Toward Deeper Learning: A Systematic Review of Cognitive Load Management in Mobile Learning

Mobile Learning Practices’ Contribution to Different IL and EL Levels (RQ1)

The first step prepared the data for IL and EL analyses. While some studies adopted the full three-construct framework, others used alternative tools. For consistency, cognitive load measures from all of the studies were mapped to the three constructs, IL, EL, and GL, based on the construct labels, item descriptions, and theoretical interpretations provided in the original studies.

Next, all mobile learning practices applied across the studies were identified. Grant’s (2019) framework, which outlines six design characteristics of effective teaching in mobile learning contexts, was adapted and adopted. It was selected for its comprehensive and multidimensional perspective on mobile learning. The original seventh dimension, “data services,” was excluded as it concerns technical infrastructure.

The six dimensions: (1) Learner empowerment, (2) Device functionalities, (3) Mobile content with pedagogical impact, (4) Tutor accessibility and interaction, (5) Contextual influence, and (6) Learner engagement, served as initial broad categories. A hybrid thematic coding approach was applied, beginning with inductive coding to identify emerging patterns, followed by deductive mapping to the predefined dimensions. Inductive coding was conducted using qualitative data from each study’s descriptions of mobile technology use, instructional design, and learning activities. In vivo coding was employed to extract important words or short phrases from the original texts (Saldaña, 2015). This was followed by a focused coding phase, during which related patterns across studies were iteratively clustered and refined into researcher-generated concepts representing mobile learning practices. Finally, these synthesized practices were deductively mapped onto Grant’s predefined dimensions, consistent with the thematic categorization procedure by Auerbach and Silverstein (2003). The resulting checklist of mobile learning practices was used for two purposes: to document commonly adopted practices for RQ1 and to support the identification of cognitive purposes associated with GL-aligned strategies for RQ2.

Then, to determine the most adopted practices, the frequency of each practice across all the instructional designs of the 53 studies was recorded; some studies employed the same practice in multiple experimental groups. The high-frequency practices were then selected for subsequent analysis of their connection with IL and EL through Epistemic Network Analyses (ENA).

The process of employing ENA (provided by https://app.epistemicnetwork.org/) focused on the contrasts between high and equal/low levels of IL and EL. The effectiveness of instructional design is commonly interpreted as its potential to yield equal or improved learning outcomes with reduced or maintained cognitive load (Paas et al., 2003; Sweller et al., 1998). Therefore, for each construct, studies were categorized as reporting “higher” or “equal/lower” levels of IL/EL compared to control groups. This binary categorization was adopted also to facilitate network comparison in ENA, which requires two analytically distinct conditions.

The following protocol was adopted:

- A construct was categorized as “higher” if all experimental groups in the study reported higher levels of that load compared to all control groups.

These categorizations, along with qualitative data from instructional design descriptions, were processed through separate ENA for IL and EL. These analyses showed contrasts in how the observed instructional practices contributed to IL and EL under two opposite conditions: burdened and unburdened mental demands. ENA requires a qualitative data set for connecting nodes, but the studies presented their designs in varying formats with examples, illustrations and repeated key terms. Thus, to ensure consistency, all instructional designs were converted into concise researcher-made versions following the same pattern: identified practices were arranged in a fixed order, with a one-sentence description based on the original version being assigned to each practice. These descriptions were derived through in vivo coding of the original study texts and prioritized authors’ original terminology, and explicitly described design features. They were intended to summarize observable instructional elements. The first author and an independent researcher independently summarized five studies. Agreement averaging 89% indicated strong inter-rater agreement (McHugh, 2012). The first author then completed the remaining coding. An example of a researcher-made version is provided in Appendix C.

Designs Reflecting Germane Processing (RQ2)

The mapping process described in RQ1 established the basis for documenting the exclusion of GL measurement. Studies that did not include any construct aligned with GL were identified as excluding GL. The subsequent analyses included all studies regardless of whether GL was directly measured.

