Productive struggle in children's app-based digital reading

1 Introduction

In recent years, public and policy discourse has increasingly framed children's digital media engagement as a primary driver of developmental risk, often linking screen use to rising concerns about anxiety, depression, attentional difficulties, and diminished wellbeing. Yet empirical evidence paints a more nuanced picture. Large-scale studies frequently report small or context-dependent effects, suggesting that the developmental implications of digital media use depend less on exposure alone and more on the conditions under which digital engagement occurs. This divergence between public perception and scientific evidence highlights a pressing need to examine the mechanisms and moderators that shape children's experiences in digital environments, particularly during early developmental periods when patterns of engagement, persistence, and self-regulation are still forming.

While much of the debate has centered on adolescents' social media use, comparatively less attention has been paid to younger children's encounters with digitally mediated challenge. Early childhood digital environments are often designed to minimize friction (Ekambaranathan et al., 2023), accelerate gratification, or simplify tasks, potentially limiting opportunities for effortful engagement. From a developmental perspective, however, the capacity to navigate difficulty and persist through challenge constitutes a foundational form of resilience. In digital contexts where disengagement is immediate and effortless, understanding how design features shape children's responses to difficulty becomes critical. Rather than treating digital media as uniformly harmful or benign, it is therefore essential to investigate how specific design conditions may either undermine or support the development of persistence, agency, and adaptive help-seeking.

The present study contributes to this agenda by examining patterns of engagement, perceived difficulty, scaffold use, and disengagement within a large-scale digital reading environment for young children. By advancing a design-mediated model of productive struggle, we investigate how calibrated challenge, embedded support, and learner agency interact within digitally mediated contexts where real-time adult guidance cannot be presumed. In doing so, this study shifts attention away from screen time alone and toward the design features of digital environments. Rather than treating screen use as a monolithic exposure, the study argues that what children are doing on screens may matter more than how long they spend there. This perspective aims to inform more balanced discussions of children's digital media use, grounded in developmental theory and behavioral evidence rather than moral concern or technological optimism.

Experiences of challenge play a pivotal role in cognitive development. Within developmental psychology, challenge has long been understood as a condition for learning when appropriately structured and supported. Vygotsky's (1978) concept of the zone of proximal development (ZPD) provides a foundational account of this process. The ZPD refers to the distance between a child's independent performance and their potential performance when supported by a more knowledgeable other. Learning, from this perspective, occurs when tasks are positioned slightly beyond current competence and are scaffolded through contingent guidance. Struggle, when accompanied by support, is therefore developmentally productive.

Yet such moments of productive struggle do not occur in isolation. They are embedded within a constellation of relational, institutional, and cultural contexts that shape both the availability and the quality of support. Ecological systems theory has long underscored that child development is influenced by multiple, interacting layers of environment, extending beyond the individual child to encompass family dynamics, community characteristics, and broader sociocultural and policy structures (Bronfenbrenner, 1979). Research across developmental science and early childhood education has demonstrated that children's social and emotional development is shaped not only within intimate family settings but also by meso-level features such as neighborhood conditions and community services, as well as macro-level forces including policy frameworks, social norms, and cultural belief systems (Xu et al., 2025). Indeed, healthy social-emotional development and long-term outcomes are shaped by determinants operating across these multiple environmental levels (Poon et al., 2022).

Recent developmental scholarship has further refined this perspective through dynamic systems and lifeworld approach (Kärtner and Köster, 2024). From this vantage point, children's development is understood as emerging from their ongoing interactions with the social world. The lifeworld encompasses the entirety of forces that constitute children's lived experience in a given situation, including relational, material, cultural, and historical conditions, which together shape behavior and developmental trajectories over time (Kärtner and Köster, 2024). In this sense, challenge is not merely a property of a task calibrated to individual competence, but a situated experience arising within a broader ecology of relationships, expectations, and resources.

Scaffolding is typically defined as the temporary, adaptive support provided by a more knowledgeable other that enables a child to accomplish a task they could not complete alone (Wood et al., 1976). Such support operates cognitively, through prompts and strategy cues, and affectively, through reassurance and motivational regulation. Importantly, scaffolding not only supports task completion but also contributes to the development of self-regulation, including persistence, strategic flexibility, and adaptive help-seeking. Within this framework, productive struggle may be defined as sustained engagement with cognitively demanding material that promotes conceptual understanding when supported within an optimal range of challenge (Warshauer, 2015; Young et al., 2024). In early childhood, these processes are typically mediated by caregivers and teachers who interpret children's cognitive and emotional signals and adjust support accordingly. The calibration of challenge, and the responsiveness of support, are therefore central mechanisms through which developmental change occurs.

Today, children increasingly encounter learning contexts that are digitally mediated. Interactive applications, game-based platforms, and artificial intelligence (AI)-enabled systems now form part of many children's everyday learning ecologies. In such environments, the structure of challenge differs in important ways. When difficulty arises, children may disengage by exiting an activity or shifting to alternative digital content. In the absence of immediate interpersonal scaffolding, persistence depends more heavily on individual self-regulatory capacity. Research suggests that digital environments can place substantial demands on children's attentional control and motivational regulation (Aparicio et al., 2017), raising questions about how challenge is navigated in these contexts.

