A Culturally Responsive Digital Differentiated Learning framework is associated with greater engagement and narrower learning disparities in higher education: a quasi-experimental mixed-methods study

1 Introduction

Student engagement and equity in higher education have become central concerns in increasingly diverse and digitally mediated learning environments (Gulko et al., 2024). Recent research continues to emphasize that engagement is shaped not only by students' individual characteristics but also by the extent to which instructional approaches respond to diverse learner needs and provide meaningful opportunities for participation. Differentiated learning has increasingly been recognized as an approach for tailoring instructional content, processes, and learning activities to variation in learner readiness and needs (Goyibova et al., 2025). At the same time, technology-supported approaches, including interactive and gamified digital environments, provide additional opportunities to increase participation and create more flexible learning experiences (Fatima et al., 2025). Nevertheless, persistent gaps in engagement and academic success remain between students from different socioeconomic, cultural, and linguistic backgrounds, even when they study in the same institutions and programmes (Crabtree, 2021; Gulko et al., 2024). During and after the COVID-19 pandemic, large student surveys documented that lower-income students were more likely to report low engagement and a stronger sense of “falling behind” than their better-resourced peers, illustrating how digital learning arrangements can amplify pre-existing inequities (Hope, 2021). Similar concerns have been reported across higher education contexts, suggesting that increasing student diversity has not always been accompanied by sufficiently responsive pedagogical and student-support systems (Crabtree, 2021; Gulko et al., 2024).

The rapid growth of online, blended, and other technology-enhanced forms of teaching has added another layer to this picture (Hope, 2021; Rehman, 2023). Digital environments, including learning management systems (LMS), open up possibilities for flexible access, varied pacing, and more personalized learning experiences, but they also make visible, and in some cases intensify, disparities in access to devices and connectivity, differences in digital literacy, and uneven availability of culturally relevant academic and pastoral support (Hope, 2021; Rehman, 2023; Shadiev et al., 2023). Policy reports and practice guides increasingly urge universities to move beyond a narrow “upload the lecture” mindset and to design online learning with explicit attention to students' mental well-being, diverse needs, and fair chances to succeed (Al-Zahrani, 2025; Barry et al., 2023; Hope, 2021). Across this literature there is a growing agreement that simply transferring existing courses into digital formats is not enough; higher education needs pedagogical frameworks that deliberately combine differentiated instruction, cultural responsiveness, and thoughtful digital design to tackle engagement and learning disparities in a more systematic and sustainable way (Barry et al., 2023; Cochrane et al., 2017; Mills, 2025).

Against this background, this study develops and empirically evaluates a Culturally Responsive Digital Differentiated Learning (CR-DDL) Framework for LMS-based higher education. The framework integrates differentiated learning, culturally responsive teaching, and purposeful digital learning design to provide flexible learning pathways while incorporating students' cultural, community, and lived experiences into learning activities. The study makes three contributions. First, it provides an empirically specified framework that translates these complementary principles into a structured LMS-based learning design. Second, it evaluates the framework using multiple indicators, including self-reported student engagement, LMS-derived behavioural engagement, and academic achievement. Third, it examines whether the framework contributes to reducing engagement disparities across socioeconomic and cultural groups and uses qualitative interviews to explain students' experiences of differentiated learning, cultural relevance, inclusion, and perceived fairness. The remainder of the paper first reviews the theoretical foundations and research gap underlying the framework, followed by the Methods section describing the mixed-methods design, participants, framework development and implementation, instruments, and analytical procedures. The Results section then presents quantitative, qualitative, and integrated findings, followed by a Discussion of their theoretical and practical implications, limitations, and directions for future research.

2 Literature review

2.1 Differentiated and technology-enhanced learning

Differentiated learning aims to adapt content, process, and learning products to students' readiness, interests, and learning profiles, with the goal of maximizing each learner's potential (Azis and Maharani, 2024; Harmilawati et al., 2024). In technology-enhanced learning environments, digital tools can support differentiated instruction through adaptive pathways, multimodal resources, and personalized feedback, allowing instructors to respond more flexibly to individual differences at scale (Rehman, 2023). Recent studies note that data-driven platforms, learning analytics, and virtual environments can tailor instruction to varied linguistic proficiencies, cognitive styles, and learning needs, while also enabling more precise monitoring of engagement and performance over time (Al-Zahrani, 2025; Gulko et al., 2024; Rehman, 2023). Taken together, these affordances suggest that LMS-based differentiation can improve both engagement and academic outcomes, particularly when anchored in deliberate course design and continuous, data-informed adjustment of tasks and supports (Al-Zahrani, 2025; Keith et al., 2023).

In practice, however, many digital differentiation efforts still concentrate on cognitive and performance indicators, such as prior knowledge, pacing, or assessment scores, while giving far less systematic attention to students' cultural, linguistic, and community-based resources (Azis and Maharani, 2024; Cochrane et al., 2017; Harmilawati et al., 2024). Studies of differentiated learning in school and higher education contexts frequently report gains in motivation and achievement, but culture is often treated as a static background variable rather than a core design dimension that actively shapes materials, tasks, and interactions (Azis and Maharani, 2024; Harmilawati et al., 2024). In such cases, digitally differentiated environments may privilege dominant cultural norms and preferred ways of learning, thereby risking the reproduction of existing engagement and outcome gaps rather than closing them (Crabtree, 2021; Ebede, 2018; Gulko et al., 2024).

2.2 Culturally responsive teaching in higher and digital education

Culturally responsive teaching (CRT) positions students' cultural, linguistic, and experiential backgrounds as assets and essential starting points for learning (Cochrane et al., 2017; Nurbatra and Masyhud, 2022). In higher education, CRT involves maintaining high academic expectations, cultivating cultural competence, and embedding students' cultural knowledge and lived experiences within curriculum, pedagogy, and assessment (Cochrane et al., 2017; Johnson, 2022; Keith et al., 2023). Emerging work in this area shows that culturally responsive approaches can strengthen students' sense of belonging, engagement, and academic confidence, particularly among historically marginalized or culturally diverse learners (Johnson, 2022; Keith et al., 2023; Nurbatra and Masyhud, 2022).

The shift toward online and hybrid learning has prompted growing interest in culturally responsive pedagogy in digital spaces (Keith et al., 2023; Mills, 2025). Recent studies and conceptual frameworks describe how culturally responsive principles can be translated into online course design, assessment, and community-building by emphasizing inclusive communication norms, multimodal representation of content, and opportunities for students to bring their languages, identities, and cultural practices into digital classrooms (Kumi-Yeboah and Samuel, 2023; Mills, 2025). Equity-focused guides for digital learning similarly argue that culturally responsive practice should be embedded into the design of digital activities, assessment strategies, and interaction patterns, treating students' cultural knowledge as a resource for meaning-making rather than a barrier to academic success (Keith et al., 2023; Kumi-Yeboah and Samuel, 2023).

At the same time, research on culturally responsive teaching in digital settings highlights persistent barriers, including unequal access to technology, differing levels of digital literacy, and institutional practices that do not adequately support culturally responsive design (Al-Zahrani, 2025; Gulko et al., 2024; Hope, 2021). Qualitative and conceptual work in school and higher education contexts suggests that culturally responsive digital pedagogy often depends on individual instructors' efforts rather than coherent institutional frameworks, which leads to substantial variation in students' learning experiences and outcomes (Boison and Burke, 2025; Keith et al., 2023; Mills, 2025).

2.3 Intersection of digital differentiation, CRT, and engagement

Several strands of scholarship suggest that differentiated instruction, technology-enhanced learning, and culturally responsive pedagogy can work together rather than in isolation (Azis and Maharani, 2024; Cochrane et al., 2017; Harmilawati et al., 2024; Rehman, 2023). In school, language, and higher education contexts, researchers show that when instructors embed culturally responsive strategies into differentiated tasks, learning activities become more connected to students’ cultural and linguistic worlds, which can increase interest, engagement, and academic outcomes (Azis and Maharani, 2024; Harmilawati et al., 2024; Nurbatra and Masyhud, 2022). Work on technology-enhanced learning in multicultural settings adds another important dimension: digital tools can support differentiated and culturally responsive approaches by making it easier to offer multiple representations, languages, and modes of interaction within the same learning environment (Rehman, 2023; Shadiev et al., 2023). Equity-minded digital strategy guides further emphasize practical approaches such as activating culturally grounded prior knowledge, drawing explicitly on students' community-based experiences, and using cultural referents in online tasks to humanize digital learning and keep equity at the center (Keith et al., 2023; Kumi-Yeboah and Samuel, 2023; Mills, 2025).