Frameworks guiding germane processing in mobile learning emphasize schema construction and deep engagement (Crompton et al., 2018; Curum & Khedo, 2021; Grant, 2019). To investigate the design emphasis of cognitive purposes that GL-aligned designs serve, Grant’s (2019) framework was adapted for its focus on meaningful interaction and knowledge integration.

A shortened version of Grant’s checklist (see Appendix B) was applied, retaining only practices that directly support GL. Several items were excluded because they primarily managed IL or EL. For example, Instruction, Interactive learning, and Study tools in the Device functionalities dimension emphasized content delivery and navigation; Documents and databases and Evaluation and progress monitoring focused on administration and assessment; Traditional tutor and Digital tutor offered guidance that mainly reduced IL/EL; and Physical context influenced task complexity linked to IL (Sweller et al., 1998). With these eight practices removed, the checklist encompassed 19 items. While managing IL and EL can indeed foster germane processing indirectly by freeing cognitive resources (Paas & van Merriënboer, 2020; Sweller, 2005), these practices were excluded so that the analysis only focused on items with a direct intent to support germane processing.

Guided by a CLT perspective that germane load reflects learners’ investment of cognitive resources in active learning, design features were coded as GL-aligned when they were explicitly intended to elicit learning-relevant cognitive processes (e.g., attention regulation, effort regulation, or information integration), rather than merely to reduce task demands. The analysis was built directly on the set of retained mobile practices identified through in vivo coding in RQ1. For RQ2, additional in vivo coding was applied to each study’s stated learning objectives and pedagogical intentions. These qualitative codes were subsequently refined through focused coding and were then deductively mapped onto four predefined core cognitive purposes of germane processing in technology-enhanced environments (Table 3): schema construction, schema automation, deep learning, and transfer of learning (de Jong, 2010; Howland et al., 2011; Sweller et al., 1998).

OPEN IN VIEWER

Table 3.

Cognitive Purposes and Their Definitions

Cognitive purpose	Definition
Schema construction	The process of building knowledge in long-term memory by combining simpler elements into more complex schemas.
Schema automation	The process of making knowledge or skills, through repeated practice, automatic and be performed with minimal working memory use.
Deep learning	Learning that is active, constructive, intentional, authentic, and cooperative.
Transfer of learning	The application of knowledge or skills acquired in one context to a new, but related, context or situation.

After coding the cognitive purposes associated with the mobile practices in each study, the frequency distribution was summarized across the full dataset. Frequencies for each cognitive purpose were then aggregated across the 53 studies. These aggregated frequencies were used as weighting indicators to reflect the instructional design emphasis placed on different cognitive purposes of germane processing across mobile practices.

After identifying the design emphasis, implementation challenges were then thematically coded using the same Grant’s framework. The Device functionalities dimension was excluded to separate technology-related issues from the analysis, which centered on instructional design features directly addressing germane processing. For the coding, subthemes and codes were drawn from previous reviews of mobile learning (Al-Amri et al., 2020; Benali & Ally, 2020; Faudzi et al., 2023; Gumbheer et al., 2022), e-learning (Regmi & Jones, 2020), and VR/AR-supported learning (Jiang et al., 2025). During deductive coding, all Learner Empowerment subthemes were excluded because reported issues (e.g., technological constraints) corresponded to EL. The coding scheme (Table 4) remained open to additional inductive codes.

OPEN IN VIEWER

Table 4.

