Online university courses often give every student the same readings, assignments and deadlines, even when their academic preparation and everyday circumstances differ. An Indonesian study suggests that a more flexible approach may be associated with stronger engagement and higher course marks.
The researchers followed 252 undergraduates across six courses. Students in redesigned courses reported a larger increase in engagement than those receiving conventional online instruction, while their final marks were also higher after statistical adjustment.
Published in Frontiers in Education on 8 October 2026, the research examines teaching choices rather than the introduction of a new commercial platform. Its findings are promising, but the non-randomised comparison does not establish that the redesign caused the differences.
Why uploading lectures is not enough
Learning management systems can distribute materials, collect assignments, host discussions and record student activity. Those functions can make teaching more convenient without addressing differences in students’ backgrounds, language, access to resources or familiarity with course examples.
The study investigated a combined teaching framework called Culturally Responsive Digital Differentiated Learning, or CR-DDL. It joins flexible learning pathways with teaching materials that draw on students’ communities and experiences.
The central idea is that learners should encounter the same core academic objectives through activities that reflect different starting points. Rather than equating fairness with identical tasks, the approach tries to make demanding material accessible without abandoning academic standards.
Digital access remains important, but connectivity alone cannot ensure participation. Students may be able to open a course page and still find its examples unfamiliar, its task difficulty unsuitable or its opportunities for interaction limited.
Six connected changes to course design
The first element of CR-DDL introduced course topics using local cases and culturally familiar contexts. Students were invited to connect abstract concepts with their own communities rather than treating their experiences as irrelevant to academic study.
The second element provided core information in multiple formats, including written materials, videos and visual resources. The third offered tasks at basic, intermediate and advanced levels so students could select an appropriate degree of challenge.
A fourth element encouraged reflection on cultural and community connections. Collaborative discussions and group problem-solving formed the fifth, while flexible assessments and feedback made up the sixth element.
These components operated as one instructional package. The research cannot determine which individual component was most influential, nor whether a smaller set of changes would have produced comparable outcomes.
Comparison courses used familiar learning-management-system functions, such as lecture uploads, standard assignments, quizzes and routine instructor communication. They did not receive the structured integration of local examples, differentiated tasks, cultural reflection and flexible assessment.
Who participated in the research
The study involved three higher education institutions in Pekanbaru, Indonesia. Participants were undergraduates enrolled in public health and agrotechnology courses, two areas where online and blended learning already featured in ordinary teaching.
The final quantitative sample included 252 students nested within six course clusters. Three redesigned courses contained 129 students, and three comparison courses contained 123 students.
Course allocation depended on scheduling and instructors’ willingness to implement the framework, rather than random assignment. This makes the investigation quasi-experimental, not a randomised controlled trial.
The distinction is consequential. An instructor who volunteered to redesign a course might differ from another instructor in experience, enthusiasm or familiarity with digital teaching, and these differences could influence student outcomes.
Baseline assessments took place in weeks one and two. The intervention operated during weeks three to twelve, followed by engagement measurement in weeks thirteen and fourteen and final course marks at semester end.
Eighteen students also completed semi-structured interviews intended to illuminate how they experienced the redesigned activities. Those accounts were supplementary and were not designed to establish the intervention’s effectiveness independently.
How engagement was measured
The researchers used the 23-item Higher Education Student Engagement Scale, which captures behavioural, emotional and cognitive engagement. Items are rated on a five-point scale, with higher average scores indicating stronger self-reported engagement.
The questionnaire was translated and adapted for the Indonesian context. In a 60-student pilot, the adapted total scale had a reported reliability coefficient of 0.88, while subscale reliability coefficients ranged from 0.81 to 0.86.
The team also analysed platform records: weekly logins, minutes spent on course content, and discussion posts or replies. Academic outcomes came from final course marks, with prior-semester grade point average included as a statistical control.
Mixed-effects models accounted for students belonging to different course clusters. That approach reduces one problem created by grouped data, but it cannot remove unmeasured differences between instructors or make non-random groups fully equivalent.