Recent advances in generative artificial intelligence (AI) have intensified interest in personalized learning systems that dynamically adapt task difficulty and content (Li and Wong, 2023; Akintola et al., 2025). These systems aim to optimize learning trajectories by responding to performance data. However, from a developmental perspective, adaptation must be evaluated not only in terms of efficiency but in relation to how it structures challenge. A substantial body of research indicates that deep learning is supported when tasks are neither too easy nor overwhelmingly difficult but are calibrated within an optimal zone that stretches competence while maintaining engagement (Vygotsky, 1978; Wood et al., 1976; Anghileri, 2006). Excessive simplification may reduce opportunities for effortful processing and strategy development, while excessive difficulty may prompt disengagement. The developmental significance of digital personalization therefore depends on how it preserves, rather than eliminates, productive challenge.

The importance of calibrated challenge is echoed across several influential theories of learning and motivation. Cognitive Load Theory suggests that learning is impaired when task demands exceed the learner's available cognitive resources, highlighting the need to balance challenge with processing capacity (Sweller, 1988). Similarly, Flow Theory proposes that optimal engagement occurs when task difficulty is closely matched to an individual's skills, whereas excessive challenge may generate anxiety and insufficient challenge may result in boredom (Csikszentmihalyi, 1990). Self-Determination Theory further emphasizes that sustained motivation depends upon learners experiencing both competence and autonomy (Ryan and Deci, 2000). Taken together, these perspectives suggest that educational environments should not seek to eliminate difficulty, but rather to calibrate challenge in ways that maintain engagement, support competence, and encourage persistence. This principle is central to contemporary conceptualizations of productive struggle.

Empirical research on productive struggle has demonstrated that perseverance, sense-making, and guided reflection contribute to conceptual growth (Warshauer, 2015; Sengupta-Irving and Agarwal, 2017; Baker et al., 2020). These processes are linked to the development of executive functions and motivational resilience. However, much of this literature presupposes the presence of an adult who can provide real-time interpretive support. Comparatively less attention has been directed toward understanding how productive struggle unfolds in digital environments where adult mediation may be absent, intermittent, or indirect. Although digital games have shown potential to scaffold problem-solving and reasoning (Sun et al., 2011; Kao et al., 2017; Barzilai and Ina, 2014), the mechanisms through which digital design can sustain engagement with challenge remain underexplored. This issue carries implications for developmental equity. If sustained engagement with cognitively demanding tasks depends primarily on access to adult co-regulation, then opportunities for deep learning may be unevenly distributed across children's home and school contexts. Embedding elements of calibrated challenge and scaffolded support within digital design may therefore contribute to more consistent access to developmentally supportive learning experiences.

The present study investigates whether productive struggle can be supported through intentional design within a game-based digital reading platform for young children. In this study, productive struggle is operationalized behaviorally through patterns of persistence, scaffold utilization, and disengagement in response to perceived difficulty. Rather than conceptualizing technology as a replacement for adult guidance, we examine whether design features such as adaptive difficulty, embedded hints, and structured learner choice can partially replicate aspects of scaffolded challenge in the absence of real-time interpersonal support.

This study is novel in both its conceptual framing and its exploration of an emerging educational phenomenon. While productive struggle has been extensively examined in classroom settings where teachers provide real-time scaffolding, comparatively little attention has been paid to how challenge is encountered and navigated in digital learning environments where adult mediation cannot be presumed. Existing research provides limited guidance, particularly in relation to early childhood digital reading, the role of design in supporting productive struggle beyond teacher-mediated settings, and large-scale behavioral patterns of learner engagement with challenge. This gap is particularly important given the well-documented novelty effect, whereby new technologies may temporarily increase engagement, motivation, and persistence because they are perceived as new rather than because they embody effective pedagogical principles (Elston, 2021; Jeno et al., 2019). As children's learning increasingly occurs through apps, digital books, games, and AI-enabled technologies (Bus et al., 2020), a fundamental question remains largely unexplored: what does productive struggle look like in digital environments where adult mediation cannot be presumed? Before researchers can determine whether particular design features are effective through experimental investigation, there is a need to establish conceptual frameworks and descriptive evidence that help define the phenomenon itself. The present study responds to this challenge. It does not seek to test a predefined causal model or establish the effectiveness of a particular intervention. Instead, it offers an exploratory and agenda-setting examination of productive struggle in a digital reading environment, advancing a design-mediated conceptual model intended to inform future experimental and longitudinal research.

2 Productive struggle, mediation, and learning in digital contexts

Despite growing interest in productive struggle and digital learning, several important gaps remain. First, while productive struggle has been extensively examined in mathematics classroom settings, comparatively little research has explored how young children encounter and navigate challenge within digital reading environments. Second, existing theoretical accounts largely assume the presence of teachers, caregivers, or other knowledgeable others who provide real-time scaffolding, leaving unanswered questions about how productive struggle may be supported when adult mediation cannot be presumed. Third, there is limited large-scale behavioral evidence describing how young learners persist, seek support, or disengage when confronted with challenge in digitally mediated learning contexts. These gaps are particularly significant as digital technologies, AI-enabled systems, and interactive media become increasingly embedded within children's everyday learning ecologies (Bus et al., 2020). Given the exploratory nature of this emerging area, the present study aims to provide an initial conceptual and empirical examination of an underexplored phenomenon. By advancing a design-mediated model of productive struggle and examining behavioral patterns across 419,700 digital reading sessions, the study seeks to establish a foundation for future experimental and longitudinal research investigating how digital environments may be designed to support meaningful engagement with challenge.

