao & Wagner, 2022). This can lead to expectancy violations, which may prompt individuals to disengage. Conversely, if the (recommended) content is too accurate, users will notice the presence of algorithmic recommender systems or will feel too intensely “seen” by the recommendations. This is due to a “privacy breach,” which indicates an uncomfortable feeling of being spied on, based on previous online or offline behavior (Oeldorf-Hirsch et al., 2025; see also D. Zhang et al., 2025). However, this effect is also influenced by high self-relevance (Lee et al., 2022) and a perceived loss of distance (Fahr & Böcking, 2009) of the recommended content. If such content is perceived as negatively intimate or overtly real, it may make users feel uncomfortable, vulnerable, and “caught in the act.”
Finally, unsuccessful personal curation refers to users’ ambition to shape their user experience and recommendations in the presence of other responsive curators, such as platform algorithms (Thorson & Wells, 2016). Increasingly, users attempt to refine algorithmic recommendations by signaling their preferences to the recommender system through their likes, watch duration, or other engagement behavior toward the content they wish to be exposed to (e.g., Lee et al., 2022). If users notice that, despite their efforts, the content does not change, this will likely lead to frustration and can prompt them to stop their current media use activity to let the recommender system “recalibrate” (Chen, 2023, p. 12).
The Present Study
In this study, we aim to explore our stated premises within typical types of disengagement from mobile apps in daily life. Premise 1 states that there are cues for disengagement, which can (consequential) but may not always be (inconsequential) succeeded by disengagement behavior and which may co-occur within situations. As outlined, there is a magnitude of possible environmental, internal, and media context cues that could trigger disengagement. First and foremost, it is essential to investigate the relative importance and relevance of these disengagement cues for actual disengagement behavior. Therefore, we ask:
RQ1: Which disengagement cues (a) occur most frequently and (b) are most consequential for disengagement behavior?
Second, Premise 1 reflects that media use in everyday life is shaped by a complex interplay of environmental, media-related, and internal factors (Schnauber-Stockmann, Bayer et al., 2025). Disengagement cues do not occur in isolation but in combination (c.f., Rixen et al., 2023). This means that consequential cues (i.e., cues that lead to disengagement in a given situation) may coincide and form typical co-occurring patterns. Therefore, we aim to go beyond the relative frequency of single cues and consider their co-occurrence to delineate and explore types of disengagement. We ask:
RQ2: Which distinct types of disengagement can be identified in everyday life?
These types of disengagement are of focal interest because they reflect the complexity of disengagement situations in realistic settings and are thus key to understanding how individuals stop using media in their everyday lives. Therefore, we aim to explore their impact on disengagement behavior in more detail.
Premise 2 states that disengagement can be motivated by turning away from the focal media use behavior or by turning to an alternative activity. Building on this premise, we aim to explore whether different types of disengagement map onto this distinction. Therefore, our third research question asks:
RQ3: Which types of disengagement are more likely motivated by turning away from the focal media use behavior or by turning to an alternative activity?
Finally, Premise 3 states that disengagement behavior varies according to effort. This is indicated, for one, by the degree of automaticity of the disengagement process. For the other, longer use episodes are indicative of higher levels of stickiness associated with more effortful disengagement. We assume that different types of disengagement may vary in terms of effort, given that their specific cue combinations may impact how disengagement behavior is performed. Therefore, we ask:
RQ4: Which types of disengagement differ in (a) their degree of automaticity and (b) the duration of the preceding use episode?
Method
Materials, data, code, and the appendix are available online in the Open Science Framework (OSF): https://osf.io/v3yqj/.
Procedure and Participants
In an event-based mobile experience sampling study (MESM), German Instagram or TikTok users answered questions about use sessions directly after their Instagram or TikTok use for eight days. The app movisensXS was used for data collection. MovisensXS tracked (1) when the respective app (either Instagram or TikTok) was opened, (2) when it was closed, and (3) when the display was turned on and off (since turning off the display can be a version of closing the app). Based on the tracking data, the MESM protocols were automatically triggered by closing Instagram or TikTok (either by closing the app or by turning off the display). To reduce the overall burden of participation, the number of sampled episodes was limited to a maximum of six per day (a maximum of two protocols between 12 am and 8 am, 8 am and 4 pm, and 4 pm and 12 am to ensure that different parts of the day were covered). The tracking data were also used to calculate the number of use episodes (defined by the time between opening the app and closing it or turning off the display) and their duration.
