The dominant explanation for the attenuation of DBCI effects centers on engagement: users stop using the intervention, and behavior change dissipates accordingly [18]. Although engagement is undeniably important, this explanation treats the problem as primarily technical, involving the design of more compelling interfaces and smart notification systems. A recent international consensus study identified a lack of agreed-upon metrics for engagement and evidence on how better engagement improves outcomes as key unresolved challenges [13]. Therefore, we propose that deeper structural issues should be addressed. Even when users remain engaged, the architecture of many DBCIs can systematically undermine the 3 experiential processes critical for sustained behavior change. Importantly, DBCIs are heterogeneous, and not all of them displace user interpretation in the same way; some already incorporate reflective, autonomy-supportive, or meaning-based components. The concern developed here is therefore directed not at automation as such, but at what we term substitutive automation that replaces the user’s own interpretation, as distinct from supportive automation that prompts reflection, offers choices, or helps users recognize patterns [19]. The 3 mechanisms below describe a structural tendency in prevailing designs rather than a uniform property of all DBCIs.

The first mechanism arises from algorithmic systems in contemporary DBCIs that detect behavioral deviations in real time and respond with immediate corrective guidance [16,20]. When a user’s dietary intake deviates from a target, the system recalculates the meal plan, and when step counts fall short, an adaptive prompt suggests a brief walk. This automated responsiveness is efficient but carries an unexamined cost: it removes the temporal space in which users might reflect on why a lapse occurred and situates it within their broader trajectory of change.

Behavior change, particularly in chronic illness self-management, is not experienced as a series of isolated behavioral events but as an unfolding process that includes hesitation, interruption, and renewed effort [21,22]. A recent narrative review of qualitative studies on the self-management of chronic illness highlights how individuals construct meaning by caring for themselves over time, a process that requires narrative continuity [23]. When algorithmic systems process lapses instantaneously, treating them as errors to be corrected rather than experiences to be understood, users lose the opportunity to frame these events as part of an ongoing process. The lapse becomes something the system handles and not something the person lives through. Thus, the capacity to situate individual setbacks within a continuous, personally meaningful trajectory (process framing) tends to be eroded by the same efficiency that makes these systems attractive. For instance, a person attempting to establish a daily walking routine may skip several days due to work stress or family obligations. If the system immediately responds with a prompt such as “You’re behind your goal, take a short walk now,” the behavior may resume without the user ever reflecting on why the interruption occurred or how similar situations might be handled in the future.

The second mechanism operates through notification-based behavioral prompts. Push notifications are among the most widely used engagement strategies in DBCIs [18,24], and their effectiveness in triggering immediate behavioral responses is well documented [25]. However, when a notification redirects a user from a lapse toward the “correct” behavior, it performs an override rather than a repair. The user moves past the lapse without interpreting it.

Narrative repair, defined as the reinterpretation of a lapse within one’s personal narrative such that it becomes a source of learning rather than a fixed mark of failure, requires a moment of meaning-making that notification-driven redirection tends to foreclose [26,27]. In the relapse prevention literature, the capacity to reinterpret a lapse as a correctable event rather than as evidence of personal inadequacy is recognized as a critical determinant of whether a lapse escalates into full relapse [28,29]. However, most DBCI architectures bypass this interpretation process. For example, someone trying to reduce alcohol consumption may interpret an episode of heavy drinking as either proof of personal failure or a signal of a stressful context that requires new coping strategies. Narrative repair refers to the latter process of reinterpretation. In many DBCIs, however, this interpretive moment is short-circuited: lapses are not narratively repaired but are algorithmically overwritten. Consequently, the experiential residue of lapses, including self-blame, confusion, and demoralization, may accumulate beneath the surface of apparently successful behavioral redirection.

The third mechanism concerns how success is experienced and to whom or what it is attributed. Gamification elements, such as streaks, badges, leaderboards, and progress visualizations, are widely incorporated into DBCIs to sustain motivation [17,30]. Gamification is not inherently problematic; well-designed reward systems can support competence, autonomy, and meaning [31]. Our concern is narrower, focusing on reward structures that supply the interpretation of success rather than supporting the user’s own. These features not only effectively celebrate behavioral achievements but also shape the attributional structure of success in ways that may undermine the user’s sense of authorship.