2.1 Productive struggle as a condition for deep learning

The notion that learning is strengthened through engagement with difficulty is well established within educational psychology and the learning sciences, with theoretical roots in constructivist accounts of cognitive conflict (Vygotsky, 1978; Wood et al., 1976), where difficulty is viewed as a productive tension that prompts learners to reorganize existing knowledge structures. Contemporary elaborations of productive struggle have been most extensively developed within mathematics education, where it is understood as learners' sustained engagement with challenging tasks that require sense-making rather than procedural recall (Warshauer, 2015; Young et al., 2024). However, productive struggle is not a mathematics-specific construct. Rather, it is a learning construct that mathematics researchers have perationalized most thoroughly. At its core, productive struggle concerns learners' engagement with challenges that require effort, persistence, reflection, and the construction of meaning. These processes are equally relevant to reading, where learners must navigate unfamiliar vocabulary, monitor comprehension, resolve ambiguity, make inferences, and persist when understanding does not emerge immediately.

This broader interpretation aligns with insights from the Science of Learning, which emphasizes that deep and transferable learning depends upon active engagement, effortful processing, and knowledge construction rather than passive consumption (Hirsh-Pasek et al., 2015). As Hirsh-Pasek et al. (2015) argue, educational technologies should be designed not simply to engage learners, but to embody principles derived from how learning occurs. This perspective is particularly important in early childhood, where apps, digital books, games, and interactive media have become increasingly prominent features of children's learning environments (Bus et al., 2020). Yet despite the growing prevalence of these technologies, relatively little attention has been paid to how they might support the development of habits of mind associated with deep learning, including perseverance, resilience, metacognitive awareness, and a willingness to engage with challenge.

This question becomes especially salient in digital reading environments. Unlike traditional classroom settings, where teachers and caregivers may encourage learners to persist through difficulty, digital environments often allow children to abandon an activity with a single tap. Why should a child persist when a story becomes difficult if exiting the experience is easier than continuing? This question lies at the heart of the present study. If productive struggle is indeed a condition under which meaningful learning occurs, then understanding how challenge, support, and learner agency interact within digital reading environments becomes increasingly important. The present study therefore explores whether productive struggle may offer a useful conceptual lens for understanding how young children engage with challenge in app-based reading environments, while recognizing that this remains an emerging and largely uncharted area of inquiry.

2.2 Mediation, scaffolding, and the zone of proximal development

The role of mediation is foundational to most accounts of productive struggle. Vygotsky's (1978) concept of the ZPD frames learning as a socially mediated process in which learners can achieve more with guidance than they can independently. Within this framework, struggle becomes productive when a more knowledgeable other provides timely and contingent scaffolding that supports the learner in moving beyond their current level of competence. Subsequent research has elaborated on the forms such scaffolding may take, emphasizing prompts, questioning, modeling, and emotional support as mechanisms through which engagement with challenge is sustained (Van de Pol et al., 2010). In classroom contexts, teachers play a critical role in interpreting learners' difficulties, deciding when to intervene, and calibrating support in ways that preserve both challenge and motivation. Young et al. (2024) identify productive struggle as essential for deepening understanding, fostering problem-solving skills, and promoting long-term learning in mathematics classrooms. They note its' role in moving learners beyond surface-level memorization to more meaningful and retained knowledge. Similarly, in home learning environments, caregivers often mediate children's encounters with difficulty through reassurance, shared problem-solving, and encouragement to persist. However, reliance on adult mediation introduces substantial variability in access to productive struggle. The quality and consistency of scaffolding depend on adults' pedagogical knowledge, confidence, time, and emotional resources, all of which vary widely across educational settings and home contexts. As a result, opportunities to benefit from productive struggle are unevenly distributed, raising important questions about equity, access, and the conditions for productive struggle.

2.3 Productive struggle beyond adult-mediated settings

Digital environments differ from classrooms or home contexts in that adult-mediation cannot be presumed. Digital tools often afford learners greater control over pacing, content selection, and continuation, which may support autonomy and motivation. At the same time, this control introduces the possibility of immediate disengagement when tasks become difficult. Unlike classroom settings, where social norms, teacher expectations, and relational support encourage persistence, digital environments frequently allow learners to exit, restart, or abandon tasks at the first sign of frustration. This shift places greater responsibility on design choices to determine whether difficulty becomes a site of productive engagement or a trigger for disengagement. These affordances raise a critical design question: how can digital learning environments be structured to invite learners to grapple with difficulty rather than bypass it?