Engagement rose more in redesigned courses
At the start, mean total engagement was 3.02 out of five in redesigned courses and 3.05 in comparison courses. By follow-up, those scores had increased to 3.54 and 3.18 respectively.
The redesigned group therefore recorded a 0.52-point increase, compared with 0.13 points in conventional courses. The modelled group-by-time estimate was 0.31 points, with a standard error of 0.07 and an approximate 95% confidence interval from 0.17 to 0.45.
The reported p-value for that overall estimate was below 0.001. This indicates that the observed difference is unlikely to be explained by the model’s sampling uncertainty alone, although it does not resolve the study’s non-randomised design.
Emotional engagement increased by 0.47 points in the redesigned group and 0.16 in the comparison group. The corresponding model estimate was 0.24, with an approximate 95% confidence interval from 0.08 to 0.40.
Cognitive engagement rose by 0.45 points in redesigned courses and 0.16 in conventional courses. Its modelled difference was 0.18, with an approximate confidence interval from 0.02 to 0.34.
These domains are worth distinguishing. Logging into a website is a behavioural action, while emotional engagement concerns connection to learning and cognitive engagement reflects the mental effort students report investing in the work.
Platform records supported the survey pattern
Weekly logins increased from 3.8 to 5.6 in redesigned courses, a gain of 1.8. In conventional courses, they increased from 3.7 to 4.2, a gain of 0.5.
Time spent on course content rose from 96.4 to 142.3 minutes per week in redesigned courses, an increase of 45.9 minutes. The comparison group increased from 94.8 to 112.5 minutes, or 17.7 minutes.
Discussion-forum posts and replies increased from 1.2 to 3.1 per week in redesigned courses. The conventional group increased from 1.1 to 1.7, meaning the changes were 1.9 and 0.6 interactions respectively.
The reported group-by-time estimates were 0.62 for logins, 0.55 for content time and 0.44 for forum activity. Their respective p-values were 0.001, 0.002 and 0.004, indicating consistent modelled differences across these platform measures.
A standardised composite of learning-platform activity also favoured the redesigned courses, with a reported group-by-time estimate of 0.36 and p below 0.001. This combined several indicators rather than relying on one measure of participation.
More online activity does not necessarily mean more learning. Students may remain on a page because content is confusing or because they are distracted, so these behavioural indicators are best considered alongside the survey and assessment findings.
Course marks differed by 3.41 adjusted points
Final course scores averaged 82.6 out of 100 in redesigned courses and 78.9 in comparison courses. The raw difference was 3.7 points, while the adjusted difference was 3.41 points after accounting for prior GPA and course clustering.
The adjusted estimate had a standard error of 1.12, an approximate 95% confidence interval from 1.21 to 5.61, and p = 0.003. Previous-semester GPA averages were similar, at 3.20 and 3.22 in the two groups.
However, final marks came from different courses and discipline-specific assessments, not a single standardised examination. Even when scores are placed on a common 100-point scale, marking practices and assessment difficulty can differ.
The 3.41-point estimate is therefore evidence of an association with reported course performance in this setting. It should not be presented as a guaranteed improvement in examination results if universities adopt the framework elsewhere.
A possible narrowing of socioeconomic differences
Students classified as lower socioeconomic status started with lower average engagement than their more advantaged peers. The researchers examined whether the redesigned courses were associated with different changes across these groups.
Among lower-status students, mean engagement increased from 2.88 to 3.49 in redesigned courses, a gain of 0.61. In conventional courses it rose from 2.90 to 3.08, a gain of 0.18.
For middle- and higher-status students, engagement increased from 3.08 to 3.56 in redesigned courses, a gain of 0.48. Their counterparts in conventional courses increased from 3.10 to 3.25, a gain of 0.15.
These descriptive figures suggest that the engagement gap between lower-status and better-resourced students became smaller in the redesigned courses. That pattern is interesting because the framework was intended to address differences in student circumstances.
It is not proof that the framework benefited disadvantaged students more. The authors did not provide a separately verifiable three-way interaction test comparing the intervention effect across socioeconomic groups.