Coding Scheme Derived From Previous Reviews

Theme	Subtheme:- Code	Description	Example
Mobile content with pedagogical impact	Pedagogical limitations:	​	​
	- surface-level content	Content lacks pedagogical depth or adaptation to learners’ needs.	Lessons only require learners to read texts or recall facts.
	- insufficient scaffolding	Learners receive limited guidance or support during activities.	Learners complete tasks without clear steps or support.
	- lack of adaptive instruction	Instruction does not adjust to learners’ needs or abilities.	Learners with lower ability struggle to keep up because instruction is not adapted to their pace.
Tutor accessibility and interaction	Feedback and accessibility:	​	​
	- delayed/insufficient feedback	Feedback is missing or delayed, preventing timely correction and awareness of learning gaps.	Learners complete quizzes but receive no feedback on mistakes or ways to improve.
Learner engagement	Engagement:	​	​
	- low motivation	Learners show low interest to engage, often influenced by attitudes or preferences.	Female learners express lower interest in using VR tools.
	- passive participation	Learners join activities with minimal effort or reflection.	Learners guess answers without reviewing content or discussing thoroughly with peers.
	Sustained engagement:	​	​
	- tech-related discomfort	Learners experience physical or digital fatigue when using devices.	Learners experience eye strain after use of VR.

Mobile Technologies’ Complementary Role and Learning Achievement (RQ3)

Mobile technologies can serve either a primary instructional role, that is, facilitating access to information, enabling knowledge construction, and offering interactive reinforcement (Domingo & Garganté, 2016), or a complementary role, where they enhance rather than replace instructional strategies (Thornton & Houser, 2002). More recently, they have been distinguished from simple content delivery by their capacity to provide adaptive prompts or scaffolds aligned with learners’ cognitive states (Breitwieser et al., 2024).

To determine the technology role, each study was screened independently twice by the first author for the abstract, research objectives, research questions/hypotheses, and instructional and tool design. Each round involved judging the role as either Role 1 (primary) or Role 2 (complementary). After both rounds, role labels were compared. If they matched, the label was assigned. If not, the study was re-reviewed for final judgement. Intra-rater reliability (Cohen’s κ = 0.73) was within the acceptable range (McHugh, 2012).

To examine achievement in studies where technologies complemented instructional strategies, reported scores for knowledge and skill gains were documented. Learning achievement refers to content-based performance (e.g., knowledge acquisition, retention) (Abuhassna et al., 2020; Ifenthaler & Yau, 2020), which is distinct from affective, behavioral or metacognitive outcomes. Results were classified as higher, equal, or lower compared to control group(s).

The Most Adopted Practices

Twenty-seven practices across six dimensions of mobile learning were identified (Figure 2). Percentages indicate occurrence rates across all instructional designs (n = 69) rather than the number of studies. Practices are grouped into six dimensions adapted from Grant’s (2019) mobile learning framework, shown along the horizontal axis. Orange bars indicate practices identified as most employed (occurrence above 50% across all instructional designs in 53 studies). The list of all mobile practices with definitions is provided in Appendix B.OPEN IN VIEWER

This frequency distribution shows that mobile learning designs most commonly emphasize learner empowerment (e.g., Autonomy), device-enabled interactivity (e.g., Interactive learning), and learner engagement (e.g., Feedback integration). Highly prevalent practices such as Instruction (96%), Interactive learning (93%), Scaffolded learning (61%), and Practical scenarios (62%) indicate that mobile technologies are typically integrated within structured instructional sequences rather than functioning as stand-alone delivery tools. Similarly, the frequent inclusion of Autonomy (68%) and Critical thinking (55%) reflects that many designs incorporate features associated with increased learner agency and cognitively demanding activities. Taken together, this distribution highlights the coexistence of practices associated with both instruction and learner-driven complexity, which provides the basis for examining how these commonly adopted practices co-occur under higher versus equal/lower intrinsic and extraneous load conditions in the subsequent analysis.

Practices’ Contribution to IL and EL

The left and right panels (Figure 3) present two conditions based on whether experimental groups reported higher levels of IL and EL than control groups, respectively. Node size represents the frequency with which a mobile practice appears within each condition, and line thickness represents the strength of co-occurrence between practices. Accordingly, the networks highlight practice bundles that tend to appear together when cognitive demand is higher (orange-coded) versus when it is maintained or reduced (blue-coded).OPEN IN VIEWER

Across the IL panel, equal or lower IL conditions are more prevalent, shown in the blue-coded networks, suggesting that managing inherent complexity is common. This leaves resources for EL-related challenges, as reflected in the dominant orange-coded EL patterns.