Participants were recruited in autumn 2024 via the online access panel SoSci Panel. They had to be 18 years or older, Android users (as movisensXS only runs on Android devices), and use either Instagram or TikTok at least two to three times a week. Participants were assigned to answer all pre-survey questions as well as MESM protocols for either Instagram or TikTok. Due to the lower prevalence of TikTok use in Germany (We Are Social et al., 2024), all participants who used TikTok at least two to three times a week were assigned to the TikTok condition. Before giving informed consent and starting the MESM phase, participants were informed about the MESM study and received an FAQ document as well as contact details in case of technical or other issues. After the eight-day MESM phase, participants were invited to a post-survey where they could sign up for an incentive of 20 euros if they had answered at least 60% of all MESM protocols that were sent to them.
In total, 425 participants met the inclusion criteria and completed the pre-survey, and 265 gave informed consent to take part in the MESM phase. Of those, 143 enrolled in movisensXS, and 118 completed at least one MESM protocol (55% attrition based on informed consent). Thus, the final sample for the following analyses consisted of 118 participants—89 in the Instagram condition and 29 in the TikTok condition. Sixty-four percent identified as female, 33% as male, and 2% as diverse. The sample was highly educated (76% high school graduates), and the mean age was 35.8 years (SD = 11.4).
Participants used Instagram or TikTok an average of 45.6 times (SD = 46.6, Min = 3, Max = 227). Of all the use episodes tracked, 53% per participant were followed by a MESM protocol on average, resulting in a total of 1,893 completed MESM protocols (1,554 for Instagram and 339 for TikTok; protocols per participant: M = 16.0, SD = 10.5, Min = 1, Max = 41). The average compliance rate was 66%. Participants took 1.11 min (SD = 2.08) to complete the MESM protocol.
Measures
Pre-Survey
The pre-survey was part of a larger project on mobile media use. For the present study, we used only sociodemographic information on gender, age, and education level for sample description from the pre-survey.
MESM Protocol
The MESM protocol focused primarily on disengagement cues. In a first step, participants indicated which of the following disengagement cues applied to the past use episode (“Which of the following experiences have occurred during your [Instagram/TikTok] use? Please select all that apply.”). The response options were as follows: (a) environmental context: “other activities seemed more important to me,” “I changed location,” “timing was bad (very early, very late),” “I met someone/started a conversation,” (b) internal: “using the app brought me closer to my initial goal,” “using the app did not fulfill my initial goal,” “using the app was exhausting,” “I felt like I lost control,” “I felt guilty/ashamed/had regrets,” and (c) media context: “I received a push notification from another app,” “I received a notification that my time limit was up,” “the content no longer suited me,” “I felt too intensely seen/caught by the content/app,” “I felt that I should be persuaded by products/things,” “the content was very challenging or problematic for me,” “the content was too similar/homogeneous or not interesting/entertaining (anymore),” “I failed to improve the content recommended to me”).
In a second step, participants selected among the cues that had applied to the current situation whether they were a reason to end the use episode (“Below, all experiences that have occurred are listed. Please indicate whether each experience was a reason to close [Instagram/TikTok]. You can select one, several, or all experiences.”). In both steps, participants could choose as many cues as they felt applied. Additionally, they could indicate that they had closed the app accidentally or had only been interrupted while using it.
Furthermore, the MESM protocol contained a question on the degree of automaticity of disengagement (“I closed the app automatically,” 5-point agreement scale, based on Gardner et al., 2012) (M = 3.77, SD = 1.56) as well as additional questions on the use session nit relevant for the current analyses.
Data Analysis
All analyses were conducted in R (R Core Team, 2021). To answer RQ1, we aggregated for each participant separately how often a cue had applied to a use episode and how often it was the reason for disengagement (see Table A3 in the OSF for the sample distribution of the cues). We used this information to calculate the likelihood of closing the app when the respective cue applied. To answer RQ2 and RQ3, we conducted a multilevel latent class analysis (MLCA; Lukočienė et al., 2010; Vermunt, 2008) using the package multilevLCA (version 1.5.1, Lyrvall et al., 2023). The analysis included variables indicating whether a cue was perceived as consequential—that is, whether it served as a reason for disengagement. To ensure that the variables had at least some variance while not discarding rare cues, we excluded cues that were selected in less than 1% of all protocols (“The content was very challenging or problematic for me” and “I failed to improve the content recommended to me”).