A significant result in one subgroup and a non-significant result in another do not establish a significant difference between them. The socioeconomic pattern should therefore be described as exploratory, not as a confirmed reduction in inequality.
Socioeconomic status was also measured using self-reported information and grouped into lower versus middle-or-higher categories. Other definitions of disadvantage might change the size or direction of the observed subgroup differences.
What the interviews add
Eighteen students participated in interviews lasting approximately 45 to 60 minutes. Retained comments suggested that some appreciated being able to select tasks at suitable difficulty levels and connect academic ideas with their own communities.
Other comments described flexible assessment and a greater sense that students’ backgrounds were acknowledged. These accounts are compatible with the quantitative findings but do not show which part of the framework produced the measured differences.
An important limitation is that the original audio recordings, complete transcripts and historical coding records were unavailable to the researchers when the paper was prepared. Independent reviewers therefore cannot reconstruct the full qualitative analysis.
The published excerpts should be treated as contextual observations, not as a complete or independently reproducible qualitative dataset. The questionnaire, platform and academic data provide the primary evidence for the study’s conclusions.
What institutions can reasonably learn
The findings suggest a practical area for experimentation: course design within existing learning platforms. Institutions could test locally relevant examples, meaningful task choices, multiple formats and structured discussion without assuming that a new software subscription is necessary.
Flexibility should not become a euphemism for lower academic standards. Different task formats should assess comparable learning objectives and use clear marking criteria so that choice does not undermine fairness or academic rigour.
Instructors would also need training and time to prepare alternative pathways and feedback. Offering more formats and options may increase workload, so the approach should be evaluated for sustainability as well as short-term engagement.
For universities serving multilingual and economically diverse students, culturally relevant examples may be worth testing. But the Indonesian results cannot be assumed to transfer unchanged to South African institutions with different curricula, resources and student populations.
Learning analytics can flag changing patterns of participation, yet low login counts may reflect limited connectivity, offline study or work commitments rather than disinterest. Platform records should be interpreted with students’ circumstances in mind.
The limitations that prevent a causal conclusion
The sample was restricted to 252 students, three institutions and six courses in one Indonesian city. The course-level assignment was not random, and instructor willingness could have influenced who received the redesigned teaching.
The original power calculation treated students as independent observations and did not include a prospective adjustment for clustering. Later models accounted for courses, but only six clusters were available for estimating between-course variation.
The study did not use a formal prospective checklist to quantify how consistently every instructor delivered all six components. Differences in implementation intensity could have contributed to the observed course-level outcomes.
Course marks were not formally equated across disciplines, the socioeconomic analysis remained exploratory and the qualitative audit trail was incomplete. The follow-up period also did not establish whether engagement gains persisted into later semesters.
Stronger evidence would come from a larger cluster-randomised trial, standardised or equated assessments, explicit implementation checks and longer follow-up. Complete interview records would allow the proposed mechanisms to be scrutinised independently.
A useful signal, not a final verdict
Across questionnaires, platform activity and course scores, the redesigned courses showed a broadly consistent advantage over conventional online instruction. The evidence supports further testing of a framework that treats student differences as part of course design.
The defensible conclusion is that culturally responsive, differentiated digital teaching was associated with stronger engagement and somewhat higher marks in this setting. Causation, lasting academic gains and reduced socioeconomic inequality remain questions for future research.
The study poses a useful challenge for universities: improving digital education may require more thoughtful teaching pathways and culturally meaningful examples, not simply placing additional material on the same learning platform.
Source Information
Study title: 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
Authors: Reno Renaldi, Aldiga Rienarti Abidin, Syafrani and Dwi Sapta Aryantiningsih
Journal: Frontiers in Education, volume 11
Publication date: 8 October 2026
Study design: Non-randomised quasi-experimental explanatory sequential mixed-methods study
Study location: Three higher education institutions in Pekanbaru, Indonesia
Sample: 252 students in six course clusters, with 129 in redesigned courses and 123 in comparison courses; 18 follow-up interviews
Measurements: Engagement questionnaire, LMS behavioural records, prior GPA, final course marks and mixed-effects models
Original study DOI: 10.3389/feduc.2026.1865489