At the same time, the literature indicates that digital differentiation, CRT, and student engagement are still frequently treated as separate conversations rather than as an integrated pedagogical agenda (Cochrane et al., 2017; Johnson, 2022; Rehman, 2023). Studies on engagement in higher education underline its multidimensional nature, including behavioral, emotional, cognitive, and social and show that engagement varies substantially across student groups, reinforcing the need for inclusive institutional strategies rather than isolated interactive activities (Crabtree, 2021; Ebede, 2018; Gulko et al., 2024). Research on online engagement and equity likewise stresses intentional design, but much of it still concentrates on course structure or generic use of technology rather than on tightly integrating cultural responsiveness into differentiated digital practices (Hope, 2021; Keith et al., 2023; Mills, 2025). Where CRT and digital learning are brought together explicitly, the work is often conceptual, small-scale, or focused on specific contexts, with relatively limited empirical testing of structured frameworks that simultaneously seek to raise engagement and reduce documented learning gaps (Boison and Burke, 2025; Kumi-Yeboah and Samuel, 2023; Mills, 2025).

2.4 Research gap: need for an integrated CR-DDL framework

A growing body of work shows a consistent pattern: engagement and achievement gaps in higher education remain persistent and are closely tied to structural inequities rather than simply to individual effort (Crabtree, 2021; Ebede, 2018; Gulko et al., 2024). The same literature also makes clear that digital learning is a double-edged sword; under some conditions, it can worsen disparities, while under more inclusive and well-designed conditions, it may help reduce them (Al-Zahrani, 2025; Gulko et al., 2024; Hope, 2021). At the same time, differentiated instruction and culturally responsive teaching continue to appear as promising approaches for improving engagement and outcomes among diverse learners (Boison and Burke, 2025; Harmilawati et al., 2024; Keith et al., 2023). However, few studies have integrated these approaches into a single framework for higher education. In particular, limited research has examined how LMS features can support both culturally responsive and differentiated learning. Evidence is also limited on whether such an integrated approach can improve engagement while reducing disparities between student groups (Keith et al., 2023; Mills, 2025; Rehman, 2023).

First, differentiated learning with digital tools is not new; there is already a sizeable literature on technology-enhanced and adaptive learning. However, most models still treat cultural responsiveness as an additional consideration rather than as the core engine of instructional design (Azis and Maharani, 2024; Cochrane et al., 2017; Harmilawati et al., 2024). Digital differentiation is commonly driven by test scores, prior performance, or preferred learning styles, while students' cultural and community-based knowledge often remains peripheral rather than being used as a core design resource (Harmilawati et al., 2024; Rehman, 2023). Approaches designed in this way may be less capable of addressing engagement and achievement gaps that arise from structural and cultural inequities, not merely from variation in academic readiness (Crabtree, 2021; Ebede, 2018; Gulko et al., 2024).

Second, the CRT literature has advanced considerably, particularly in higher and online education, but it often stops short of large-scale LMS-based empirical work capable of measuring its impact on learning disparities (Cochrane et al., 2017; Johnson, 2022; Keith et al., 2023). There are rich descriptions of effective practices, strong conceptual models, and smaller studies focusing on student perceptions and inclusive course design (Keith et al., 2023; Mills, 2025). What remains limited is research that uses robust engagement and achievement indicators to examine how culturally responsive design in LMS-supported courses changes the size of outcome gaps between student subgroups (Boison and Burke, 2025; Gulko et al., 2024; Hope, 2021).

Third, where CRT and digital learning intersect, the emphasis often falls on equity-oriented course design or family–school partnerships, frequently in school rather than university contexts (Boison and Burke, 2025). These contributions are valuable, but they do not easily translate into a scalable higher education model that brings together LMS functions, differentiated tasks, culturally responsive pedagogy, and measurable equity outcomes in one coherent framework (Keith et al., 2023). As a result, institutions and instructors committed to equity still face difficulty translating this commitment into concrete, data-informed practices across courses and programs (Crabtree, 2021; Gulko et al., 2024; Rehman, 2023).

This study responds to this gap by developing and empirically evaluating a Culturally Responsive Digital Differentiated Learning (CR-DDL) Framework for LMS-based higher education. The framework brings together three complementary principles: differentiated learning, culturally responsive teaching, and the purposeful use of digital learning affordances. Differentiated learning provides flexible pathways that accommodate variation in students' readiness and learning needs; culturally responsive teaching positions students’ cultural and community experiences as resources for learning; and LMS-based digital design provides the infrastructure through which these principles can be implemented, monitored, and adapted. The framework therefore goes beyond the provision of technology or isolated instructional choices by linking pedagogical differentiation and cultural responsiveness to an explicit equity objective: improving engagement while reducing disparities between student subgroups.

The study makes three contributions to the literature. First, it provides an empirically specified framework that integrates differentiated learning and culturally responsive teaching within an LMS-based higher education environment. Second, it evaluates the framework using complementary indicators of student engagement, including self-reported engagement and LMS-derived behavioural indicators, together with academic achievement. Third, it examines whether the intervention is associated with changes in engagement disparities, particularly according to socioeconomic and cultural background, while using qualitative interviews to explain how students experienced differentiated tasks, culturally relevant learning materials, and assessment practices. The study therefore addresses both the effectiveness of the framework and the mechanisms through which students perceived it as supporting engagement and inclusion.

The remainder of this paper is structured as follows. The next section reviews the theoretical and empirical foundations of differentiated learning, culturally responsive teaching, and their intersection in digital higher education and presents the rationale for the CR-DDL Framework. The Methods section then describes the explanatory sequential mixed-methods design, participants, implementation of the framework, quantitative analyses, and qualitative thematic analysis. The Results section presents quantitative outcomes, subgroup disparities, qualitative themes, and their integration. The Discussion interprets the findings in relation to previous research, considers implications for higher education practice, and outlines the study limitations and directions for future research. Through this structure, the study evaluates CR-DDL not only by asking whether engagement and learning outcomes improve, but also by examining whether the framework contributes to more equitable engagement across student groups.

2.5 Development and theoretical foundation of the CR-DDL framework

The Culturally Responsive Digital Differentiated Learning (CR-DDL) Framework was developed by the authors specifically for this study and was not adopted as a pre-existing intervention package. The framework represents an integrative pedagogical model designed to address student engagement and engagement disparities in LMS-based higher education. Its development was theoretically informed by three complementary strands of literature: differentiated learning, culturally responsive teaching (CRT), and technology-enhanced digital learning. Differentiated learning informed the provision of alternative learning tasks, levels of challenge, and learning pathways according to students' readiness and learning needs. CRT informed the incorporation of students' cultural knowledge, community experiences, and locally relevant contexts into learning activities and interactions. Digital learning principles informed the use of LMS functions to provide multimodal content, flexible learning pathways, asynchronous interaction, collaborative activities, and continuous monitoring of student participation.

The framework was developed through a conceptual integration and instructional operationalisation process. First, the research team reviewed the pedagogical functions represented in the three theoretical foundations and identified complementary functions relevant to engagement, inclusion, and learning equity. Second, these functions were translated into observable instructional practices that could be implemented within an LMS. Third, the practices were organised into an interconnected instructional sequence that could be consistently implemented across intervention courses. This process resulted in six operational components: (1) culturally responsive introduction, (2) multimodal core content, (3) differentiated learning activities, (4) cultural reflection and connection tasks, (5) collaborative learning, and (6) assessment and feedback. The framework was therefore not treated as a pre-existing validated intervention, but as a study-specific framework developed from established pedagogical principles and operationalised for empirical evaluation in the participating higher education contexts.

Table 1 summarises how the three theoretical foundations were translated into six operational components of the CR-DDL Framework. The components were designed to function as an integrated learning sequence rather than as independent instructional techniques. Culturally responsive introduction and cultural reflection established contextual relevance, multimodal content provided multiple routes to accessing learning materials, differentiated activities provided appropriate levels of challenge and choice, collaborative learning supported social engagement, and flexible assessment and feedback provided opportunities for students to demonstrate and improve their learning. The LMS served as the delivery and monitoring environment through which these components were implemented and documented.