MLCA was used to identify unobserved (latent) classes while accounting for the hierarchical data structure (i.e., situations nested in participants). multilevLCA thereby allows for random variation in class membership probabilities across persons. This accounts for individual-level heterogeneity by modeling probabilistic class assignment at both levels rather than assuming fixed-class structures. To this end, class selection was done in three steps. First, single-level models were computed to select the optimal number of situation-level classes. Second, multilevel models were fitted with the number of situation-level classes held fixed to identify the optimal number of person-level classes. Finally, multilevel models with the number of person-level classes held fixed were run, re-evaluating the optimal number of situation-level classes (Lukočienė et al., 2010; Lyrvall et al., 2023). As we focus on the situation level, the person-level classes are not interpreted in this paper. We merely used MLCA to account for the multilevel structure of our data (see Table A4 in the OSF for the person-level classes).
Only participants with more than one MESM protocol were included, as MLCA requires multiple measurements per person, resulting in a sample size of 115 participants and an average of 16.54 (SD = 10.3) protocols per participant. Based on simulation studies, our sample size is thus sufficient to detect classes at the situational level (Park & Yu, 2018).
To answer RQ4, we assigned the modal class to all episodes. 1 We ran multilevel models (using the R-package lme4, version 1.1-35.5, Bates et al., 2015) with disengagement automaticity and use duration as dependent and the classes as independent variables to test for class differences. We used estimated marginal means for post hoc pairwise comparisons (Bonferroni correction) to determine significant differences between disengagement types (using the R-package emmeans, version 1.10.4, Lenth, 2017).
Results
During the field period, we tracked an average of 45.6 Instagram or TikTok use episodes per participant (SD = 46.6, Min = 3, Max = 227). On average, a use episode lasted 5.65 min (SD = 4.14, Min = .23, Max = 22.1). Based on the sampled use episodes, the most frequently encountered disengagement cues were competing activities (69.5%), goal achievement (43.6%), and push notifications (19.7%, RQ1a; see Table 1). On average, 2.11 cues were encountered per episode (SD = 1.44, Min = 0, Max = 9) and 1.73 were consequential, leading to the termination of the use episode (SD = 1.19, Min = 0, Max = 8). There are substantial differences between the cues in terms of whether they lead to disengagement (RQ1b; see Table 1). Whereas, for instance, participants only reported in 10.1% of their use episodes that they were interrupted by other people with whom they started a conversation, this disengagement cue was consequential in almost all such situations (93.1%). The likelihood of disengagement upon encountering a cue was high (≥75%) for all environmental cues as well as for push notifications (media context cue). It was low (≤25%) for the media context cue of unsuccessful personal curation (however, this cue was seldom encountered, and only five participants chose it at least once).
Cues Perceived, Consequential Cues, and Share of Consequential/Perceived Cues Per Person.
Note. Means of person-level percentages. See Table A3 in the OSF for the sample distribution.
The MLCA identified five types of disengagement (RQ2 and RQ3), which are displayed in Table 2. The first type can be described as non-self-determined disengagement. In this disengagement type, individuals likely experience negative self-related states, particularly loss of control, and feelings of guilt, shame, and regret, but also a depletion of mental resources. Adding to the lack of self-determination, they likely perceive their use as badly timed. As in most other disengagement types, their mobile app use competed with other activities. This rather problematic type was the most likely (modal class) in 8.6% of all use episodes. It is strongly determined by turning away from current media use behavior, which is perceived as averse.
Types of Disengagement (Multilevel Latent Class Analysis).
Note. NPerson = 115, NSituation = 1890. AIC = 15161.83, BIClow = 15738.44; BIChigh = 15447.30; R2entrlow = .77; R2entrhigh = .84.