2.4 From mediation to design: a conceptual reframing

Emerging research in learning design and human–computer interaction suggests that digital systems can be intentionally structured to support learning processes traditionally associated with human mediation. Features such as adjustable difficulty, optional hints, and embedded prompts (all correlates to good pedagogy in classrooms) can provide forms of contingent support that learners access at their own discretion. Importantly, such supports differ from direct instruction in that they preserve learner agency, allowing individuals to decide whether, when, and how to seek assistance. This perspective invites a conceptual reframing of productive struggle, shifting attention from interpersonal mediation alone to the role of intentional design in shaping learners' encounters with challenge and raises the question of whether digitally mediated challenge should be understood as an extension of productive struggle or as a related but distinct phenomenon. The present study does not seek to resolve this question, but rather to provide an initial conceptual and empirical basis for investigating it. Rather than locating struggle exclusively within interpersonal interactions between learners and adults, it becomes possible to consider how struggle may be shaped through interaction with designed systems. Design-mediated support does not seek to replace teachers or caregivers, but may partially assume mediating functions by calibrating challenge, making support visible and accessible, and preserving learners' sense of control. This reframing shifts attention away from learner deficits and toward the responsibilities and possibilities of design.

The ZPD has long been used to inform the design of educational software. Drawing on a Vygotskian framework, Luckin and Du Boulay (1999) developed the Ecolab system and introduced two complementary constructs to operationalize the ZPD in digital contexts: the Zone of Available Assistance (ZAA) and the Zone of Proximal Adjustment (ZPA). The ZAA refers to “the variety of qualities and quantities of assistance which need to be available to enable the more able partner (whether human or computer) to offer appropriate assistance to the child” (Luckin and Du Boulay, 1999, p. 200). With the advent of generative AI, it is increasingly plausible that digital systems can provide such scalable and personalized assistance in real time. However, Luckin and Du Boulay (1999) emphasized that the availability of assistance alone is insufficient. The ZPA, which concerns selecting the appropriate form and level of assistance for a given learning situation, remains critical. In Ecolab, designers therefore sought not only to maximize the ZAA, but to align it as closely as possible with the learner's ZPD through careful design decisions. This distinction underscores the pedagogical importance of design in mediating how and when support is made available to learners.

2.5 A design-mediated model of productive struggle

In response, the present study advances a design-mediated model of productive struggle in digital learning environments. The model, presented in Figure 1, synthesizes insights from research on productive struggle, scaffolding, and learner agency to conceptualize how challenge may be encountered and sustained through intentional design when real-time adult mediation cannot be presumed.

The model foregrounds design conditions as the primary mediating layer through which productive struggle may be invited rather than imposed. Three interrelated design conditions are central. The model does not assume a fixed hierarchical or causal relationship among these conditions. Rather, they are conceptualized as mutually shaping aspects of the learner–system interaction whose relative importance may vary across learners, tasks, and contexts.

Calibrated challenge refers to the structuring of task difficulty within developmentally appropriate bounds, allowing learners to encounter material that is neither trivially easy nor prohibitively difficult, and to experience challenge as subjectively meaningful rather than externally enforced. Embedded scaffolding captures the provision of optional, nondirective supports that are available at the point of need, enabling learners to seek assistance without disrupting their sense of agency or flow. Learner agency refers to the preservation of choice over content, pacing, modality, and continuation, ensuring that engagement with challenge remains voluntary rather than compulsory.

These design conditions shape the learner–system interaction, which constitutes the conceptual core of the model. At this level, productive struggle is understood as a dynamic process arising from the learner's appraisal of perceived difficulty in relation to available support and their own capacity to choose how to proceed. This interaction is not directly observable and should not be equated with discrete behaviors or outcomes. Rather, it represents an ongoing decision-making process in which learners weigh challenge against support and determine whether to persist, seek help, or disengage.

The lower layer of the model identifies observable behavioral indicators that may be consistent with engagement in design-mediated struggle. These include persistence through task completion, strategic use of embedded scaffolds, and patterns of early disengagement when challenge exceeds the learner's capacity or when available supports are insufficient. Importantly, these indicators are treated as signals rather than definitive measures of productive struggle. Similar behaviors may arise from a range of contextual and motivational factors unrelated to learning-oriented struggle, and their interpretation therefore requires caution.

The model also acknowledges the role of contextual moderators, such as language proficiency, cultural familiarity with content, and the physical or social environment in which digital engagement occurs. These factors may amplify or constrain the effectiveness of design-mediated support and help explain variation in learners' responses to challenge. By explicitly situating contextual moderators alongside the core design conditions, the model resists universal claims and foregrounds the importance of sensitivity to the learner's lived contexts.

2.6 Design, equity, and AI-mediated personalization

By conceptualizing productive struggle as a design-mediated process, the model carries important implications for equity and for the development of AI-enabled learning systems. If opportunities to engage in productive struggle depend solely on access to high-quality adult mediation, then such opportunities are unevenly distributed. Design-mediated approaches offer the possibility of embedding baseline access to challenge and support within digital systems, while acknowledging that design alone cannot eliminate broader structural inequities.

