From insight to intervention: learning analytics and generative AI support for learning design and educational decision-making

Synthesis: Rienties, Divjak and Law (2026) edit a special issue on learning analytics and generative AI in learning design, and their editorial's argument is that neither technology becomes educationally valuable through technical capability alone. Synthesizing four contributions across student, teacher, disciplinary and institutional levels, they identify the mediators of value as learning design, community and interaction, digital literacy, educator and learner agency, ethical fitness, organizational strategy, resources and stakeholder engagement. Their sharpest claim is methodological: the evidence base is dominated by cross-sectional, self-report, discourse-based and expert-judgment studies, so the field must move from demonstrating associations and proposing frameworks to intervention research that tests causal mechanisms, implementation conditions, longer-term outcomes and distributional effects.

Key Findings

1. The special issue spans four levels of analysis. The four contributions address the student level (Dong and colleagues on generative learning analytics tools and engagement), the disciplinary level (Tang and Wang on film education), the educator level (Claassen and colleagues on learning design reasoning), and the institutional level (Svetec, Divjak and Kadoić on enablers of intervention).
2. Engagement is mediated, not produced. Dong and colleagues report a direct positive relationship between generative learning analytics tool use and engagement, with partial mediation through the three Community of Inquiry presences — cognitive, social and teaching. Digital literacy strengthened the relationship between tool use and each presence dimension.
3. A configurational finding qualifies the main effect. Dong and colleagues' fuzzy-set qualitative comparative analysis identified several alternative pathways to high engagement, including one in which strong Community of Inquiry sustains engagement even when tool use is limited. This rejects a simple technology-effects model.
4. Ethical fitness predicts creative outcomes in film education. Tang and Wang's LA-EDM framework, tested with 508 students across five Chinese institutions, 10 educator interviews and 10 student films, found ethical fitness to be a significant predictor of technical-artistic balance and an indirect contributor to educational effectiveness. The qualitative strand documents both empowerment and creative deskilling from AI-generated content.
5. Educators integrate technical affordances with local knowledge. Claassen and colleagues' analysis of 11 focus groups found that learning analytics discussions clustered around contextual information, course-level design, creative problem solving, student support and activity redesign, while generative AI discussions connected more strongly to assessment design, ideation, contextual judgment and designing for student self-determination.
6. Institutional enablers are interdependent conditions for action. Svetec and colleagues used a survey, an international Delphi process, and a group decision-making approach combining the Analytic Hierarchy Process with network influence to identify seven enablers. Prioritization placed institutional strategic orientation and available resources first, followed by pedagogical and research foundations and stakeholder engagement.
7. The editorial calls for causal, longitudinal and distributional research. The six-direction agenda asks for intervention and causal explanation, multilevel evidence, quality measures beyond engagement and efficiency, treatment of GenAI systems as heterogeneous and changing, participatory and culturally responsive infrastructures, and research on professional and organizational learning.

Why the mediation findings matter more than the main effects

Dong and colleagues' contribution is the clearest example of the collection's logic. A direct relationship between tool use and engagement would be easy to report and easy to misread as a technology effect. The mediation structure through cognitive, social and teaching presence locates the mechanism in pedagogy and interaction: the tool's Feedback, visualizations and recommendations acquire meaning through social and communicative learning designs. The configurational result reinforces this, because it shows that high engagement is reachable by more than one route — including routes where the tool contributes little. The practical consequence is that adopting a generative learning analytics tool without attending to presence and design may produce no engagement gain at all. The same reading applies across the collection: Svetec and colleagues find that resources and strategic orientation must be in place before analytics insights convert into intervention, which is the institutional analogue of the same argument.

Ethics as productive rather than restrictive

Tang and Wang's treatment of ethics in film education runs against the usual framing. Ethical fitness — copyright, originality, bias, cultural respect and human oversight — is modeled as a condition for high-quality creative work rather than as a constraint on it, and it significantly predicts technical-artistic balance. The qualitative findings give the claim texture: AI-generated content can empower students and increase efficiency while simultaneously producing creative deskilling, shallow work, stereotyped cultural representations and dependence on algorithmic conventions. Cultural adaptability functions as a localisation filter in the framework, which is the collection's reminder that models trained on global data do not enter culturally neutral classrooms. The editorial connects this to Law and Liang's ecological research on sustainable educational innovation, where technology use is shaped by institutional, cultural, community and policy priorities — a reading that makes ethical fitness a design variable rather than a compliance checkbox.

From institutional readiness to a testable theory of change

Svetec and colleagues' seven enablers — institutional strategic orientation, pedagogical and other research foundations, available resources, pedagogical support, ethics and data AI Governance, stakeholder engagement, and quality assurance — are presented as interdependent conditions rather than as a maturity checklist. The editorial's agenda takes this further by asking for the enablers to be operationalized into maturity indicators and tested for whether changes in those indicators predict implementation quality and educational outcomes. That move would convert institutional readiness from a descriptive framework into a theory of change. The editorial is equally pointed about the theoretical ambition of the field as a whole: frameworks connecting resources, capabilities, social interaction, Motivation, design and institutional change are useful, but breadth can become accumulation without explanation. Its recommendation is that studies state what each theory adds, which mechanism it predicts, what evidence could falsify the claim, and at which level it operates — a demand for parsimonious explanations rather than ever more comprehensive models.

What this means for practice

• Instructors. Attend to presence and design before adopting a tool. Dong and colleagues' mediation and configurational findings both indicate that engagement gains depend on the social and pedagogical structures around the technology.
• Faculty developers. Claassen and colleagues' results show that educators already combine technical affordances with local knowledge. Professional learning should build on that reasoning rather than replace it with tool training.
• Assessment and learning designers. Treat ethical fitness as a design requirement in creative disciplines. Tang and Wang's finding that it predicts technical-artistic balance makes it part of quality, not a separate compliance step.
• Institutional leaders. Use the seven enablers as an audit of preconditions for intervention. Strategic orientation and resources were prioritized first, which suggests that analytics initiatives fail at the institutional layer rather than the analytical one.
• Researchers. Move from association to intervention. The editorial's agenda is explicit that longitudinal, quasi-experimental and design-based research is needed, with process evaluation of implementation fidelity, educator adaptation and local constraints.

Limitations

• The special issue comprises four contributions. The editorial's synthesis rests on a small number of studies, and its claims about the field's overall state are inferred from that collection plus prior literature rather than from a systematic review.
• The evidence base it critiques is its own. The editorial states that the field is dominated by cross-sectional, self-report, discourse-based and expert-judgment studies, and the four contributions largely fall within those designs, so the collection illustrates the problem it identifies.
• Cross-national adaptation is proposed but not tested. The editorial calls for frameworks such as Community of Inquiry, ethical fitness, digital literacy and institutional enablers to be tested for measurement equivalence and adapted through local co-design, which means equivalence is currently unknown.
• The multilevel connection remains aspirational. The four articles operate across levels but, as the editorial notes, rarely connect them empirically, so claims about how institutional conditions interact with individual capability are not established.
• Distributional effects are identified as a gap. The agenda asks who receives better support, who is misclassified and whose participation remains invisible, but the collection does not report those analyses.
• The editorial's theoretical recommendation is programmatic. Its call for parsimonious, falsifiable explanations is a standard for future work rather than a description of the current collection.

Citation

Rienties, B., Divjak, B., & Law, N. (2026). From insight to intervention: learning analytics and generative AI support for learning design and educational decision-making. International Journal of Educational Technology in Higher Education, 23(1).