Table 1. Components and operationalisation of the CR-DDL framework.

The first component, culturally responsive introduction, was designed to activate students' prior knowledge through local cases and contextual prompts. The second component, multimodal core content, provided learning materials in multiple formats to accommodate differences in students' access, preferences, and ways of engaging with information. The third component, differentiated learning activities, provided task options at different levels of complexity and allowed students to select an appropriate pathway. The fourth component, cultural reflection and connection tasks, required students to connect theoretical concepts with their cultural, community, or lived experiences. The fifth component, collaborative learning, provided opportunities for peer interaction and shared problem-solving. The sixth component, assessment and feedback, provided flexible assessment formats and timely feedback to support learning progression.

The six components were implemented as an interconnected cycle rather than as isolated activities. Culturally relevant entry points were used to activate prior knowledge, multimodal resources provided access to core content, differentiated tasks allowed students to select an appropriate level of challenge, reflection activities connected academic concepts with students' contexts, collaborative activities supported social engagement, and flexible assessment with feedback provided opportunities for students to demonstrate and improve their learning. LMS analytics were used alongside survey measures and academic outcomes to evaluate whether the implementation was associated with changes in engagement and learning disparities.

The framework consequently informed the study at four levels: learning design, implementation, data collection, and evaluation. At the learning-design level, the framework specified the content, activities, interaction, and assessment features used in intervention courses. At the implementation level, it guided instructor preparation and LMS configuration. At the data-collection level, the framework informed the selection of engagement indicators and interview domains concerning differentiation, cultural relevance, inclusion, and fairness. At the evaluation level, the framework provided the basis for examining changes in overall engagement, LMS participation, academic achievement, and subgroup disparities. This alignment ensured that the framework was evaluated as an integrated pedagogical intervention rather than as a collection of unrelated digital activities.

Figure 1 illustrates the conceptual structure and evaluation pathway of the CR-DDL Framework. The framework integrates culturally responsive teaching, differentiated learning, and digital learning principles into a connected instructional process. Culturally responsive elements establish contextual relevance by connecting learning materials with students' cultural, community, and lived experiences. Differentiated learning elements provide students with choices in task level and learning activities according to their perceived readiness. Digital learning features within the LMS provide the infrastructure for delivering multimodal materials, differentiated pathways, collaborative activities, reflection, assessment, and feedback.

The six components are interconnected rather than implemented as isolated instructional strategies. Cultural relevance provides the contextual foundation for learning, while multimodal content and differentiated activities provide multiple ways for students to access and engage with course materials. Cultural reflection and collaborative learning extend this process by encouraging students to connect academic concepts with their experiences and interact with peers. Assessment and feedback provide opportunities for students to demonstrate learning and receive guidance for further improvement.

The framework also provided the organising structure for the evaluation of the intervention. The implementation of the six components informed the design of the LMS-based learning activities, while the resulting learning experiences were examined through quantitative indicators of engagement and academic performance and qualitative accounts of students' experiences. This integrated evaluation allowed the study to examine both whether the CR-DDL intervention was associated with improved engagement and learning outcomes and how students perceived the mechanisms contributing to those outcomes.

3 Methods

3.1 Study design

This study used an explanatory sequential mixed-methods design consisting of a quasi-experimental quantitative phase followed by a qualitative explanatory phase (Creswell and Creswell, 2018). In the quantitative phase, comparable courses were allocated to either the CR-DDL intervention condition or the standard LMS-based comparison condition, and students were assessed before and after the intervention. Quantitative data comprised self-reported engagement, LMS-derived behavioural indicators, and academic outcomes. The primary quantitative analyses examined changes over time between conditions and changes in engagement disparities across student subgroups.

The qualitative phase followed the preliminary quantitative analysis. Individual semi-structured interviews were conducted with purposively selected students representing different engagement trajectories and subgroup characteristics. The qualitative findings were used to explain and contextualise the quantitative patterns, particularly the observed differences in engagement and the experiences of students from lower socioeconomic and culturally diverse backgrounds. Integration occurred during interpretation through comparison of quantitative results with qualitative findings and through joint displays (Fetters et al., 2013).

Because course clusters, rather than individual students, were assigned to the intervention or comparison condition on the basis of scheduling feasibility and instructor willingness, the quantitative component was considered quasi-experimental rather than individually randomised. This design was selected to evaluate the CR-DDL Framework under routine educational conditions while accounting analytically for clustering by course.

The qualitative component was conducted after completion of the quantitative assessments to provide contextual insight into students' experiences of the CR-DDL Framework. Although the interviews were conducted as planned, the original audio recordings and complete individual transcript files are currently unavailable in the research records accessible to the authors. This limitation is acknowledged explicitly because it affects the extent to which the qualitative analytical process can now be independently audited.

3.2 Setting and participants

The setting was three higher education institutions in Pekanbaru, Indonesia: Universitas Hang Tuah Pekanbaru, Universitas Lancang Kuning, and Institut Kesehatan Payung Negeri. We focused on undergraduate students enrolled in LMS-based courses in public health and agrotechnology, because these programmes already relied heavily on online and blended learning.

3.2.1 Inclusion and exclusion criteria

Students were included if they were officially registered in one of the selected courses during the study semester, used the institutional LMS as their main learning platform for that course, and gave informed consent to participate. Those who were on academic leave, had previously participated in CR-DDL pilot activities, or did not complete both the baseline and follow-up surveys were excluded from the quantitative analyses. For the interview phase, students also needed to be willing to join a 45–60-minute conversation in Bahasa Indonesia or English.

3.2.2 Sample size and sampling technique

The initial quantitative sample-size calculation was conducted in G*Power for a two-group, two-time-point mixed ANOVA (Faul et al., 2009). Assuming a small-to-medium effect size (f = 0.20), α = 0.05, statistical power of 0.80, and a correlation of 0.50 between repeated measurements, the estimated minimum student-level sample was 196 participants (98 per group). To allow for attrition and incomplete data, recruitment targeted at least 240 students.

Importantly, this initial calculation was performed at the individual-student level and did not incorporate a design effect or an assumed intraclass correlation coefficient (ICC) for clustering within courses. Course-level assignment became an important feature of the implemented quasi-experimental design; therefore, the original power calculation may have overestimated the effective amount of independent information if outcomes were correlated within courses. The final sample included 252 students nested within six course clusters. Clustering was subsequently addressed in the primary outcome analyses by including course as a random intercept in the linear mixed-effects models. Given the small number of course clusters, however, estimates of between-course variance and clustering effects should be interpreted cautiously. This limitation is acknowledged when interpreting the precision and generalisability of the intervention estimates.

Courses, rather than individual students, were sampled first. Within each institution, programme coordinators identified two compulsory courses at a similar academic level with established LMS use, resulting in six course clusters. One course within each institutional setting received the CR-DDL redesign and the other served as the comparison condition. Allocation was not random and was determined by scheduling feasibility, instructor availability, and willingness to implement the intervention. Accordingly, the study is described as a quasi-experimental study with course-level assignment rather than as an individually randomised trial.

Because intervention allocation occurred at the course level and partly depended on instructor willingness, instructor-related differences represent a potential source of confounding. All intervention instructors received the same CR-DDL orientation and implementation guidance, but instructor characteristics such as teaching experience, pedagogical style, digital teaching proficiency, and prior familiarity with culturally responsive or differentiated instruction were not measured systematically in a form that permitted formal statistical adjustment. Consequently, the analytical models account for course-level clustering but cannot fully separate the influence of the CR-DDL framework from unmeasured instructor-level variation.

3.3 Instruments

3.3.1 Higher Education Student Engagement Scale (HESES)

Engagement was measured with the Higher Education Student Engagement Scale (HESES) (Kim and Sax, 2019; Kim and Westine, 2019). The original HESES was developed for university students and assesses three domains: behavioural, emotional, and cognitive engagement. It contains 23 items rated on a 5-point Likert scale from 1 (“strongly disagree”) to 5 (“strongly agree”), with higher values indicating stronger engagement (Kim and Westine, 2019). The developers reported a stable three-factor structure and good internal consistency, with Cronbach's alpha coefficients typically in the 0.80–0.90 range for the subscales (Kim and Sax, 2019; Kim and Westine, 2019).