In the high IL region, Multimodal learning was the largest node, distinctly linked to Autonomy. Autonomy also paired strongly with Critical thinking, indicating that designs that increase learner control while simultaneously expanding representational input and higher-order processing demands tend to be associated with greater inherent task complexity. In contrast, the co-occurrence of Practical scenarios and Scaffolded learning is most connected to equal/low IL, followed by strong links between Practical scenarios with Multimodal learning and with Interactive learning. This pattern shows that when contextualized tasks are paired not only with structured guidance but also with multimodal representations and interactive elements, intrinsic complexity is more often maintained at equal or lower levels. The contrast between the two networks in the IL panel suggests that learner autonomy is not lightweight by default, and that it becomes cognitively demanding when paired with multimodal inputs and cognitively complex tasks, whereas scaffolding plus contextualization is more often associated with manageable intrinsic complexity.

The network of higher EL is characterized by Physical context and Interactive learning as the largest nodes, which share the strongest co-occurrence. The second strongest link is between Physical context with Multimodal learning and with Autonomy. These patterns indicate that designs combining real-world context cues with interface actions and multiple representations are frequently linked with increased processing demands. Conversely, the reduced EL side places Interactive learning at the center when it co-occurs with Learning resources and Scaffolded learning. This suggests that interactivity reduces unnecessary burden when it is accompanied by clear guidance and well-organized support. To put it differently, interactivity is a “double-edged” design feature: it is associated with higher EL when combined with contextual integration, but it is associated with managed EL when embedded within a scaffolded and resource-rich design.

GL in Cognitive Load Measurement (Context Setting)

Thirty-four studies excluded GL from cognitive load measurement, making up 64% of the included studies (Figure 4).OPEN IN VIEWER

Although nearly two thirds of all studies excluded GL from the measurements, all the studies incorporated instructional design features aligned with germane processing to varying extents.

Cognitive Purposes

The Sankey diagram (Figure 5) shows how individual practices are distributed across four cognitive purposes. The link width represents the aggregated frequency to which each practice was connected with a given cognitive purpose across the included studies, thereby indicating relative design emphasis.OPEN IN VIEWER

Among the four purposes, Deep Learning received the greatest attention in GL-aligned design. This cognitive purpose linked fairly evenly to multiple practices such as Metacognition, Multimodal learning, Critical thinking, and Feedback integration. This indicates that this purpose was supported through a diverse set of design strategies rather than concentrated on a single approach. Schema Construction ranked second and was strongly supported through practices including Multimodal learning, Learning resources, Collaboration, and Embodied interaction, reflecting an emphasis on knowledge integration through multiple representations and social engagement. Schema Automation came third with Self-regulation, Feedback and Autonomy as the main contributors, indicating an emphasis on regulating and refining learning processes rather than introducing additional learning content. Finally, Transfer of Learning ranked last, most often supported through Practical scenarios and Autonomy, suggesting that fewer studies explicitly targeted the application of learning beyond the immediate instructional context.

Challenges and Limitations in Implementation

As seen in Figure 6, challenges and limitations in implementing GL-aligned features were identified across three main themes. No challenges were coded under Learner Empowerment, reflecting the absence of reported issues related to learner agency in the included studies. While a range of limitations was identified, three challenges emerged most prominently across the subthemes of Pedagogical Limitations and Engagement: restricted affordances, mismatches between design and intended cognitive processes, and excessive external support.OPEN IN VIEWER

Within Mobile content with pedagogical impact, issues most frequently pointed out were restricted affordances and mismatches between design and intended cognitive processes, alongside other codes including insufficient scaffolding. These patterns indicate that challenges at this level were primarily related to how content features constrained or misaligned with intended cognitive processes. Under Tutor accessibility and interaction, challenges were delayed/insufficient feedback and the newly identified discouraging response mechanism. This points to limitations in how instructional support was provided during learning activities. For Learner engagement, the most prominent issues were excessive external support, alongside low motivation and passive participation. This suggests that obstacles to cognitive engagement cannot be attributed solely to the structure of support, but also involve learners’ willingness to engage. Sustained engagement was hindered by tech-related discomfort, although this was not a major concern across the included studies. Overall, the results indicate that limitations in GL-aligned designs were concentrated less on learner-related factors and more on how technological affordances, instructional structures, and support mechanisms were designed in relation to intended cognitive processes.