At the same time, the model provides a lens through which to critically examine claims about AI-driven personalization. While adaptive systems are often presented as mechanisms for smoothing learning pathways, an alternative perspective suggests that personalization should instead aim to calibrate difficulty in ways that preserve challenge and invite sense-making. From this standpoint, the educational value of AI in this context lies not in removing struggle, but in supporting it responsibly through design choices grounded in established theories of learning. Together, this conceptual framing provides the foundation for the empirical analysis that follows, examining how young children engage with digitally mediated challenge in a game-based reading environment, focusing on patterns of persistence, scaffold use, and disengagement in contexts where real-time adult mediation cannot be presumed.

3 Materials and method

3.1 Study context and design

This study employed a naturalistic, large-scale observational design using de-identified dashboard data drawn from a game-based digital reading environment embedded within a children's edutainment platform. The design was exploratory and descriptive in nature, aimed at examining patterns of engagement, persistence, and disengagement as children interacted with digitally mediated reading challenges. Rather than seeking to infer internal cognitive states or causal mechanisms, the study focused on observable behavioral indicators that may be consistent with engagement in design-mediated productive struggle.

The study was guided by a design-oriented conceptual framework (see Figure 1), which treats productive struggle as an emergent experience shaped by task difficulty, available support, and learner agency within the digital environment. As such, the analysis foregrounds system-level design features and learner–system interactions, rather than individual learner traits. Guided by this framework, the study addressed three research questions:

RQ1: how do young learners engage with calibrated challenge in a game-based digital reading environment when real-time adult mediation cannot be presumed? RQ2: what patterns of persistence, scaffold use, and disengagement emerge when children encounter tasks they perceive as challenging within this environment? RQ3: how do contextual factors, particularly language availability, shape children's engagement with challenge and their decisions to persist or disengage?

3.2 Participants

Participants comprised young users of the digital reading platform who engaged with AI-generated storybook content in Let's Story during a defined 30-day period from 1 to 30 November 2024. The platform is designed for children aged 4–9 years, and users self-reported their age range upon entering the app. As data were fully de-identified, individual ages could not be verified, and analyses were conducted using age-band groupings as provided by the platform. Participation in data collection was governed by the platform's terms and conditions, to which caregivers' consent for the data to be used for research purposes at the point of app registration. No personally identifiable information was collected, stored, or analyzed for the purposes of this study. Only aggregated, summary-level data were accessed, ensuring compliance with data protection and ethical standards for research involving children's digital data. Story sessions were available across 18 languages. As no individual-level language background data were collected and the study was conducted in a naturalistic setting, language was treated as a characteristic of the reading session rather than as an indicator of first-language status.

3.3 Digital reading environment and design features

The digital reading environment allows users to generate personalized storybooks by selecting narrative variables such as characters, settings, and plot elements. Stories are generated at one of three complexity levels aligned with broad developmental reading bands, namely, 4–5 years, 5–7 years, and 7–9 years. These bands are intended to approximate developmentally appropriate challenge rather than to function as precise measures of reading ability.

During each reading session, users may choose to read the story independently or listen to an AI-generated voiceover. Embedded within the story are optional scaffolding features, including highlighted vocabulary linked to a dictionary function and discussion prompts that appear intermittently to support comprehension and reflection. These supports are nonmandatory and learner-initiated, allowing users to decide whether and when to access assistance.

Importantly, users retain full control over pacing and continuation. At any point during the session, they may pause, exit, or abandon the story, making persistence voluntary rather than enforced behavior. These design features were intentionally developed to preserve learner agency while providing access to calibrated challenge and contingent support, consistent with the study's focus on design-mediated struggle.

3.4 Data sources and measures

The primary data source was the platform's internal analytics dashboard, which records aggregated user interactions with the digital reading environment. For the period under study, the dataset included the total number of story sessions initiated and completed, points of disengagement for incomplete sessions, user-selected language, and post-session ratings of perceived story difficulty. Although data collection on the platform is ongoing (currently seven million story sessions), the present analysis focuses on a defined 30-day period following the launch of the digital storybook feature. This window was selected to ensure consistency in design features, user experience, and system behavior, and to examine early patterns of engagement in relation to research-informed design choices. Focusing on this bounded period allows for a clearer interpretation of how learners interact with calibrated challenge and embedded scaffolding under stable design conditions, while avoiding confounds introduced by iterative platform updates or evolving user familiarity over time.

Perceived difficulty was measured using a three-point self-report scale presented at the end of each story session, with response options labeled “too easy,” “just right,” and “too difficult.” While this measure does not capture objective task difficulty or comprehension, it provides insight into learners' subjective experience of challenge. Difficulty ratings were available for a subset of sessions and therefore may not be representative of all users or reading experiences on the platform.

Behavioral indicators examined in the analysis included story completion rates, the proportion of users disengaging at different stages of the story, and aggregate patterns of scaffold use. These indicators were interpreted cautiously as signals that may be consistent with engagement in productive struggle, rather than as direct measures of learning or self-regulation.

3.5 Adult mediation and epistemic constraints

A critical limitation of dashboard-level data is that it does not capture the social context in which digital engagement occurs. It is therefore not possible to determine whether a caregiver, teacher, or other adult was present during individual reading sessions, nor the nature or extent of any support they may have provided. As such, real-time adult mediation cannot be presumed to be either present or absent. This epistemic constraint is central to the interpretation of findings. Rather than making claims about learning in the absence of adult support, the study examines how children interact with digitally mediated challenge in contexts where adult mediation is not directly observable and cannot be presumed. The focus is therefore on the potential of design features to support persistence and sense-making under conditions of uncertainty regarding external support.