We adapted HESES into Bahasa Indonesia using a forward–backward translation procedure. Two bilingual experts independently translated the items, a third reconciled differences, and a fourth translator produced a back-translation to check for meaning shifts (Kaur et al., 2022). Three senior academics in education and public health then reviewed the items for clarity and cultural relevance, especially around examples of online participation and group work. A pilot with 60 students from similar programmes showed good reliability for the adapted version (α = 0.88 for the total scale; 0.81–0.86 for the three subscales), which is comparable to the original studies (Kaur et al., 2022).

For the main study, subscale scores were computed by averaging item scores within each domain. A total engagement score was also calculated as the mean of all 23 items. As a practical guide for interpretation, mean scores below 2.5 were treated as low engagement, 2.5–3.4 as moderate, and 3.5 or higher as high engagement, consistent with previous work using Likert-type engagement scales (An et al., 2022).

3.3.2 LMS-based engagement indicators

To complement self-report data, we extracted LMS log indicators for each participant over the semester. These included (1) number of logins per week, (2) total time spent on course content pages, (3) number of posts and replies in discussion forums, and (4) completion of CR-DDL activities such as differentiated tasks and reflective prompts. Prior research suggests that when interpreted carefully, such indicators can provide useful proxies for behavioural engagement in online and blended courses (Bond et al., 2020; Shadiev et al., 2023). We standardized each indicator and combined them into a composite LMS engagement index used in the analyses.

3.3.3 Learning outcomes and disparity indicators

Final course performance was obtained from routine course-specific assessments, including quizzes, assignments, and final examinations, and was recorded using the 0–100 score scale available in the participating courses. Because the participating courses represented different academic disciplines and used discipline-specific assessment tasks, the assessments were not based on a common examination and were not formally equated for assessment difficulty across courses. Therefore, final course score was treated as a secondary supportive academic outcome rather than as a perfectly standardised cross-course achievement measure. Analyses of final course score adjusted for prior-semester GPA and incorporated a random intercept for course to reduce, although not eliminate, the influence of pre-existing academic differences and course-level heterogeneity (Gulko et al., 2024).

Socioeconomic status (SES) was assessed at baseline using information collected in the study questionnaire concerning parental education and students' perceived financial sufficiency. The study dataset classified participants into three ordinal SES categories: lower, middle, and higher SES. The distribution was 85 students (33.7%) in the lower-SES category, 138 (54.8%) in the middle-SES category, and 29 (11.5%) in the higher-SES category.

For the primary disparity analysis, the middle- and higher-SES categories were combined and contrasted with the lower-SES group. This analytical grouping was chosen because the higher-SES category was relatively small and separate estimation of multiple interaction terms across only six course clusters would have produced unstable subgroup estimates. Accordingly, the lower versus middle/higher SES classification should be interpreted as a pragmatic analytical grouping rather than as a validated socioeconomic index.

3.3.4 Qualitative interview guide

The interview guide was developed jointly by the research team to explore students' experiences with the CR-DDL Framework. Questions covered perceived relevance of tasks, cultural resonance of examples and case studies, feelings of inclusion and fairness in group activities, and perceived changes in motivation and confidence over the semester. Additional prompts invited students to describe specific LMS features or differentiated options they found helpful or challenging (Johnson, 2022; Nurbatra and Masyhud, 2022). The guide was piloted with three students; minor wording changes were made for clarity before the main round of interviews.

3.4 Procedure

3.4.1 Ethical approval and consent

The study protocol was reviewed and approved by the Research Ethics Committee of Universitas Hang Tuah Pekanbaru (approval number: 514.A/KEPK/UHTP/VIII/2025). Administrative permission for data collection was obtained from the participating institutions. The study was conducted in accordance with established ethical principles for research involving human participants (Bierer et al., 2024).

Before participation, students received an information sheet describing the study purpose, procedures, voluntary nature of participation, potential risks and benefits, confidentiality procedures, and the right to decline participation or withdraw without academic consequences. Written informed consent was obtained electronically before participation in the quantitative survey. Students invited to the qualitative interviews were additionally informed about the interview procedures, audio recording, use of anonymised quotations, and their right to decline individual questions or withdraw from the interview.

Participation was voluntary and was not linked to students' academic assessment or relationship with their instructors. Interview transcripts and quantitative records were de-identified before analysis. Participant codes rather than names were used in qualitative analysis and reporting. Only authorised members of the research team had access to the research data, which were stored using password-protected research storage.

Because the study involved students who were also enrolled in courses participating in the intervention, particular attention was given to avoiding perceived pressure to participate. Students were informed that participation or non-participation would not affect their grades, access to course materials, or relationship with instructors.

3.4.2 CR-DDL framework development and operationalisation

The CR-DDL Framework was developed by the research team specifically for the present study based on the integration of differentiated learning, culturally responsive teaching, and digital learning principles. The framework was developed as a study-specific pedagogical intervention rather than adopted from an existing intervention package. The development process translated the theoretical principles into observable learning activities and LMS functions that could be implemented consistently across the intervention courses.

The operationalisation process involved four stages. First, the research team identified the pedagogical functions of differentiated learning, culturally responsive teaching, and digital learning that were directly relevant to student engagement and learning equity. Second, these functions were mapped into specific instructional practices that could be delivered through the LMS. Third, the practices were organised into six interconnected components: culturally responsive introduction, multimodal core content, differentiated learning activities, cultural reflection and connection tasks, collaborative learning, and assessment and feedback. Fourth, the components were incorporated into the intervention modules and aligned with specific LMS features and learning activities (Table 1).

The six components served complementary functions within the intervention. Culturally responsive introduction established contextual relevance by connecting course topics with local cases and students' prior experiences. Multimodal core content provided learning materials in different formats, including text, video, and visual resources. Differentiated learning activities provided students with task options at different levels of complexity to accommodate variation in readiness and learning needs. Cultural reflection and connection tasks encouraged students to relate academic concepts to their cultural, community, and lived experiences. Collaborative learning supported peer interaction and shared problem-solving. Assessment and feedback provided opportunities for students to demonstrate learning through flexible formats and receive timely feedback for improvement.

The components were implemented as an integrated learning sequence rather than as independent instructional techniques. The LMS served as the delivery and monitoring environment, allowing instructors to provide multimodal materials, differentiated activities, reflective tasks, collaborative interaction, and assessment and feedback. The implementation was supported by instructor preparation and course redesign before the intervention. The framework was subsequently evaluated through self-reported engagement, LMS-derived behavioural indicators, academic achievement, and qualitative interviews. Thus, the framework provided the conceptual and operational basis for both the intervention and the evaluation strategy.

3.4.3 Implementation of the CR-DDL framework

The CR-DDL Framework was implemented in the intervention courses through a structured LMS-based learning process over approximately 10 weeks. Before the intervention, instructors assigned to the CR-DDL courses participated in a structured workshop conducted by the research team. The workshop introduced the theoretical rationale of the framework, the six operational components, and procedures for implementing culturally responsive and differentiated learning activities within the LMS.

Following the workshop, instructors worked with the research team to redesign selected course modules. The redesigned modules incorporated culturally relevant case studies, locally grounded examples, multimodal learning materials, differentiated task options, cultural reflection activities, collaborative learning opportunities, and flexible assessment and feedback (Azis and Maharani, 2024; Cochrane et al., 2017; Keith et al., 2023). The LMS was configured to support these activities through modular content organisation, multimedia resources, discussion forums, reflective activities, differentiated or branching tasks, online assessments, and instructor feedback.

Implementation consistency was supported through several procedural strategies. All intervention instructors received a common orientation to the six CR-DDL components, worked from the redesigned learning modules, and received implementation guidance from the research team. LMS activity was monitored throughout the intervention period, including login frequency, time spent on course content, discussion participation, and completion of intervention-related activities.

However, a formal implementation-fidelity assessment was not prospectively conducted using a prespecified component checklist, independent observer ratings, or a validated fidelity scale. The LMS indicators primarily reflected student participation and activity completion and therefore should not be interpreted as a formal measure of instructor adherence to all six CR-DDL components. Consequently, the present study cannot provide a quantitative fidelity score or determine precisely whether the six components were implemented with equivalent intensity across all intervention courses.