Roles of Mobile Technologies

Studies with mobile technologies taking the supporting role outnumbered those having digital tools as the primary approach. The distribution is 58% (31 studies) and 42% (22 studies), respectively (Figure 7). This reflects a dominant pattern in mobile learning research toward integrating mobile technologies as supportive elements within instructional designs, rather than positioning them as primary instructional approaches.OPEN IN VIEWER

Positive Achievement

Of the 31 studies where technologies supported instructional design, 26 were analyzed; five were excluded due to lacking control groups or inconsistent subscale results. As shown in Figure 8, 22 studies (85%) reported higher scores for learning achievement, by far higher than those with equal or mixed outcomes. This concentration of positive outcomes highlights a clear pattern in which studies employing mobile technologies in a complementary role most often reported improved learning achievement.OPEN IN VIEWER

The Most Adopted Practices

Across the six dimensions, Learner empowerment, Device functionalities, and Learner engagement emerged as the most emphasized areas. Within the first two, practices such as Autonomy, Multimodal learning, Scaffolded learning, and Interactive learning were most frequently employed. These findings indicate a clear focus on learner agency and interaction, consistent with the shift from tool-centric views to contextual learning interactions (Crompton, 2013; Thornton & Houser, 2002). Although mobile learning has been implemented across educational levels, the literature indicates a predominant focus on higher education contexts, where greater learner self-direction is typically expected; consequently, mobile learning practices increasingly emphasize interaction and pedagogical support rather than mere technological functionality (Fang, 2025; Garzón et al., 2025). The Learner engagement dimension further reinforces this design orientation with frequent use of Practical scenarios, Critical thinking, and Feedback integration for active participation and sustained involvement.

Within Mobile content with pedagogical impact, only Learning resources stood out, showing how mobile affordances are utilized through quizzes and tasks. Flexibility is a core principle of mobile learning, given that personal and situational contexts are essential for effective learning (Palalas & Wark, 2020). However, Contextual adaptation and Learner-dependent adaptability were not often employed.

No practices were highly adopted under Tutor accessibility and interaction, whereas in Contextual influence, only Physical context emerged. This stands in contrast to a recent study by Ariza & Hernández (2025), which positioned teacher presence and instructional guidance as central components of mobile learning design. Taken together, these patterns reflect a reduced reliance on direct instructor guidance while confirming the certain influence of the physical setting in mobile learning, where learners process information from both digital and physical sources.

Practices’ Contribution to IL and EL

The ENA showed Autonomy and Multimodal learning appeared together in high IL contexts. Without sufficient guidance, high autonomy can overwhelm inexperienced learners due to material complexity (Castro-Alonso et al., 2021), as seen in empirical VR studies where low prior-knowledge students experienced higher load (Liu et al., 2024). Integrating multiple multimedia sources also increases processing demands (Sweller, 1994), and tasks involving critical thinking are inherently complex, making guidance essential when combined with high autonomy (Curum & Khedo, 2021; Skulmowski & Xu, 2021). These links reflect the challenge of balancing self-navigation with instructional support.