3.6 Analytical approach

Analyses were conducted at an aggregate level to describe patterns of engagement across the user population. In line with the theoretical framing of productive struggle, behavioral engagement metrics were treated as indirect indicators of children's responses to challenge. Descriptive statistics were used to examine distributions of perceived difficulty ratings, completion rates, and disengagement points across story sessions. Perceived difficulty was interpreted as a subjective index of experienced challenge, while completion and disengagement patterns were examined as behavioral markers of persistence or withdrawal. Productive struggle was operationalized as engagement with content rated as challenging that was not followed by immediate disengagement.

As these measures are derived from behavioral trace data, they provide indirect rather than direct evidence of productive struggle and do not capture the quality of scaffolding or affective experience. Where relevant, data were segmented by story language to explore whether challenge and persistence varied across language contexts. Given the exploratory and descriptive nature of the study and the limitations of the available data, no causal claims were made. Instead, findings were interpreted in relation to the study's conceptual framework, with particular attention to how observed behaviors might inform the design of digital learning environments intended to support productive struggle.

4 Results

The results presented below describe aggregate patterns of engagement, perceived difficulty, persistence, and disengagement observed during a 30-day period following the launch of the digital storybook feature, from 1 to 30 November 2024. In keeping with the exploratory nature of the study and the use of dashboard-level data, findings are reported descriptively and without attribution of intent, learning outcomes, or motivational states. Interpretation of these patterns in relation to design-mediated productive struggle is taken up in the subsequent discussion.

4.1 Engagement with calibrated challenge

Across the 30-day period, a total of 155,700 story sessions included a post-session rating of perceived story complexity. Of these, 73,045 sessions, representing 46.91% of ratings, were classified by users as “just right.” A further 58,271 sessions, or 37.42%, were rated as “too easy,” while 24,399 sessions, corresponding to 15.67%, were rated as “too difficult.”

These distributions indicate that nearly half of the rated sessions were perceived as appropriately challenging within the available complexity bands, while a smaller proportion were experienced as exceeding the learner's perceived capacity. Ratings were captured across all available languages and age bands, reflecting subjective appraisals of difficulty rather than externally assessed reading level.

4.2 Story initiation and completion

During the same period, 419,700 story sessions were initiated within the digital reading environment. Of these sessions, 329,591 were completed in full, yielding an overall completion rate of 78.53%. Completion rates reflect aggregate engagement across users and sessions and do not account for individual differences in reading ability, prior exposure, or contextual factors. Nevertheless, the overall distribution provides an indication of persistence within the digital reading environment during the defined analysis window.

4.3 Patterns of disengagement during incomplete sessions

For sessions that were not completed, disengagement was examined in relation to progression through the story. Incomplete sessions were categorized according to the point at which users exited the story, defined as the first third of the story (0%−33% complete), the middle third (33%−66% complete), or the final third (66%−99% complete).

Figure 2 shows the distribution of user attrition across digital storybook progression. The majority of users completed the storybook (78.53%), while 15.99% exited within the first third (0%−33%), 3.30% exited between one-third and two-thirds (33%−66%), and 2.18% exited during the final third (66%−99%). Across the 30-day analysis period, this pattern indicates that disengagement was most likely to occur early in the reading experience, with progressively fewer exits observed as users advanced through the story. Few sessions were abandoned close to completion.

4.4 Use of embedded scaffolding features

Interaction with optional scaffolding features was examined across the dataset. On average, users accessed the embedded dictionary feature 4.58 times/ session. Discussion prompts, which appear intermittently within stories, were accessed an average of 1.36 times/session. These interactions demonstrate that users made use of nonmandatory support features during reading sessions. However, the aggregate nature of the data does not allow for analysis of the timing, sequencing, or individual-level strategic use of scaffolding in relation to perceived difficulty or disengagement.

4.5 Language as a contextual moderator

Story sessions during the analysis period were conducted across 18 languages. Variation in early disengagement and completion patterns was observed across story languages. Given the absence of individual-level language background data, these observations are reported descriptively and should not be interpreted as evidence of first-language effects. Rather, language is treated as a contextual characteristic of the reading environment that may shape how challenge is experienced and navigated.

Given variation in language availability across regions and the absence of individual-level language background data, these findings are reported descriptively and interpreted cautiously. Language is therefore treated as a contextual moderator rather than as an independent explanatory variable. These findings provide a descriptive foundation for the discussion that follows, which examines their implications for understanding design-mediated productive struggle in digital learning contexts.

5 Discussion

This study set out to examine how productive struggle might be supported through design within a game-based digital reading environment for young children, focusing on contexts in which real-time adult mediation cannot be presumed. Guided by a design-mediated conceptual model of productive struggle, the analysis addressed three interrelated research questions concerning learners' engagement with calibrated challenge, their patterns of persistence, scaffold use, and disengagement, and the role of contextual factors, particularly language availability, in shaping these behaviors.