3.4.3.1 CR-DDL learning design in the LMS

The CR-DDL Framework was operationalised through structured learning modules delivered over approximately 10 weeks. Each module followed the six-component CR-DDL learning sequence described in Table 1: culturally responsive introduction, multimodal core content, differentiated learning activities, cultural reflection and connection tasks, collaborative learning, and assessment and feedback.

Culturally responsive introductory activities used local case scenarios and contextual prompts to activate students’ prior knowledge. Multimodal core content was provided through readings, videos, infographics, and other visual resources. Differentiated learning activities provided students with task options at different levels of complexity, allowing them to select an appropriate learning pathway according to their perceived readiness. Cultural reflection and connection activities required students to relate theoretical concepts to their cultural, social, community, or lived experiences. Collaborative learning was facilitated through discussion forums, peer exchange, and group problem-solving activities. Finally, assessment and feedback activities provided flexible opportunities to demonstrate learning and receive automated and/or instructor feedback.

The LMS provided the digital infrastructure through which these components were delivered, monitored, and documented. The learning design therefore linked the pedagogical principles of cultural responsiveness and differentiation with concrete digital learning functions rather than treating LMS use as a separate technological component.

3.4.3.2 Comparison condition

Students in the comparison courses received standard LMS-based instruction reflecting the usual instructional approach in the participating programmes. Course materials were primarily delivered through conventional LMS functions, including uploaded lecture materials, routine assignments, quizzes, and standard instructor communication. The comparison courses did not receive the structured integration of culturally responsive introductions, differentiated task pathways, cultural reflection activities, and flexible assessment incorporated into the CR-DDL intervention. This condition therefore represented the usual LMS-based learning approach and provided the comparison condition for evaluating the CR-DDL Framework.

3.4.3.3 Implementation timeline and monitoring

The implementation followed a structured timeline. Baseline data were collected during weeks 1–2, followed by the CR-DDL intervention during weeks 3–12, and post-intervention assessment during weeks 13–14. Academic achievement data were obtained at the end of the semester.

Throughout the implementation period, LMS analytics were used to continuously monitor student engagement. The monitored indicators included login frequency, time-on-task, participation in discussions, and completion of learning activities. These LMS-derived indicators were considered alongside survey-based engagement measures and academic performance records to provide a comprehensive evaluation of the intervention.

3.4.4 Interviews and participant feedback

The qualitative phase was conducted after preliminary examination of the quantitative findings. Participants were purposively selected from students who had completed both quantitative assessments and had indicated their willingness to participate in an interview. Selection sought variation in engagement trajectories and student characteristics, including socioeconomic background and cultural or ethnic identity. This sampling strategy was intended to capture a range of experiences rather than to recruit only students who reported positive experiences with the CR-DDL Framework.

Individual semi-structured interviews were conducted online using a secure videoconferencing platform. Each interview lasted approximately 45–60 min and was conducted in Bahasa Indonesia or English according to participant preference. The interviews followed a semi-structured guide covering differentiated learning choices, cultural relevance of learning materials, perceived inclusion and fairness, engagement, motivation, and experiences with LMS-based learning. The interview guide was piloted before the main interviews, and minor wording adjustments were made to improve clarity.

Before each interview, participants were reminded of the voluntary nature of participation, the purpose of the interview, their right to decline any question, and their right to withdraw without academic consequences. Participants provided consent before the interview, and audio recording was conducted only with participant permission. Interview data were transcribed verbatim. Where interviews were conducted in Bahasa Indonesia, transcripts were translated into English for reporting and manuscript preparation while retaining the original-language transcript as the primary source for analysis.

The interviews were conducted using the same general interview guide and procedures across participants, while allowing participants to elaborate on experiences that they considered important. Interviewers used neutral prompts and avoided evaluative responses to minimise the influence of interviewer expectations on participants' responses. Reflexive notes were maintained during the qualitative process to record emerging assumptions, contextual observations, and potential sources of interpretive influence.

The qualitative dataset comprised 18 completed interviews. Interviews were not treated as independent evidence of intervention effectiveness. Instead, they were used to provide contextual explanations for the quantitative findings and to identify how students experienced the different components of the CR-DDL Framework.

To protect participants' confidentiality, transcripts were de-identified before analysis. Each participant was assigned a study identification code (e.g., P01–P18), and identifying information was removed from quotations and analytic files. Interview data and related research files were stored in password-protected research storage accessible only to authorised members of the research team.

At the end of each interview, participants were invited to comment on whether any question was unclear or uncomfortable and whether an important aspect of their learning experience had not been discussed. These responses were considered during subsequent interviews and during interpretation of the qualitative dataset.

3.5 Data analysis

Quantitative analyses were conducted using SPSS version 24. Descriptive statistics were used to summarise participant characteristics, engagement scores, LMS indicators, and academic outcomes. Baseline characteristics of the CR-DDL and comparison groups were examined descriptively to assess the extent of observed balance before outcome comparisons. Because course allocation was not random, baseline similarity was not interpreted as evidence that the groups were equivalent with respect to unmeasured characteristics.

Linear mixed-effects models included a random intercept for course to account for students nested within course clusters (Luke, 2017). Comparable mixed-effects models were used for LMS-based engagement outcomes. The final-course-score model included intervention condition and prior-semester GPA while retaining the random intercept for course.

The original student-level sample-size calculation did not incorporate a cluster design effect or a prespecified ICC. Clustering was therefore addressed analytically rather than through prospective inflation of the required sample size. Because only six course clusters were included, the precision of course-level variance estimates and cluster-adjusted inference should be interpreted cautiously.

To evaluate socioeconomic disparities, SES and its interaction terms with condition and time were added to the engagement model. The primary analysis compared lower SES with the combined middle/higher SES group. Estimated marginal means and contrasts were used to characterise changes within SES strata and differences in changes between conditions.

All principal model estimates are reported with standard errors, 95% confidence intervals, and exact two-sided p-values. Statistical inference was based on α = 0.05; however, interpretation emphasised effect magnitude, confidence intervals, design limitations, and consistency across outcomes rather than statistical significance alone (Kim and Sax, 2019).

As a sensitivity analysis, the SES disparity model was re-estimated using the original three-category SES classification (lower, middle, and higher) rather than the dichotomised lower versus middle/higher classification. This analysis was conducted to determine whether the direction of the primary SES findings depended on the chosen grouping strategy. Because the higher-SES subgroup was small, the sensitivity analysis was interpreted primarily in terms of the direction and consistency of estimated effects and their confidence intervals rather than statistical significance alone.

3.5.2 Qualitative data analysis: thematic analysis

Qualitative data were analysed thematically following the general principles of Braun and Clarke's approach (Braun and Clarke, 2019). At the time of analysis, the research team reviewed the interview material, identified recurrent patterns related to differentiated learning, cultural relevance, inclusion, fairness, and engagement, and developed thematic interpretations from these patterns. The qualitative analysis was intended to complement the quantitative findings by providing contextual insight into students' experiences rather than to independently estimate intervention effects.

The analytic process involved familiarisation with the interview material, identification of relevant segments, generation and comparison of codes, development and refinement of candidate themes, definition of the final thematic structure, and selection of representative participant excerpts. The analysis combined attention to the predefined interview domains with concepts that emerged from participants' accounts.

We acknowledge an important documentation limitation. The original audio recordings, complete individual interview transcripts, and the full historical coding/audit files are currently unavailable in the research records accessible to the authors. Consequently, although the interviews and qualitative analysis were conducted as part of the study, the complete chain from original transcript to initial codes, sub-codes, categories, and final themes cannot now be independently reconstructed. We therefore do not present a newly reconstructed coding framework or transcript dataset.

The qualitative findings reported below are restricted to the interview material and participant excerpts that remain documented in the manuscript and associated research documentation. These excerpts are presented as previously documented qualitative evidence and should not be interpreted as a substitute for the unavailable complete raw interview dataset.