The second type of disengagement represents—opposed to the first type—self-determined disengagement. In these situations, individuals are likely to have reached the goals they had when starting their app use, representing a—from an idiosyncratic point of view—“successful” media use episode. At the same time, priorities shifted, and other competing activities became more important. Self-determined disengagement was the most likely type in 26.2% of all use episodes. It is characterized by both turning away from app use and turning to another activity. Turning away, in this case, was not caused by aversion as in the first type but by needs being satisfied—meaning the media use has fulfilled and thereby lost its initial purpose. This may have led to a shift in priorities, turning attention and resources to another activity.
The third type of disengagement shares with the second the similarity that individuals turned to a competing activity. However, unlike the second type, disengagement was not accompanied by a feeling of goal achievement. Thus, it is characterized by turning to a new activity without feelings of aversion (type 1) or satisfaction (type 2), indicating a priority shift that is rather independent of the focal media use behavior, for instance, because the other activity became more pressing. Disengagement due to priority shifts was the most likely type in 38.9% of all use episodes.
The fourth type of disengagement represents temporary disengagement. It is very likely that individuals closed the app only by accident or took a short break from its use. This was likely caused by push notifications of other apps or by starting a conversation with another person. Thus, unlike the other types, temporary disengagement does not represent final disengagement but covers instances of short breaks within longer media use sessions. It was the most likely type in 15.2% of all use episodes. This disengagement type was mainly driven by (at least shortly) turning to another activity, be it talking to people present or using another app.
The fifth type of disengagement does not include any specific reasons for disengagement. In tendency, it is characterized by competing activities, location changes, and push notifications but lacks a clear pattern. Thus, it presents a form of “mindless” disengagement. It was the most likely type in 11.0% of all cases. Due to its lack of characteristics, it is unclear whether disengagement was driven by turning away from current media use behavior or by turning to another activity.
Finally, the degree of automaticity of closing the app (RQ4a) and the use duration (RQ4b) varied between disengagement types (Table 3). Regarding the automaticity of closing, temporary disengagement differed significantly from all other types, showing the lowest level of automaticity in disengagement. Additionally, self-determined disengagement was characterized by a significantly higher degree of automaticity than non-self-determined disengagement. Regarding use duration, non-self-determined disengagement differs from all other disengagement types, showing the longest use duration. Furthermore, temporary disengagement differs significantly from all but the “mindless” type, with the shortest use episodes.
Automaticity of Disengagement and Use Duration—Estimated Marginal Mean Comparisons Between Types of Disengagement.
Note. Estimated marginal means with different superscripts differ significantly, with p < .05 (multilevel models with degree of automaticity/use duration as dependent variables and types of disengagement as independent variables, post-hoc-test for group differences: Bonferroni).
Discussion
Summary
Complementing media avoidance and disconnection research and building upon initial empirical work on media disengagement, this study used an “in situ” lens to understand disengaging from mobile media use, such as Instagram and TikTok, as a mundane everyday behavior. Setting a cornerstone for investigating (mobile) media disengagement, we derived theoretical premises from the literature. First, we introduced the idea of disengagement cues, which can be (1) consequential or inconsequential, (2) co-occur, and (3) become salient either distinctly or gradually in the form of thresholds. Second, we stated that disengagement incorporates turning away from the focal media use activity and turning toward a new activity. The third premise states that disengagement varies by effort. Based on these premises, we collected, systematized, and empirically tested 17 disengagement cues that originated from the environmental context, the internal states of the user, and the media context.
The exploratory results of our event-based MESM study underline the relevance of disengagement cue categories. The most prevalent cues were distributed across all three cue categories: competing activities (environmental context), goal achievement (internal), and receiving push notifications (media context) occurred most frequently (RQ1a). In addition, our study revealed that environmental cues are usually consequential (RQ1b). For instance, if users felt that other activities were more important, it was a bad time, or they wanted to devote their attention to the people around them, this often led to the termination of use. Additionally, reaching an app limit—a media context cue—frequently stopped users from continuing their media use. Notably, many content-related cues from the media context category, such as too much persuasion from ads or influencers, were not indicative of users terminating their use.
Exploring mobile media disengagement in situ as situations in which multiple consequential cues co-occur, we identified five types of disengagement (RQ2): (1) non-self-determined disengagement, (2) self-determined disengagement, (3) disengagement through priority shifts, (4) temporary disengagement, and (5) “mindless” disengagement. In response to RQ3, which addressed the attentional focus of disengagement types—turning away from the focal media activity vs. toward a new activity—we found that most types reflected a shift toward a new activity without the media activity being perceived as aversive (types 2, 3, and 4). Only type 1 involved a clear turn away from the focal media use, whereas type 2 included elements of both directions. Finally, we found that disengagement behavior varied in how effortful it was (RQ4).