The first research question asked how young learners engage with calibrated challenge in a digital reading platform when adult mediation cannot be presumed. Findings relating to perceived difficulty indicate that learners encountered a range of challenge levels, with a substantial proportion experiencing the content as “just right,” while others perceived it as too easy or too difficult. Interpreted through the design-mediated model, these appraisals suggest that challenge was neither eliminated nor uniformly imposed but instead encountered within broadly calibrated design constraints that allowed learners to appraise difficulty subjectively. The substantial proportion of sessions rated as “too easy” further highlights the challenge of designing calibrated difficulty for diverse learners in large-scale digital environments.

The second research question focused on patterns of persistence, scaffold use, and disengagement when children encounter tasks they perceive as challenging. The results show that disengagement was concentrated early in the reading experience, while most learners who progressed beyond initial thresholds persisted to completion. At the same time, learners made use of optional scaffolding features, despite these being nonmandatory. Together, these patterns illuminate how learners navigate challenge, support, and choice within the digital environment, without implying learning outcomes or internal cognitive states.

The third question explored how contextual factors, including language availability, shape children's engagement with challenge and their decisions to persist or exit. Variation observed across languages, particularly in early disengagement patterns, highlights the importance of context in understanding how challenge is experienced. Rather than treating such variation as noise, the design-mediated model foregrounds contextual moderators as integral to learner–system interactions, with implications for equity and inclusion.

Taken together, the findings do not constitute evidence of productive struggle as a psychological construct. Rather, they provide insight into how design features shape the conditions under which struggle may be encountered and navigated within digital learning environments. The sections that follow interpret these patterns in greater depth, considering their implications for design-mediated struggle, equity, and AI-enabled personalization.

5.1 Design-mediated struggle and early thresholds of engagement (RQ2)

One of the most salient patterns observed in the data was the concentration of disengagement in the earliest phase of the reading experience, with substantially fewer exits occurring once learners had progressed beyond the initial third of the story. As illustrated in Figure 2, very few sessions were abandoned close to completion, suggesting that persistence, once established, was often sustained. Interpreted through the design-mediated model of productive struggle, this pattern highlights the importance of early interactional thresholds in digital learning environments.

When learners first encounter a task, they are likely to appraise the perceived difficulty of the content, the availability and visibility of support, and their own willingness to persist in the absence of external prompting. Early disengagement may therefore reflect a complex set of judgements rather than a simple lack of perseverance. Importantly, the data does not allow for inference about why learners exited at this point, nor whether disengagement reflects unproductive struggle. Alternative explanations, including caregiver interruption, exploratory use of the platform, navigation behavior, competing activities, or technical factors, cannot be ruled out and should be considered when interpreting patterns of early disengagement.

Nevertheless, the distribution of disengagement suggests that design choices at the point of entry, including initial calibration of difficulty, narrative hooks, and the salience of scaffolding features, may play a disproportionate role in shaping whether learners remain engaged long enough for struggle to potentially become productive. From a design perspective, this finding underscores the importance of how challenge is framed and introduced, rather than focusing solely on overall task difficulty.

5.2 Calibrated challenge and learners' appraisal of difficulty

Nearly half of the rated story sessions were perceived by users as “just right,” indicating that a substantial proportion of learners experienced the available content as appropriately challenging within the provided complexity bands. Although subjective ratings of difficulty cannot be equated with learning outcomes or developmental appropriateness, they offer insight into how learners appraise challenge in situ. Within the design-mediated model (see Figure 1), such appraisals are understood as a vital component of the learner–system interaction, shaping decisions about persistence, help-seeking, or disengagement.

The presence of sessions rated as “too difficult” further highlights the importance of preserving challenge rather than eliminating it through over-personalization. Rather than smoothing all difficulty, the system allowed learners to encounter content that exceeded their perceived capacity, while also retaining the option to exit or seek support. This balance is critical. Productive struggle requires that learners encounter challenge that is neither trivial nor overwhelming, but its productive potential depends on how learners experience and respond to that challenge, not on difficulty alone.

These findings suggest that calibrated challenge functioned as an experiential design feature rather than a prescriptive constraint, allowing learners to engage with difficulty in ways that were subjectively meaningful, even when adult mediation could not be presumed.

5.3 Embedded scaffolding and voluntary support-seeking

Patterns of interaction with embedded scaffolding features provide further insight into how design-mediated support may function in practice. Learners accessed dictionary definitions and discussion prompts despite these features being optional and nondirective, indicating that support was sufficiently visible and accessible to be taken up when needed. At the same time, variability in scaffold use cautions against assuming uniform engagement with support across learners or sessions.

Within the design-mediated model (see Figure 1), embedded scaffolding is intended to preserve learner agency by allowing assistance to be accessed at the learner's discretion, rather than being imposed in response to error or difficulty. This contrasts with directive forms of support that may interrupt engagement or prematurely remove challenge. While learners made use of the available scaffolding features, the present data do not permit analysis of how scaffold use related to persistence, perceived difficulty, or completion. Scaffold use is therefore interpreted cautiously as a behavioral indicator of engagement with challenge rather than as evidence of productive struggle itself. Examining these relationships remains an important direction for future research.