In the equal/low IL zone, Practical scenarios co-occurred with Scaffolded learning and linked to Multimodal learning and Interactive learning. Real-world tasks boost relevance and motivation but can be inherently demanding. However, scaffolding helps learners handle task complexity (Könings et al., 2019), which allows IL to be managed rather than being overwhelming. When combined with interactive elements, scaffolding can lessen cognitive demands, as found in studies by Kao et al. (2017) and Sysoev et al. (2022). In both digital and mobile learning contexts, prior studies have largely conceptualized scaffolding as real-time, task-embedded support accessed during task implementation to manage learners’ cognitive load (Gong et al., 2026; Karabay & Mese, 2025). Accordingly, the present findings suggest that scaffolding should also be considered at the instructional design stage as part of IL management, particularly for structuring task complexity.

Working memory is easily burdened when learners process both real and digital sources. As physical context itself can impose EL, combining it with interactive elements, multimodal formats, or high autonomy would further increase processing demands, in line with findings from AR/VR studies where multimedia manipulations increased cognitive load (Tugtekin & Odabasi, 2022). This reflects the split-attention effect, which occurs when learners integrate different sources, and the modality effect, where multiple presentation modes overwhelm learners if poorly designed (Chandler & Sweller, 1991; Sweller, 1994). These effects likely explain strong links between Physical context and Interactive learning, Multimodal learning, and Autonomy in the high EL region.

Under equal/low EL conditions, Interactive learning was central, closely connected with Learning resources and Scaffolded learning. Learning resources here include quizzes, tasks, and games embedded in mobile systems. When combined with scaffolded support, these activities reduce unnecessary load, consistent with studies of scaffolded game- and task-based learning (Chang & Yang, 2023; Faber et al., 2024).

The ENAs indicate that for both high and low levels of IL and EL, mobile learning design should not be understood in terms of single strategies, but rather as combinations of co-occurring instructional practices. By contrast, prior mobile learning studies have predominantly examined cognitive load through either isolated design attributes or interaction complexity (Toh & Tasir, 2024). When interaction complexity was considered, these prior studies did not unpack the specific instructional practices that contribute to such complexity, or examine how these practices are combined within instructional designs (Toh & Tasir, 2024). The ENA findings suggest that while IL is primarily addressed through the structuring and sequencing of task complexity, EL in mobile learning emerges from how multiple instructional features are integrated. This highlights the importance of design coherence across physical context, interactivity, and instructional support.

Cognitive Purposes

The strong emphasis on Deep Learning suggests the importance of cognitive engagement in mobile learning designs. Learning activities now frequently promote higher-order thinking (Crompton et al., 2018). In digital environments, deep and meaningful learning arises when activities are active, constructive, intentional, authentic, and cooperative (Howland et al., 2011). This explains why Deep Learning is consistently targeted through practices such as Self-regulation, Critical thinking, Feedback integration, Practical scenarios, and Collaboration.

Schema Construction ranked second, reflecting its central role in CLT (Sweller et al., 1998). It is supported by designs featuring personalized learning (Wang et al., 2024), multimedia enhancement (Hwang, 2024), and related strategies. This aligns with practices tied to knowledge building, including Multimodal learning, Learning resources, Scaffolded learning, and Embodied interaction.

Recent instructional design models recommend prioritizing schema activation and construction before other processes (Jung et al., 2022), explaining the third and fourth positions of Schema Automation and Transfer of Learning. Authentic learning contexts and repeated practice are key strategies for automation and transfer (Merriënboer & Paas, 1990; Sheldon et al., 2023). High autonomy, associated with greater motivation and involvement (Chen, 2022), further supports sustained engagement. This explains why Autonomy, alongside Practical scenarios and Contextual adaptation, typically supports these purposes.

To the best of our knowledge, most empirical studies do not explicitly link their designs to germane processing purposes; however, this intention can be interpreted from empirical research on higher-order thinking, conceptual understanding, practice effects, and transfer in mobile learning contexts (Chang & Hwang, 2018; Chin et al., 2019; Huang et al., 2024; Yang et al., 2024).

The overall findings indicate that mobile instructional designs prioritize germane processing for deeper learning and schema construction, while automation and transfer are more often supported as longer-term outcomes.

Challenges and Limitations in Implementation

Three prominent codes which emerged across the subthemes of Pedagogical limitations and Engagement were restricted affordances, mismatches between design and intended cognitive processes, and excessive external support.