Implications for Mobile Media Disengagement
Disengagement Can, But Does Not Necessarily Have to, be “Problematic”
Only one type of disengagement (type 1) was characterized by goal conflicts and adverse emotional reactions—such as guilt or regret about using Instagram or TikTok—as well as by users’ desire to turn away from media because they felt exhausted and experienced a loss of control during use. Strikingly, this type of disengagement has been a dominating scientific and societal narrative surrounding everyday mobile media use, often focusing on such media use behavior that originates from self-control failures and leads to deteriorations in individuals’ mental health and well-being (as discussed by Vanden Abeele, 2021). Our exploratory analysis suggests that situations corresponding to disengagement type 1 are part of mobile media users’ everyday lives. However, they represent the predominant pattern in only about 8% of all situations.
Type 1 is in stark contrast to type 2. The second disengagement type represents “non-problematic” media use. Users find what they have been looking for and consequently close Instagram or TikTok, turning to other, higher-prioritized activities. Such self-determined mobile media use relates to early ideas from uses and gratifications research, conceptualizing autonomous users utilizing media to fulfill their needs (Katz et al., 1973), and to related work on disengagement, which similarly finds that individuals often experience “closure” when stopping rewarding entertainment media use (Alexandrovsky et al., 2024). The termination of daily mobile media use being very often self-determined suggests that using TikTok and Instagram is often an indulging treat, a “guilty pleasure” without guilt, or an assistant who helps gather instrumental or social information that individuals are capable of putting away if no longer needed.
The Availability of Other Mediated or Non-Mediated Activities Seems Crucial for Disengagement Behavior
Another overall finding is that competing activities from the external environment and push notifications from the media context are prevalent cues for disengagement behavior. Thus, the mere salience of alternative activities can serve as a powerful driver of media disengagement. This is underlined with type 3, “disengagement through priority shifts” being the most frequent type of disengagement.
For most disengagement types—especially types 1, 2, and 3—the likelihood that individuals terminate app use due to competing activities was consistently high. Overall, environmental context cues frequently played a consequential role in disengagement behavior. This finding broadly aligns with related work (Rixen et al., 2023). Interestingly, in type 4, competing activities are not a driver; instead, a push notification from another app is—because it probably triggers individuals’ connection habits (Bayer et al., 2016). However, there is an important distinction between the two: The app’s “pull factors” (such as enjoyment during use; Gilbert et al., 2024) are likely not pervasive enough to ignore the availability of competing non-mediated activities because, if triggered, they often lead to an ultimate termination of a use session. In comparison, media availability seems less persistent, as it may rather lead to quickly checking the notification’s origin but returning to the respective app afterwards. Looking at the bigger picture, both imply that current app use often “loses the battle of attention” if other (non-)mediated activities signal availability and are persistent—even if it is only briefly.
Content-Related Cues Are Unlikely Drivers of Mobile Media Disengagement
Throughout all types of disengagement, content-related cues originating from the media context were rarely perceived and thus played a minimal role in disengagement. At least in the context of Instagram and TikTok use, these cues appear to have little evocative potential to be perceived as expectancy violations or sources of friction (Kang et al., 2024; Min, 2019; Natarajan, 2024) and to drive disengagement from apps. There are two explanations for this finding. First, the content of the investigated mobile apps may be too fleeting to stick with individuals. Other than disengagement in long-format audiovisual media (Baumgartner & Kühne, 2024; Gilbert et al., 2024), mobile media apps entail very short content units (e.g., a short video, a story) of which individuals consume plenty per use session. Thus, individuals may easily forget what they were exposed to during the session, and even if, for instance, particular content evoked individuals’ reactance, such content remained inconsequential.