Consider the experiential sequence: when a user reaches a step count target, the app announces the achievement, frames it within its visual language of progress bars and congratulatory messages, and delivers the reward through its incentive system. The interpretive framework through which success is understood (what it means, why it matters, and how it connects to the user’s life) is supplied by the system rather than constructed by the user. Research on personal informatics systems has shown that the manner in which data are collected, visualized, and reflected back to users fundamentally shapes their self-understanding [32,33]. Over repeated cycles, this externalization of the interpretive framework may prevent behavioral achievements from being integrated into the user’s self-understanding. The achievement remains legible within the app’s logic but not within the user’s own narrative of who they are becoming [34].

This concern resonates with long-standing critiques of nudge-based interventions, which have been noted to risk relocating agency from the individual to the choice architect [35,36]. This does not suggest that external feedback is inherently detrimental; it becomes problematic when it substitutes for, rather than supports, a user’s own interpretation of success. In the context of DBCIs, the issue is not that users are coerced, but that the felt sense of success, the experience of having personally achieved something, is displaced onto the system. Users may meet their behavioral targets while experiencing achievement as something that happened to them rather than something they authored. This state, behavioral success without experiential ownership, constitutes what might be called brittle success: sustainable only as long as external scaffolding remains in place [8]. When the app is deleted, the notification stream ceases, or the device is set aside, the behavior that was never owned may be readily abandoned.

Together, the 3 mechanisms (the erosion of process framing, the foreclosure of narrative repair, and the hollowing of authorship) describe not a failure of technology, but a structural pattern in which the experiential conditions for lasting behavior change are systematically displaced by the same features designed to promote it.

As mentioned previously, narrative ownership refers to the extent to which individuals experience behavior change as an ongoing, interpretable, and self-authored process. More precisely, narrative ownership is defined as a state in which the experience of behavior change, including lapses and interruptions, is framed as part of an ongoing process (process framing), remains repairable through reinterpretation within one’s personal narrative (narrative repair), and is held with a sense of authorship as one’s own story (authorship).

The concept of narrative ownership draws upon several strands of narrative research. Narrative identity research explains how individuals understand themselves through the stories they construct about their lives [37], whereas work on narrative coherence highlights the importance of temporal continuity and interpretability in personal narratives [38]. In the health context, the illness narrative tradition illustrates how experiences of disruption and recovery are interpreted and reconstructed over time [26,27].

Narrative ownership, in contrast, is narrower in scope and more specific in application; it targets how a behavior change trajectory, including its lapses, is experienced and interpreted by the individual and asks whether that trajectory is held as the individual’s own ongoing process. It is not a theory of identity or a description of illness experience, but an evaluative concept designed to assess whether the conditions for sustainable behavior change are present in a given intervention context. Narrative ownership comprises 3 components.

Process framing is the capacity to situate individual lapses and setbacks within a continuous trajectory that includes the surrounding context that triggers them and what is learned from them. When process framing is operative, a lapse is comprehensible not as an isolated failure, but as a moment within an unfolding experience. Narrative repair is the reinterpretation of lapse episodes within one’s personal narrative such that they are reconnected to renewed effort rather than fixed as irreversible failures. Authorship is the attribution of change, not as an externally imposed result, but as one’s own story, one that can be told in one’s own words.

These 3 components are conceptually distinct but functionally interdependent. Process framing provides the temporal structure within which narrative repair can occur. Narrative repair, in turn, sustains a sense of authorship that makes continued engagement meaningful rather than merely compliant.

Three points clarify the status of the concept. First, narrative ownership is a user-level psychological construct: a state, realized through the aforementioned 3 processes, rather than an intervention feature. We therefore distinguish narrative ownership as a property of the user’s experience from narrative ownership–supportive design as an intervention-level set of features intended to cultivate that state; a DBCI does not contain narrative ownership but can create conditions that make it more or less likely. Second, the construct is design-sensitive: unlike a stable trait, it can be strengthened or eroded by the architecture of the intervention, which is precisely what makes it useful as an evaluative lens for design. Finally, authorship need not be strictly individual. Although the language of personal narrative may suggest an individualistic frame, in relational and collectivist contexts the relevant narrative may be familial, communal, or shared [39], and authorship may be exercised jointly with family members, care partners, or community networks. Recognizing these relational forms at the level of conceptual development, rather than treating them only as a caveat, extends the construct rather than weakening it.