3.5.3 Integration of quantitative and qualitative findings

The qualitative findings were used to contextualise the quantitative results rather than to establish independent causal effects. Integration focused on whether the documented student accounts were consistent with and could help contextualise observed patterns in engagement, LMS participation, academic performance, and socioeconomic disparities. Given the current unavailability of the complete qualitative source records, the integrated interpretation is intentionally cautious and does not treat the qualitative material as independently verifiable evidence equivalent to the quantitative dataset.

3.5.4 Reflexivity and trustworthiness

The research team recognised that its involvement in developing and implementing the CR-DDL Framework could influence interpretation of students' accounts. Reflexive discussion was therefore incorporated during interpretation, particularly when considering students' perceptions of differentiation, cultural relevance, fairness, inclusion, and engagement.

The qualitative analysis was discussed among members of the research team to consider alternative interpretations and to reduce the risk that findings reflected a single researcher's perspective. Representative participant excerpts were used in the original manuscript to illustrate the reported thematic interpretations.

However, the authors acknowledge that the current unavailability of the original audio recordings, complete transcripts, and historical coding records limits retrospective assessment of dependability and confirmability. Accordingly, the qualitative component should be interpreted as contextual and exploratory evidence rather than as a fully auditable qualitative dataset.

4 Results

4.1 Participant characteristics

Of the 268 students initially invited, 252 met the inclusion criteria and completed both baseline and follow-up surveys, producing a final response rate of 94.0%. The CR-DDL group included 129 students and the comparison group 123 students.

The CR-DDL and comparison groups showed broadly similar observed baseline characteristics. Women constituted 66.7% of the CR-DDL group and 63.4% of the comparison group, and the distributions of year of study, programme, SES category, and cultural/ethnic background were also similar across conditions (Table 2). Baseline total engagement was 3.02 in the CR-DDL group and 3.05 in the comparison group, while prior-semester GPA was 3.20 and 3.22, respectively.

Table 2. Demographic characteristics of participants (N = 252).

These observed similarities reduce concern about large measured baseline imbalances but should not be interpreted as establishing equivalence between groups because course allocation was non-random. In particular, unmeasured course- and instructor-level characteristics may still have differed between conditions (Table 2).

4.2 Changes in HESES engagement scores

At baseline, mean total engagement was similar in the CR-DDL and comparison groups (3.02 vs. 3.05). At follow-up, mean total engagement was 3.54 in the CR-DDL group and 3.18 in the comparison group, corresponding to mean changes of +0.52 and +0.13, respectively. The estimated group × time effect was 0.31 (SE = 0.07; approximate Wald 95% CI 0.17–0.45; p < 0.001), indicating a larger increase in engagement over time in the CR-DDL courses.

A similar pattern was observed across the HESES domains. Behavioural engagement showed a group × time estimate of 0.35 (SE = 0.09; approximate Wald 95% CI 0.17–0.53; p < 0.001). The corresponding estimate was 0.24 for emotional engagement (SE = 0.08; approximate Wald 95% CI 0.08–0.40; p = 0.003) and 0.18 for cognitive engagement (SE = 0.08; approximate Wald 95% CI 0.02–0.34; p = 0.024). Thus, all engagement domains showed larger estimated changes in the CR-DDL condition, with the largest model estimate observed for behavioural engagement (Table 3).

Table 3. HESES total and subscale scores at baseline and follow-up.

Δ, change from baseline to follow-up; SE, standard error. Approximate Wald 95% confidence intervals were calculated from the reported model estimates and standard errors as estimate ± 1.96 × SE. Where the retained result was reported as p < 0.001, a more granular p-value was not reconstructed in order to avoid false precision.

4.3 LMS-based engagement indicators

The LMS data show that students in the CR-DDL courses became more active on the platform over time than students in the comparison courses. Average logins per week in the CR-DDL group increased from 3.8 to 5.6 (a change of +1.8), whereas the comparison group increased from 3.7 to 4.2 (+0.5). Time spent on course content pages rose by about 46 min per week in the CR-DDL group, compared with about 18 min in the comparison group. Forum participation followed the same pattern. Students in the CR-DDL group increased their mean number of posts and replies from 1.2 to 3.1 per week, while those in the comparison group increased from 1.1 to 1.7. When these indicators are combined into the standardized composite index, the CR-DDL group shows a clear shift from slightly below zero (−0.03) to almost half a standard deviation above the mean (0.47), whereas the comparison group moves only minimally (0.01 to 0.07) (Table 4).

Table 4. LMS engagement indicators by group.

Standardized composite of logins, time on content, and forum activity. Δ, change from baseline to follow-up; SE, standard error. Approximate Wald 95% confidence intervals were calculated as estimate ± 1.96 × SE. Where the retained result was reported as p < 0.001, a more granular p-value was not reconstructed.

Model estimates were consistent with the descriptive LMS patterns. The group × time estimate was 0.62 for weekly logins (SE = 0.18; approximate Wald 95% CI 0.27–0.97; p = 0.001), 0.55 for time spent on course content (SE = 0.17; approximate Wald 95% CI 0.22–0.88; p = 0.002), and 0.44 for forum posts and replies (SE = 0.15; approximate Wald 95% CI 0.15–0.73; p = 0.004). For the composite LMS engagement index, the group × time estimate was 0.36 (SE = 0.09; approximate Wald 95% CI 0.18–0.54; p < 0.001). These estimates indicate larger changes in LMS-based behavioural engagement in the CR-DDL condition while accounting for course-level clustering.

4.4 Learning outcomes

Prior-semester GPA was similar between the CR-DDL and comparison groups (3.20 vs. 3.22). Mean final course scores on the 0–100 reporting scale were 82.6 (SD = 6.8) in the CR-DDL group and 78.9 (SD = 7.1) in the comparison group. After adjustment for prior-semester GPA and course-level clustering, CR-DDL participation was associated with a 3.41-point higher final course score (SE = 1.12; approximate Wald 95% CI 1.21–5.61; p = 0.003).

This result should be interpreted cautiously because final scores were derived from routine assessments in different courses and disciplines rather than from a common standardised examination. Adjustment for prior GPA and course-level clustering reduces some sources of between-group and between-course variation but does not establish equivalence in assessment difficulty or grading practices across courses. The final course score is therefore interpreted as a supportive indicator of academic performance rather than as a fully standardised cross-course achievement measure (Table 5).

Table 5. Learning outcomes by group.

SE, standard error. Approximate Wald 95% confidence intervals were calculated from the reported adjusted estimates and standard errors as estimate ± 1.96 × SE. The adjusted final-course-score estimate accounts for prior academic performance and course-level clustering as specified in the statistical model. Because final course scores were derived from different courses and discipline-specific assessments, they should be interpreted as supportive indicators of academic performance rather than as fully standardized cross-course achievement measures.

4.5 Changes in engagement by socioeconomic status

At baseline, lower-SES students reported lower mean total engagement than middle/higher-SES students in both study conditions. Among lower-SES students, mean engagement increased from 2.88 to 3.49 (+0.61) in the CR-DDL condition and from 2.90 to 3.08 (+0.18) in the comparison condition. Among middle/higher-SES students, mean engagement increased from 3.08 to 3.56 (+0.48) in the CR-DDL condition and from 3.10 to 3.25 (+0.15) in the comparison condition.

Within the lower-SES stratum, the reported model estimate was 0.27 (SE = 0.11; approximate Wald 95% CI 0.05–0.49; p = 0.015). Within the middle/higher-SES stratum, the corresponding reported estimate was 0.10 (SE = 0.09; approximate Wald 95% CI −0.08 to 0.28; p = 0.281). Descriptively, the difference in change between CR-DDL and comparison courses was larger among lower-SES students.

These stratum-specific findings should not, however, be interpreted as a formal demonstration that the intervention effect differed significantly according to SES. A statistically significant estimate within one SES stratum and a non-significant estimate within another do not themselves establish a statistically significant difference between strata. Because a separately verifiable group × time × SES interaction estimate is not available in the currently reported results, the SES findings are interpreted as exploratory and descriptive evidence of a possible equity-related pattern (Table 6).

Table 6. Total engagement by SES subgroup and condition.