Second, the mobile media apps we investigated—Instagram and TikTok—are both high in stickiness and entertainment value and are optimized to keep users engaged on the app while preventing them from disengaging (Alter, 2017; A. X. Wu et al., 2021). Our results confirm that such sticky apps’ content is successful in not being (too) disruptive. Instead, they seem to effectively keep individuals’ flow (c.f., A. X. Wu et al., 2021; Zhao & Wagner, 2022). Perhaps such content-based disengagement cues may become more salient on other apps that are morally and emotionally more challenging, political, or news driven, such as X (previously Twitter), or that could more quickly feel homogenous because they are more specialized in certain types of content, like LinkedIn, or text-based modalities, like Reddit. It is thus likely that content, in general, does not drive mobile media disengagement but that there potentially are differences between media options (see also Rixen et al., 2023).
In Some Situations, Mobile Media Disengagement Is More Effortful Than in Others
Aligning with our premise and with suggestions from previous research (Gilbert et al., 2024), the types of disengagement varied by effort. Significant differences occurred between self-determined and non-self-determined disengagement, with self-determined disengagement being characterized by higher automaticity and a shorter use duration. Individuals having had “enough” and consequently terminating their use in a more automatic fashion may be tied to consumption habits (e.g., a tendency to use media for a particular length, as discussed by Baumgartner & Kühne, 2024). Conversely, non-self-determined disengagement, being a rather conscious and deliberate behavior, can be explained with a dual-systems lens from psychology, stating that encountering goal conflicts through competing activities can prompt more elaborate processing and decision-making (see Hofmann et al., 2012, for instance). This type exhibited a longer use duration compared to self-determined disengagement and all other types. Overall, this suggests that individuals need to exert considerable effort to disengage from the ‘sticky’ app, resulting in longer session durations—for example, struggling to stop usage right before going to sleep.
Additionally, temporary disengagement showed the lowest automaticity levels compared to the other types and the shortest use duration (except for the mindless disengagement type). In this disengagement type, individuals are interrupted by other tasks, other people, or notifications demanding their attention while using a mobile app to which they return after a short while. It also involves closing apps by accident. This resonates with very fragmented app use patterns and frequent task switching between apps after short visits. Such different forms of friction from both environmental and media contexts can pull individuals, often unexpectedly, out of their immersive and automatic flow state (Kang et al., 2024; Natarajan, 2024; Rixen et al., 2023). Thus, after being “pulled out” of a flow state, disengagement not being automatic in this type seems intuitive. However, it should be noted that long use sessions are not always associated with effortful disengagement nor shorter sessions with low effort. Nonetheless, our results support this general tendency.
Taken together, our typology illustrates the diverse nature of disengagement situations in everyday life. Since the content of sticky entertaining mobile media apps does not per se afford disengagement, individuals could use them endlessly were it not for the obligations and interruptions of daily life. Our results show that mobile media use is in constant competition with other demands or opportunities. Competing activities shape almost all disengagement types. This suggests that activity conflict may be one of the most pervasive companions to everyday technology use. Overall, the unique contribution of the disengagement types is that they all outline different ways of resolving conflicts between a current media activity and other activities. Type 1 (non-self-determined disengagement) represents a “worst of all worlds” situation in which the conflict is not satisfactorily resolved: Disengaging is effortful, accompanied by negative emotions, and goals are not achieved. Conversely, type 2 (self-determined disengagement) does not require conflict resolution because there is no (perceived) competition for goal attainment. Type 3 (priority shift disengagement) provides an effortless resolution to the conflict because priorities and attention can be easily reorganized and individuals choose to disengage from the current media activity. In these situations, mobile media use may have served primarily as a distraction rather than being driven by a specific goal, making continued use less important. Type 4 (temporary disengagement) represents a compromise in the conflict: individuals switch tasks to briefly engage in another activity but soon return to their original task, thus not fully deciding against it.
Building upon this narrative of activity conflict management, further theorizing of disengagement is a crucial next step for future research. After discussing the limitations of our study, we will outline recommendations for building theories on disengagement.
Limitations and Future Perspectives
This study has three major limitations. First, the typical methodological limitations of MESM research, such as reactivity, also apply to our study. Participants may have used Instagram or TikTok less during the field period to avoid responding to the MESM protocols. Alternatively, participants were likely reactive by spending more time on the app or ignoring push notifications from other apps more often to avoid triggering a protocol. This is a limitation we tried to address by limiting surveys to a maximum of six per day, but we still could not fully control for it. However, such limitations are balanced against the relative strengths of MESM studies, such as high external validity.