Under the Pedagogical Limitations subtheme, Restricted affordances were most frequently reported as obstacles to germane processing, which suggests that constrained features across platforms, modalities, and interaction types limited opportunities for deeper cognitive engagement. Usability issues such as responsiveness and interface design undermined benefits (Rad, 2025). For example, Albus and Seufert (2023) found that when learners could not revisit auditory information, GL was lower than in text-based conditions that allowed rereading, while Poupard et al. (2025) noted that VR limited to simple manipulations was insufficient for deeper learning.

Mismatches between design and intended cognitive processes were the second major obstacle within the same subtheme. This issue reflects Børte and Lillejord’s (2024) view that teaching strategies and technology must be aligned from the outset, which is coupled with findings that learning outcomes depend on appropriate teaching techniques (Omeh et al., 2024) rather than on digital tools alone. For instance, Thees et al. (2020) found that although AR visualizations were provided in a physics laboratory, knowledge gains required deeper data analysis. Yu et al. (2023) reported that agency and embodiment must fit the learning topic, and rich embodiment alone did not ensure learning.

Excessive external support was the third challenge, and fell under the Engagement subtheme. Chen et al. (2022) found that high levels of sound, light, and 3D immersion in VR hindered higher-order skills such as creative thinking. Wu et al. (2021) also reported that goal setting was driven by the SV-IVR system rather than by learners themselves, which reduced opportunities for self-regulation. From a Vygotskian perspective, support beyond the “just enough” level for independence may undermine instructional effectiveness (Vygotsky, 1978).

Remaining limitations under the Pedagogical Limitations subtheme were surface-level content (e.g., Albus & Seufert, 2023), insufficient scaffolding (e.g., Albus et al., 2021), lack of adaptive instruction (e.g., Yang, 2017), and limits of full digital substitutes, pointed out in only one study; specifically, Liu et al. (2021) found that an AR-based experiment tool lacked the learning experience with real artifacts to develop deep understanding. These limitations echo findings that digital adoption is most effective when adapting to learning needs and the subject area, and when integrated with appropriate strategies (Alshammary & Alhalafawy, 2023).

Regarding Feedback and Accessibility, the only identified challenge was delayed/insufficient feedback (e.g., Hwang et al., 2020), which matters since structured feedback combined with teaching strategies can promote higher reflection (Karakas & Yükselir, 2025), and reflection supports deeper engagement (Huang et al., 2024). Under Communication Quality, only discouraging response mechanisms emerged. For instance, Lin and Lin (2016) found that the system failed to capture students’ needs. Hwang et al. (2020) noted that requiring only correct answers to progress was discouraging, while Cai et al. (2022) found that their mechanism of replies emotionally constrained learners. Together, these findings suggest that poor communication design can limit responsiveness and learner motivation.

Regarding the Engagement subtheme, beyond excessive external support, minor challenges were low motivation and passive participation. Under Sustained engagement, a minor issue of tech-related discomfort was identified. Particularly, Liu et al. (2022) reported some students with motion sickness from using IVR.

These findings extend existing research by showing that barriers to germane processing in mobile learning arise from design decisions that span three interconnected layers of technology, pedagogy, and learner engagement. This points to balance as the key to successful implementation of GL-aligned strategies. First, balance is needed between mobile technologies and their affordances. Technological affordances can both enable and constrain actions, functioning as information opportunities (Hirvonen et al., 2024), and restricted affordances may therefore significantly influence information processing at a deeper level. Second, there should be balance between affordances and intended cognitive processes. Digital tools misaligning with pedagogical aims may hinder learning. Even when technological environments provide foundational support, additional design features are often needed to ensure engagement in the intended cognitive processes (Pan et al., 2025). Third, balance is needed between external support and autonomy; excessive support may hamper learners’ autonomous cognitive engagement (Dever et al., 2024; Giacumo & Savenye, 2020).