From Learning Analytics to Educational Interventions: Enhancing Decision-Making and Learning Design

Synthesis: Svetec, Divjak, and Kadoić (2026) address the persistent gap between learning analytics (LA) and actual educational interventions — the challenge of "closing the loop" between data collection and educational change. Using a group decision-making methodology (Delphi validation, Analytic Hierarchy Process, and the Social Network Analysis Process / SNAP), an international expert panel identified, validated, and prioritized seven enablers of trustworthy LA-based educational interventions: institutional strategic orientation, pedagogical & other research foundations, available resources, pedagogical support, ethics & data governance, stakeholder engagement, and quality assurance. Institutional strategic orientation emerged as the most important enabler overall, followed by available resources, with the study positioning trustworthiness as a prerequisite without which LA-based interventions are not meaningful.

Key Findings

1. Seven enablers of LA-based educational interventions. A preliminary list of seven enablers was developed from the literature, validated and refined through a two-round international Delphi study (18 experts), and prioritized via SNAP group decision-making: institutional strategic orientation, pedagogical & other research foundations, available resources, pedagogical support, ethics & data governance, stakeholder engagement, and quality assurance.
2. Institutional strategic orientation is the central enabler. It ranked highest in overall SNAP priority (0.2072) and, crucially, was the most influential on other enablers (PageRank 0.2430) — strategic planning and leadership commitment position it as the enabler that most enables the rest.
3. Available resources ranked second (SNAP 0.1800) and highest on direct influence on intervention implementation (AHP 0.1753), pointing to the need for systematic investment in (AI-supported) LA systems, dashboards, and infrastructure.
4. Stakeholder perspectives differ systematically. Cluster analysis revealed that senior LA experts prioritized available resources, junior/technical experts prioritized ethics & data governance, and intermediate experts prioritized strategic orientation or stakeholder engagement — underlining the value of heterogeneous expert groups in LA decision-making.
5. Trustworthiness is the prerequisite, not an enabler dimension. The enablers are conditions enabling LA-based interventions, whereas trustworthiness (ethical compliance, data security, transparent and unbiased algorithms, pedagogical validity) is the precondition without which LA-informed interventions are not meaningful.

The enabler framework

The seven enablers are defined as institutional practices, capabilities, or capacities that support, facilitate, and enhance the implementation of educational interventions based on trustworthy LA:

• Institutional strategic orientation — strategic planning and leadership commitment to implementing trustworthy LA and using LA insights to inform educational interventions (e.g. student-facing LA dashboards, ML-based early-warning for at-risk students).
• Available resources — investment in AI-supported institutional LA systems, teacher/student dashboards, infrastructure, and human capacity.
• Pedagogical & other research foundations — LA grounded in contemporary pedagogical principles, learning theories, and sound, student-centered learning design based on learning outcomes and constructive alignment.
• Pedagogical support — teacher capacity and support to ensure pedagogical soundness and correct interpretation of LA results.
• Ethics & data governance — policies and practices for the ethical use of LA and AI, data privacy, security, and accountability.
• Stakeholder engagement — human-centered development including students, educators, decision-makers, and external stakeholders (e.g. employers).
• Quality assurance — procedures for regular evaluation of LA tools and the effectiveness of educational interventions.

AI and the future of LA-based interventions

The authors argue all enablers should increasingly be considered in light of AI's benefits and risks: developing LA that uses AI-based methods (LA with AI), analyzing AI use in teaching and learning (LA of AI), and developing learning design that integrates AI as a transversal topic (LD about AI). They frame the responsibilities for trustworthy LA as distributed — institutions and educational leaders hold strategic responsibility, while teachers own pedagogical soundness and LA interpretation, and keeping the "human in the loop" requires engaging all relevant stakeholders.

What this means for practice

• Administrators. Sequence the work strategically: institutional strategic orientation ranked highest in overall SNAP priority (0.2072) and was the most influential enabler of the others (PageRank 0.2430), so leadership commitment and strategic planning must precede tool procurement.
• Administrators. Fund capacity together with infrastructure — available resources ranked second (SNAP 0.1800) and first on direct influence over implementation (AHP 0.1753), covering dashboards, systems, and the human capacity to use them.
• Learning analytics designers. Ground every dashboard and early-warning model in explicit pedagogical foundations and intended learning outcomes, and provide teachers the support to interpret LA results correctly.
• Administrators. Staff LA decisions with a heterogeneous expert group: senior experts prioritized available resources, junior and technical experts prioritized ethics and data governance, and intermediate experts prioritized strategic orientation or stakeholder engagement, so any single group's ranking is partial.
• Administrators. Treat trustworthiness — ethical compliance, data security, transparent and unbiased algorithms, pedagogical validity — as a precondition to establish first, not as one enabler competing with the other seven.

Limitations

• The expert base was small and narrow: a two-round international Delphi with 18 experts, and while participants came from multiple countries, the majority were affiliated with higher education institutions in four European countries.
• The preliminary survey covered a restricted range and profile of educators; a broader range of educational backgrounds and contexts could have surfaced additional perspectives.
• The seven enablers remain high-level categories with no operational indicators or measures attached, which limits how directly they can be audited or compared across institutions.
• The group decision-making model was never tested in real-life educational decision-making, so no concrete intervention was prioritized in an actual institutional setting.

Connected Concepts

• Learning Analytics
• Higher Education
• AI Governance
• Ethics
• Privacy
• Self-Regulated Learning
• Trust Calibration
• Formative Assessment

Connected Articles

• Jointly Predicting Courses and Grades Using a Transformer-Based Model (TRACE) — Course grade prediction and learning analytics
• Analysis and Prediction of At-Risk Students Using Machine Learning Algorithms — Machine learning prediction for at-risk students

Citation

Svetec, B., Divjak, B., & Kadoić, N. (2026). From learning analytics to educational interventions: Enhancing decision-making and learning design. International Journal of Educational Technology in Higher Education, 23(45).