Second, the types of disengagement suggest that there is no single process of disengagement but rather patterns and likely a complex interplay of different mechanisms and situational conditions that ultimately shape disengagement. Gilbert et al. (2024) proposed the idea of pull mechanisms (e.g., entertainment) versus push mechanisms (e.g., guilt), suggesting that there may be cues that push disengagement and cues that pull individuals toward continued use. In this paper, we have presented only cues that push disengagement. Nevertheless, we encourage future research to supplement our cue collection with additional pull cues to unravel the dynamics between staying in or leaving a mobile media session.
Third, the apps we studied—TikTok and Instagram—were quite similar in terms of structure (Webster, 2009): Both are mobility-focused apps with similar immersive, algorithmically curated feeds and content, particularly known for being sticky, for which we found only minor differences (see additional analyses in the OSF). This may limit the generalizability of our findings. Future research should therefore examine whether our results extend to other apps. Additionally, different application types have different media context characteristics, which are likely associated with different heuristics and habits of use (Meier & Reinecke, 2021). For instance, immobile settings, such as being at home, may allow content effects to unfold, whereas “on-the-go” contexts are more influenced by environmental (and internal cues), which can shape how content is evaluated or how disruptive within-media events are. Thus, future research on disengagement should consider different structural components of media, such as immersiveness and mobility.
Recommendations for Further Theorizing Disengagement
With this exploratory study, we extend the field of media disengagement from long-format entertainment media to mobile apps. This additional context paves the way for causal models that focus on the different processes of media use disengagement. Based on our finding that worrisome forms of disengagement are not the most prevalent type and that “non-problematic” solutions to activity conflict management subsist, we suggest investigating disengagement from a holistic perspective. That is, future research may further explore disengagement as encompassing—but not being centered solely on— the experience of negative states. The overall pattern of activity conflict management in our disengagement types suggests several avenues for connecting disengagement to existing theoretical approaches, allowing a balanced view.
Types 1 and 2 represent two cases most relevant to scholars studying digital technology use, entertainment experiences, and well-being, for which the AMUSE model (Reinecke & Meier, 2021) and the concept of hedonic decline (Galak & Redden, 2018) provide a sound theoretical foundation that has already been tested in the context of long-format streaming (Baumgartner & Kühne, 2024; Gilbert et al., 2024). Non-self-determined disengagement may be the most problematic type and warrants attention due to individuals’ mental health. Research could expand on this initial work and validate the proposed push vs. pull mechanisms in the context of mobile media. Type 2 calls for extending our understanding of users’ feelings of satiety in a paradoxically “insatiable” digital world that allows for endless use. Satiety, or what Alexandrovsky et al. termed “closure” (Alexandrovsky et al., 2024), can inform ways of healthy use of digital technology, where goals are met and gratifications obtained. Types 3 and 4 may be embedded in research on media and attention that aims to explain how priorities between different tasks are organized, for instance, when users engage in task switching and multitasking to juggle the demands of multiple pursuits (e.g., Fisher et al., 2023; Z. Wang & Tchernev, 2012; Wonneberger et al., 2009) versus when they ultimately disengage and leave.
Conclusion
So, should we worry about disengagement? Studying mobile media disengagement in a real-life setting indicates that effortful disengagement is likely not a pervasive phenomenon but should warrant our attention for specific situations. Hence, even if mobile apps can be used nearly anywhere and anytime and there are hardly any built-in boundaries within apps stopping individuals from using them endlessly, there are often natural or perceived boundaries to media use originating from situational contexts in most situations. Continuing to theorize and explore them will contribute to our understanding of how individuals navigate permanent connectedness in today’s digitalized world.
Footnotes
Acknowledgements
We thank Michael Scharkow for his support during data analysis and Antonia Glaser and Charlotte Tryba for their assistance during data collection. We thank Alicia Gilbert for her expert review on an earlier version of the manuscript.
Author Contributions
Both authors contributed equally to the study and manuscript and thus share the first authorship.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Informed Consent Statement
Participants were thoroughly informed about the study’s procedure, the pseudonymized storage and publication of their data for scientific purposes, and their rights as a participant. They gave written informed consent (i.e., clicked on a button indicating they have carefully read the information provided to them and agree to participate). Informed consent materials are available here: 