SES, socioeconomic status; Δ, change from baseline to follow-up; SE, standard error. The estimates shown are the stratum-specific estimates available from the reported analysis and should not be interpreted as the formal group × time × SES interaction. A statistically significant estimate within one SES stratum and a non-significant estimate within another do not themselves constitute a statistical test of whether the intervention effect differs between SES strata. Approximate Wald 95% confidence intervals were calculated as estimate ± 1.96 × SE. Accordingly, SES subgroup findings should be interpreted as exploratory evidence of a possible equity-related pattern rather than as definitive evidence of effect modification.

The primary SES analysis used the lower versus middle/higher grouping reported in the study dataset. A formal sensitivity analysis using an alternative SES classification is not presented here because such an analysis would require re-estimation from the participant-level dataset rather than reconstruction from the available summary statistics. The SES-related findings should therefore be regarded as dependent on the operational classification used in the primary analysis and interpreted cautiously.

4.6 Documented qualitative findings

Eighteen students participated in individual interviews as part of the qualitative component. The interview material was intended to provide contextual insight into students' experiences of differentiated learning, cultural relevance, inclusion, fairness, and engagement.

Three broad patterns were documented in the qualitative material retained in the study records: (1) differentiated task choices and engagement, (2) cultural relevance and meaningful learning, and (3) flexible assessment, fairness, and inclusion. Because the complete original interview records are currently unavailable, these findings are presented cautiously as documented qualitative patterns rather than as a fully reproducible thematic dataset.

4.6.1 Differentiated task choices and engagement

The documented interview material indicated that students associated having different task options with greater involvement and a greater sense of control over learning. One previously documented participant excerpt stated:

“Having options made it easier for me to stay involved. I could choose tasks that matched my level, so I didn't feel lost or overwhelmed.” (P07)

Another documented excerpt linked choice of assignment format with motivation and responsibility:

“When I could select the type of assignment, I felt more responsible for my learning and more motivated to complete it.” (P12)

These excerpts suggest that learner choice was perceived as making participation more manageable and supporting motivation. However, because the complete source transcripts are currently unavailable, the frequency and breadth of this pattern across all 18 interviews cannot now be independently reassessed.

4.6.2 Cultural relevance and meaningful learning

The retained qualitative material also documented students' perceptions that locally relevant examples and community-based contexts helped make learning more meaningful. One participant stated:

“The cases felt close to my daily life. It helped me understand the theory better because I could relate it to my community.” (P03)

Another participant commented:

“Usually, examples are very general, but here they reflected our context. It made the learning feel more meaningful.” (P15)

These excerpts provide contextual indications that culturally relevant examples were perceived as useful for connecting academic concepts with students' experiences. The findings should nevertheless be interpreted cautiously because the complete interview dataset is not currently available for independent verification.

4.6.3 Flexible assessment, fairness, and inclusion

The documented interview material further included perceptions that flexible assessment could make learning feel fairer and more inclusive. One participant stated:

“Different task options made the assessment feel fairer because not everyone learns the same way.” (P09)

Another participant reported:

“I felt that my background and perspective were valued, not ignored, especially during discussions and assignments.” (P18)

Together, these documented excerpts indicate that flexibility in task and assessment options was associated by some participants with perceptions of fairness, recognition, and inclusion. However, these excerpts should not be interpreted as representing the complete range of views among all 18 interview participants.

4.7 Mixed-methods integration

The quantitative and documented qualitative findings showed broadly consistent patterns, although the qualitative evidence is interpreted more cautiously because the complete original interview records are currently unavailable. Quantitatively, students in CR-DDL courses demonstrated greater improvements in self-reported engagement, LMS-based participation, and academic performance, with particularly pronounced engagement gains among lower-SES students.

The remaining documented interview excerpts provide contextual indications that several features of the CR-DDL Framework were experienced positively by some students. In particular, differentiated task choices were associated with greater involvement and motivation; culturally relevant examples were described as helping students connect course content with their experiences; and flexible task options were associated with perceptions of fairness and inclusion.

These qualitative observations are consistent with, but do not independently validate, the quantitative findings. They should therefore be regarded as contextual evidence that may help interpret how students experienced the intervention rather than as evidence establishing the mechanisms responsible for the quantitative effects.

The integration also does not establish that any individual CR-DDL component caused the observed quantitative improvements. Rather, the quantitative data provide the primary evidence for differences between the intervention and comparison conditions, while the available qualitative excerpts provide limited contextual insight into students' perceptions of the intervention.

5 Discussion

5.1 Main findings

This study examined whether a Culturally Responsive Digital Differentiated Learning (CR-DDL) Framework implemented within LMS-based courses was associated with improvements in student engagement, LMS participation, academic performance, and engagement disparities. The quantitative findings showed larger improvements in total HESES engagement and its behavioural, emotional, and cognitive domains in the CR-DDL group than in the comparison group. LMS indicators showed a similar pattern, while final course performance was also higher in the CR-DDL group after adjustment for prior GPA and course-level clustering. Importantly, lower-SES students in the CR-DDL condition demonstrated particularly pronounced engagement gains, resulting in a smaller engagement gap at follow-up.

The qualitative component was originally included to contextualise these quantitative patterns. The interview material that remains documented in the study records points to three potentially relevant aspects of the intervention: learner choice, cultural relevance, and perceived fairness and inclusion. However, because the original interview recordings, complete transcripts, and historical coding records are currently unavailable, these qualitative observations should be interpreted as contextual rather than independently verifiable evidence. Accordingly, the quantitative findings constitute the primary basis for interpreting the observed associations between CR-DDL participation and the study outcomes, while the retained qualitative material is used only to provide limited contextual insight into students' experiences.

5.2 Comparison with previous research

The overall increase in engagement in the CR-DDL group aligns with prior work showing that technology, when used deliberately, can support more active and sustained student participation (Bond et al., 2020). However, much of the earlier literature on technology-enhanced learning has focused on generic design features, such as interactive content or frequent feedback without systematically integrating differentiation and cultural responsiveness (Rehman, 2023). The present findings suggest that a design which explicitly combines differentiated pathways with culturally grounded materials may yield stronger engagement effects than technology alone.

An important contribution of this study is the operationalisation of the CR-DDL Framework as an integrated rather than additive pedagogical approach. Differentiated learning, culturally responsive teaching, and digital learning were not implemented as separate instructional strategies; instead, they were connected through a common LMS-based learning sequence. This distinction is important because previous work has often examined differentiated learning, culturally responsive pedagogy, or technology-enhanced learning independently (Cochrane et al., 2017; Harmilawati et al., 2024; Rehman, 2023). The present framework translates these complementary principles into six observable instructional components and links each component to specific learning activities and LMS functions. The resulting framework provides a more explicit pathway for instructors to move from broad commitments to inclusion and responsiveness toward concrete course-design practices that can be implemented and evaluated.

The pattern of results is consistent with studies reporting that differentiated learning can improve motivation and outcomes when it offers meaningful choices and adapts to students' readiness levels (Azis and Maharani, 2024; Harmilawati et al., 2024). In the retained interview excerpts, some participants described differentiated digital tasks and flexible pathways as making participation feel more manageable and less overwhelming. These observations provide contextual illustrations rather than evidence of the frequency or prevalence of this perception across the complete interview sample because the full qualitative dataset is no longer available for re-analysis (Cochrane et al., 2017; Johnson, 2022; Nurbatra and Masyhud, 2022). The present study extends that literature by demonstrating comparable patterns in digital environments and by linking these experiences to quantitative engagement and performance indicators.

The narrowing of engagement gaps for lower-SES students is particularly noteworthy. Prior studies have documented persistent differences in engagement by socioeconomic status and other structural factors, with lower-SES students often reporting lower participation and a stronger sense of “falling behind,” especially in the wake of the COVID-19 pandemic and rapid digitalization (Crabtree, 2021; Ebede, 2018; Gulko et al., 2024; Hope, 2021). In many settings, shifts to online and blended learning have exacerbated inequities in access, support, and outcomes rather than mitigating them (Bond et al., 2020; Hope, 2021). In the present study, equity-oriented digital design incorporating differentiated supports and culturally responsive content was associated with larger observed engagement gains among lower-SES students. Given the non-random course allocation and the exploratory subgroup analysis, this pattern should be interpreted as an association rather than evidence that CR-DDL itself caused the differential improvement.