5.4 Contextual moderators, language, and equity

Variation observed across languages highlights the potential role of contextual moderators in shaping learners' engagement with challenge. Variation in early disengagement across language contexts suggests that language availability may shape how challenge is experienced, although the present dataset does not permit conclusions regarding the mechanisms underlying these differences. In early reading contexts, language is not merely a medium of instruction but a key determinant of whether difficulty is experienced as productive or overwhelming (Hoover and Gough, 1990).

Within the design-mediated model (see Figure 1), such factors are treated not as confounds to be controlled away, but as integral elements of learners lived contexts that interact with design conditions. These findings carry important equity implications. If productive struggle is contingent upon access to linguistically and culturally appropriate content, then design-mediated approaches must attend carefully to issues of localization and inclusion.

At the same time, the ability to embed baseline scaffolding and calibrated challenge within digital systems offers the potential to reduce reliance on high-quality adult mediation, which is unevenly available across homes and educational settings. Design-mediated struggle should therefore be understood not as a substitute for teachers or caregivers, but as a complementary pathway that may broaden access to learning experiences that value persistence and sense-making.

5.5 Implications for GenAI-driven personalization

The findings also invite reflection on how AI-driven personalization is conceptualized within digital learning environments. While personalization is often framed as a means of optimizing efficiency or minimizing difficulty, the present study suggests that such approaches risk undermining opportunities for productive struggle if they remove challenge too readily. From the perspective advanced here, the educational value of AI lies not in eliminating difficulty, but in supporting calibrated challenge while preserving learner agency and access to support.

The Design-mediated model of productive struggle (see Figure 1) offers a framework for thinking critically about how AI-enabled systems might be developed in alignment with established principles of learning such as Vygotsky (1978) ZPD. Rather than asking how AI can make learning easier, this perspective foregrounds the question of how AI-enabled systems can be designed to make learning meaningfully challenging, while remaining sensitive to learners' contexts and capacities.

5.6 Limitations and future directions

Several limitations should be acknowledged. The use of aggregated dashboard data precludes inference about individual learners' experiences, intentions, or learning outcomes, and does not capture the presence or nature of any adult mediation during reading sessions. In addition, perceived difficulty ratings were available for only a subset of story sessions. Consequently, these ratings should be interpreted as indicative rather than representative of all user experiences on the platform. Analyses were conducted at the session level rather than the individual learner level, and therefore repeated engagement by the same user could not be distinguished from engagement across multiple users. As such, the findings should be interpreted as descriptive patterns rather than evidence of causal mechanisms or developmental change. Future research would benefit from mixed methods approaches that combine behavioral data with qualitative observation, caregiver reports, or experimental manipulation of design features. Longitudinal analyses examining how engagement with digitally mediated challenge evolves over time, as well as design-based studies that systematically vary scaffolding and challenge, would further advance understanding of how productive struggle can be supported in digital contexts, particularly as AI-enabled systems become more prevalent in early learning environments. This study provides an important first step for further experimental research.

An additional consideration is the potential influence of novelty effects. Because the analysis was conducted during the first 30 days following the launch of the digital storybook feature, it is possible that some patterns of engagement, persistence, and scaffold use were influenced by users' initial interest in a new digital experience rather than by the pedagogical design features alone. Previous research suggests that novel technologies can temporarily increase motivation, engagement, and willingness to persist (Jeno et al., 2019). Consequently, the findings should be interpreted cautiously, as the extent to which observed behaviors reflect enduring responses to calibrated challenge, rather than short-term novelty effects, cannot be determined from the present dataset. Longitudinal research examining engagement over extended periods would help disentangle the effects of novelty from those attributable to design-mediated support for learning.

6 Conclusion

This study contributes to the literature in three important ways. First, it extends discussion of productive struggle beyond its traditional roots in classroom-based and mathematics-focused research, exploring its relevance to early childhood digital reading environments. Second, it advances a design-mediated model of productive struggle that foregrounds the interaction between calibrated challenge, embedded scaffolding, and learner agency in contexts where adult mediation cannot be presumed. Third, it provides large-scale naturalistic evidence that offers an initial empirical foundation for investigating how young children engage with challenge when persistence becomes a choice.

By examining how young learners engage with challenge, support, and choice within a digital reading environment, the study suggests that thoughtfully designed systems may create conditions under which struggle can be encountered and navigated productively, even when real-time adult mediation cannot be presumed. As educational technologies continue to evolve, grounding design choices in established learning theory and the science of learning will be essential to ensuring that digital personalization supports, rather than undermines, the role of struggle in meaningful learning.

More broadly, the study highlights the need for greater attention to the design conditions under which challenge is encountered in children's digital learning environments. Rather than asking whether digital technologies are inherently beneficial or harmful, future research may need to focus more directly on how challenge, support, and agency are balanced within these environments. The design-mediated model proposed here offers one starting point for that agenda.

Generative AI statement

The author(s) declared that Generative AI was used in the creation of this manuscript. To convert the reference list from APA to Chicago format and to review suitability against the special issue call.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher's note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Robert Hepach, Experimental Psychology, University of Oxford, United Kingdom

Reviewed by: Muchamad Muchibbuddin Waly, Universitas Negeri Yogyakarta, Indonesia

Pupung Puspa Ardini, State University of Gorontalo, Indonesia