Several established concepts in the DBCI literature address related but distinct phenomena.

Engagement has become the central construct in DBCI research; however, its operationalization has been predominantly behavioral and has been measured through use frequency, session duration, and feature use [1,18]. Yardley et al [1] have argued for a distinction between mere interaction with a DBCI and “effective engagement,” which involves cognitive and emotional investment. Engagement captures the user’s interaction with the system but does not capture whether the resulting change is experienced as personally owned.

Self-efficacy, defined as the belief in one’s capacity to execute a behavior [40], is a well-established predictor of behavior change, and many DBCIs have been explicitly designed to enhance it [41]. Self-efficacy captures capability beliefs but does not capture meaning-making, through which change experiences are integrated into one’s self-understanding.

Autonomy support, derived from SDT [15], emphasizes the provision of choice and minimization of external control. A meta-analysis of SDT-informed health interventions confirmed the importance of autonomy-supportive contexts in sustained behavior change [42]. Autonomy support captures the motivational climate in which behavior is initiated but does not address the interpretive process through which a lapse is either fixed as a failure or repaired within one’s narrative.

The sustainability problem then falls into the gap between the existing constructs. Engagement describes interactions without ownership, self-efficacy describes capabilities without meaning, and autonomy support describes motivational quality without postlapse repair. Narrative ownership is needed because the sustainability problem concerns not only whether users continue to act but also whether they can experience and reinterpret change as their own.

This positioning also clarifies the relationship between engagement and narrative ownership, which are related but nonidentical. Engagement, at least as it is usually operationalized, is largely a matter of quantity—how much and how often a user interacts—although effective engagement also involves cognitive and emotional investment [1]; narrative ownership, by contrast, concerns the quality of that interaction: whether the resulting change is experienced as one’s own. The two can dissociate. A user may be highly engaged yet experience little ownership, as when behavior is sustained by notifications and rewards but never interpreted, a state we have termed guided but unowned behavior; conversely, ownership may persist after active engagement with the app has ended, which is precisely the outcome of interest for sustainability. Narrative ownership is therefore closer to what has been described as effective engagement than to frequency of use, and it complements rather than duplicates adjacent constructs: it differs from self-regulation, which concerns the monitoring and adjustment of behavior toward a goal [43], and from reflective motivation, which concerns deliberative evaluations and intentions [44], in that it concerns how the unfolding experience of change—including its lapses—is interpreted and held as one’s own.

Positioning narrative ownership against existing accounts of maintenance further clarifies its contribution. Theoretical reviews distinguish the determinants of initiation from those of maintenance and propose mechanisms such as maintenance self-efficacy, satisfaction with outcomes, habit formation [45], and self-regulation [8]. Identity-based accounts hold that behavior persists when it is congruent with how people see themselves [46], and relapse prevention models emphasize the interpretation of lapses [28]. Narrative ownership is not an alternative to these accounts but a complement that specifies an experiential precondition they tend to leave implicit: for a behavior to become habitual, identity-congruent, or self-regulated once external support is withdrawn, the trajectory of change—including its lapses—must first be experienced as the person’s own. In this sense, narrative ownership addresses the interpretive layer through which these maintenance mechanisms can take hold.

The 3 structural mechanisms identified in the previous section can be reinterpreted as a systematic absence of the 3 components of narrative ownership. The erosion of process framing through automated guidance reflects the loss of process framing, which is the first component of narrative ownership. The foreclosure of narrative repair through notification-driven redirection undermines narrative repair, which is the second component. The hollowing of authorship through externalized success attribution weakens authorship, which is the third component. Thus, the sustainability paradox is not merely a problem of declining engagement but a problem of absent ownership. Addressing this issue requires design principles that protect and cultivate narrative ownership.