The SES-specific pattern in this study is broadly consistent with work on equity-minded digital learning and culturally responsive online pedagogy. Keith and colleagues argue that equity-oriented digital designs must deliberately center students who are structurally disadvantaged, rather than assuming that general “good practice” will automatically benefit them (Keith et al., 2023). Similarly, Boison and Burke, working in school contexts, found that culturally responsive, family-engaged digital approaches could help bridge digital equity gaps for younger learners (Boison and Burke, 2025). While our study takes place in higher education and uses a different framework, the observed narrowing of engagement gaps suggests that CR-DDL may play a comparable levelling role at the tertiary level.

The qualitative themes also resonate with recent work on culturally responsive pedagogy in online and hybrid formats. Students in our study emphasised the importance of choice, cultural relevance, and perceived fairness in assessment, echoing findings from Kumi-Yeboah and Amponsah's exploratory work on instructors' perceptions of culturally responsive online pedagogy and from Mills' discussion of humanising online learning through culturally responsive practices (Kumi-Yeboah and Amponsah, 2023; Mills, 2025). However, much of that literature has been conceptual or perception-focused. The contribution of the present study is to connect those ideas to measurable changes in engagement, LMS behaviour, and assessment scores in a quasi-experimental design.

At the same time, our findings highlight the importance of context. The CR-DDL Framework was developed and implemented in Indonesian public health and agrotechnology programmes, with a student body comprising Malay, Minangkabau, Batak, Javanese, Sundanese, and other ethnic groups. Work by Nurbatra and Masyhud has already pointed to the importance of adapting culturally responsive teaching to local multicultural realities in Indonesian higher education (Nurbatra and Masyhud, 2022). The current study is consistent with that argument and suggests that culturally responsive digital differentiation can be tailored to such contexts in ways that both respect local diversity and address structural barriers to engagement.

5.3 Practical and educational implications

The findings have several implications for higher education practice and future evaluation. First, the larger observed engagement gains in CR-DDL courses suggest that integrating differentiated learning and culturally responsive pedagogy into LMS design may be a promising approach for supporting student engagement compared with conventional LMS-based course structures. Because the study used a non-randomised design, these findings should support further implementation and evaluation rather than be interpreted as definitive evidence of causal effectiveness. For instructors, this means that equity-oriented digital design should go beyond adding multimedia or occasional choice; it requires systematic mapping of students' backgrounds and needs, designing multiple task options, and embedding local and culturally relevant examples throughout the course. Second, the larger observed gains among lower-SES students suggest that CR-DDL warrants further investigation as an equity-oriented educational strategy rather than solely as a general engagement approach. For institutions, this suggests that investments in capacity-building around culturally responsive differentiation, and in LMS configurations that can support branching paths and multiple assessment formats, may be particularly valuable for meeting inclusive engagement and completion targets. In settings where resources are constrained, starting with key high-risk courses or cohorts could be a pragmatic way to pilot and scale such approaches. Third, the alignment between self-reported engagement and LMS log indicators in this study reinforces the value of combining survey and behavioural data to monitor engagement over time. Institutions could use similar data combinations to identify courses or student groups that might benefit most from CR-DDL-style redesigns, and to track whether such interventions are narrowing or widening observed gaps.

5.4 Limitations and directions for future research

Several limitations should be considered when interpreting the findings. First, allocation occurred at the course level and was non-random, with only six course clusters included. Although the CR-DDL and comparison groups showed broadly similar observed baseline characteristics, including baseline engagement and prior-semester GPA, these similarities do not establish equivalence between conditions. Allocation also depended partly on instructor availability and willingness to implement the CR-DDL redesign. Instructor-level characteristics, including teaching experience, pedagogical style, digital teaching proficiency, and familiarity with culturally responsive or differentiated instruction, were not systematically measured and therefore could not be adjusted for. Residual confounding by course and instructor characteristics remains possible.

Second, the original sample-size calculation was based on a student-level mixed-ANOVA framework and did not incorporate a cluster design effect or a prespecified ICC. Course clustering was addressed analytically using random-intercept mixed-effects models; however, with only six course clusters, estimation of cluster-level variation and the precision of cluster-adjusted effects remain limited. Future studies should incorporate the expected ICC and design effect during prospective sample-size planning and should include a substantially larger number of course clusters.

Third, SES was operationalised using self-reported baseline socioeconomic information and was dichotomised as lower versus middle/higher SES for the primary disparity analysis to maintain adequate subgroup sizes. This classification is not equivalent to a validated multidimensional socioeconomic index, and different operational definitions may produce different subgroup patterns. Sensitivity analyses using the original three-category classification should therefore be considered when evaluating the robustness of SES-related findings.

Fourth, formal intervention fidelity was not prospectively quantified using a prespecified component checklist, independent observer ratings, or a validated fidelity instrument. Although instructors received common preparation and implementation guidance and LMS activity was monitored throughout the intervention, these procedures do not establish equivalent adherence to all six CR-DDL components across intervention courses. Variation in implementation intensity may therefore have contributed to between-course differences.

Fifth, final course scores were derived from routine assessments in different courses and disciplines. Although scores were reported on a common 0–100 scale and analyses adjusted for prior GPA and course clustering, assessment content, difficulty, and grading practices were not formally equated across courses. Final course score findings should therefore be interpreted as supportive evidence of an association with academic performance rather than as a fully standardised comparison of achievement.

Sixth, the study was conducted in public health and agrotechnology programmes within a specific Indonesian regional context. This may limit transferability to other disciplines, institutions, student populations, and educational systems. The relatively short follow-up period also prevents conclusions about whether observed engagement patterns persist across subsequent semesters or translate into outcomes such as retention or graduation.

Finally, an important limitation concerns the qualitative component. Although 18 interviews were conducted, the original audio recordings, complete individual transcripts, and historical coding/audit files are currently unavailable. The complete analytical chain from original participant responses through codes, categories, and themes therefore cannot be independently reconstructed. The retained qualitative excerpts are consequently presented only as contextual and exploratory material and should not be interpreted as independently reproducible evidence or as confirmation of the mechanisms underlying the quantitative associations.

Future research should use randomised, cluster-randomised, or stepped-wedge designs involving a larger number of courses and institutions; prospectively account for clustering during sample-size planning; incorporate formal implementation-fidelity measures; use more robust multidimensional indicators of SES; standardise or equate academic outcome measures across courses; and maintain complete, securely archived qualitative audit trails.

6 Conclusion

This quasi-experimental mixed-methods study found that participation in LMS-based courses implementing the Culturally Responsive Digital Differentiated Learning (CR-DDL) Framework was associated with greater improvements in self-reported engagement and LMS-based participation than participation in standard LMS-based courses. CR-DDL participation was also associated with higher adjusted final course scores, although this academic outcome should be interpreted cautiously because assessments were course-specific and were not formally equated across disciplines.

The observed engagement gains were particularly pronounced among lower-SES students, and the engagement gap between SES groups was smaller at follow-up in the CR-DDL condition. These findings suggest a potentially important equity-related pattern, but they do not establish that CR-DDL caused the reduction in disparity because course allocation was non-random, only six course clusters were included, and the SES subgroup analysis is sensitive to how socioeconomic status is operationalised.

The CR-DDL Framework provides a structured way of integrating culturally responsive teaching, differentiated learning, and digital learning design through six interconnected components. The present findings support further rigorous evaluation of this framework rather than definitive causal conclusions about its effectiveness. Future multi-institutional studies using a larger number of clusters, prospective cluster-adjusted sample-size planning, formal fidelity assessment, robust SES measures, standardised academic outcomes, and longer follow-up are needed to determine the reproducibility, transferability, and sustainability of the observed associations.

The retained qualitative excerpts offer limited contextual insight into how some students experienced learner choice, culturally relevant content, and flexible assessment. Because the complete original qualitative dataset and audit trail are unavailable, these observations should remain secondary to the quantitative findings and should not be interpreted as independently reproducible evidence.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

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.

Supplementary material

The Supplementary Material for this article can be found online at: https://www.frontiersin.org/articles/10.3389/feduc.2026.1865489/full#supplementary-material

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Supplementary material Supplementary material at the source

Footnotes

Edited by: Nashwa Ismail, Imperial College London, United Kingdom

Reviewed by: Sohaib Ahmed, Taylor’s University, Malaysia

Sultan M. Alshahrani, King Khalid University, Saudi Arabia