The same interpretive processes that make narrative ownership adaptive can also operate in the opposite direction. Reflecting on a lapse may help a user reframe it as a surmountable step, but it may equally consolidate a negative self-attribution—“I always fail at this”—because positive and negative self-narratives are built through the same mechanisms [47,48]. Narrative ownership is therefore not inherently benign; cultivating it without regard to content risks reinforcing the very self-blame it aims to dissolve. This carries a direct design implication: features that open space for reflection should also be able to detect markers of negative self-framing and, where appropriate, gently challenge rather than merely record them. Because behavior change is often also an attempt to change something about oneself, a narrative ownership–informed DBCI should be designed to support adaptive, rather than corrosive, self-interpretation.

If the sustainability paradox arises from a structural deficit in narrative ownership, then addressing it requires design principles that actively protect and cultivate narrative ownership within DBCIs. The following principles are not prescriptions for specific interface features, but conceptual guidelines that can be instantiated in diverse ways depending on the behavioral domain, target population, and technological platform [49]. They are also distinct from established BCTs: although self-monitoring, action planning, coping planning, values clarification, and relapse prevention exercises [5] primarily structure what users do, these principles structure how users interpret what happens to them, including their lapses. Contemporary DBCI designs have largely pursued frictionless interactions, minimizing the cognitive effort required of users at each step. The principles proposed here represent a deliberate counterpoint: they introduce what might be called “good friction”: minimal, intentional interruption that supports reflection and self-interpretation without creating unnecessary burden, with the aim of promoting long-term ownership rather than short-term ease [50]. These principles complement rather than replace the technological efficiency of current DBCIs.

The first principle targets the erosion of process framing. Rather than responding to every detected lapse with immediate algorithmic adjustments, DBCIs can incorporate brief reflective pauses, which are moments during which users are invited to interpret the lapse before receiving corrective guidance. A reflective pause might take the form of a short prompt: “Your activity was lower than usual today. Before we adjust your plan, would you take a moment to reflect on what happened?”

The critical design feature is not the specific wording but the temporal space it creates: a gap between the lapse event and the system’s corrective response, in which the user can situate the experience within their own trajectory of change.

This principle draws on the concept of effective engagement [1], which emphasizes that meaningful interactions involve cognitive and emotional processing and not merely behavioral compliance. It also resonates with calls to reconceptualize just-in-time adaptive interventions (JITAIs) [16,20] in ways that balance responsiveness with opportunities for user reflection. An important caveat is warranted: reflective pauses introduce an additional user burden, and the design challenge is to make reflection brief, optional, and contextually sensitive, while making it available as an invitation rather than imposing it as a requirement [51]. Adaptive systems can learn when a user is receptive to reflection [52] and offer selective pauses.

The second principle addresses the foreclosure of narrative repair. Rather than moving users past their lapses as quickly as possible, DBCIs can incorporate features that support the reinterpretation of lapses as learning opportunities. Postlapse prompts such as “Looking back at what happened, what would you like to do differently next time?” invite users to construct a narrative connection between a lapse and its future. Over time, the accumulation of such reflections creates what might be called a learning timeline: a record organized not as a list of failures, but as a chronological narrative of challenges encountered and lessons derived.

This principle is informed by the relapse prevention literature, in which the reinterpretation of lapses as correctable events is recognized as critical for preventing escalation to full relapse [28,29]. It is also connected to qualitative research showing that individuals who redefine relapse as part of a recovered social identity are more likely to achieve sustained abstinence [53]. In the digital context, lapse reinterpretation features translate these clinical insights into scalable design elements.

The third principle targets the hollowing of authorship. To cultivate authorship, DBCIs can incorporate meaning-linking interface design elements that explicitly connect behavioral goals to users’ self-defined values. At goal setting, users might articulate why a particular target matters to them: “What makes this goal important in your life right now?”

The resulting statements are then displayed alongside quantitative progress indicators to ensure that numerical achievements are continuously linked to their narrative meaning. For example, a step count display may be accompanied by the user’s own words: “I want to be able to play with my grandchildren without getting out of breath.”

This juxtaposition of metrics and meanings anchors the experience of success in the user’s authorship rather than in the app’s reward architecture.

This principle resonates with value-based approaches in health behavior research [54] and with the capabilities approach to person-centered care [55]. It also shares conceptual grounding with the value clarification process in acceptance and commitment therapy, which helps individuals identify and commit to personally meaningful directions for action [56]. The key distinction is that meaning-linking interfaces do not automate this process but embed it within the user’s own narrative, ensuring that values remain personally articulated rather than system-generated.

Table 1 summarizes the correspondence among the 3 components of narrative ownership, the structural mechanisms that undermine them, and the design principles proposed to address them.

Figure 1 illustrates the contrast between a conventional DBCI design pathway in which lapses are algorithmically corrected without user interpretation (Figure 1A) and a narrative ownership–informed pathway in which each design principle activates a component of narrative ownership (process framing, narrative repair, or authorship), which in turn supports a hypothesized effect that contributes to behavioral maintenance after the intervention ends (Figure 1B).

The design principles yielded the following 3 testable propositions. First, a DBCI that incorporates reflective pauses following detected lapses will produce higher rates of behavioral maintenance at 3 and 6 months after intervention than a functionally equivalent DBCI that delivers immediate algorithmic correction. The hypothesized mechanism is that reflective pauses preserve process framing. Second, users of a DBCI that includes lapse reinterpretation features will report lower self-stigma following lapses and faster re-engagement with behavioral targets than users of a DBCI that uses standard notification-based redirection. The hypothesized mechanism is that narrative repair reduces the abstinence violation effect [28]. Finally, users of a DBCI with meaning-linking interfaces will demonstrate higher behavioral persistence after discontinuing app use than users of a DBCI with standard gamification. The hypothesized mechanism is that meaning-linking cultivates authorship that persists after the external scaffolding is removed. Testing these propositions requires explicit outcome definitions. Here, behavioral maintenance refers to meeting the target behavior at 3 and 6 months after the intervention ends; re-engagement refers to the time from a lapse to resumption of the target behavior; and authorship refers to the degree to which users attribute the change to themselves rather than to the app and, beyond attribution, can describe it in their own words and connect it to goals that matter to them personally or relationally; it is operationalized separately from autonomous motivation and self-efficacy.

These propositions can be tested independently or using factorial designs that evaluate the individual and combined effects of each component [57]. Empirical progress also requires the development of measures of narrative ownership (with subscales for process framing, narrative repair, and authorship)—for example, process framing items might assess whether users view a lapse as part of a longer change trajectory, and authorship items might assess whether the change feels “my own” rather than “caused by the app,” whether users can describe it in their own words, and whether they connect it to goals that matter to them personally or relationally—that demonstrate discriminant validity relative to self-efficacy [40], autonomous motivation [15], and engagement [18]. Existing instruments for narrative identity [58] and the recently developed Narrative Identity Self-Evaluation Scale [59] may provide useful starting points.

This viewpoint has some limitations that warrant further discussion. First, the framework is entirely conceptual, and the design principles have not been empirically tested. Second, reflective pauses and lapse reinterpretation prompts require cognitive effort and may reduce engagement among users with low motivation or under high cognitive load [51]. For some users—including those experiencing depression, anxiety, low health literacy, or socioeconomic stress—repeated reflection after lapses may heighten guilt or rumination rather than promote learning, so reflective prompts should be tailored to the user and remain suppressible. Whether these features can be designed to be brief, adaptive, and optional, without diluting their reflective functions, remains an empirical question. Third, the framework was developed primarily based on chronic relapse-prone behavioral domains. However, its applicability to acute or 1-time behavior changes may be limited.

Fourth, the expression of narrative ownership varies across cultural contexts; the emphasis on individual authorship may reflect individualistic Western assumptions [39], and in collectivist contexts ownership may instead be familial or communal, so cross-cultural validation is essential for the framework’s generalizability [60]. Practical barriers also warrant consideration: reflective and meaning-linking features compete for limited screen space, may add to user fatigue, and are difficult to integrate into commercial apps built for frictionless engagement. Because narrative data are sensitive—spanning personal values, relationships, and illness experiences—their collection further raises privacy and data governance obligations that interventions must address [61]. At a minimum, such interventions should collect no more narrative data than necessary, process reflective content on the device where this is possible, encrypt what is stored, seek explicit and specific consent, and give users a straightforward way to review, export, or delete their own entries. Finally, the framework does not address the structural and socioeconomic determinants that constrain individuals’ capacity to engage in behavior change, regardless of narrative ownership [60]. It is intended to complement, not replace, the attention paid to these factors.