XCEL

At the same time, generative AI is rapidly entering education, but current AI assistants often lack educational expertise, provide generic feedback, and offer little transparency about how their recommendations are generated. This limits trust and makes them difficult to use in educational settings.

The XCEL project addresses both these challenges by developing a new generation of Learning Analytics Agents: conversational AI assistants that combine learning analytics, educational science, explainable AI, and privacy-preserving technologies. Rather than simply presenting data, these agents help teachers and learners interpret information, understand what it means, and translate it into concrete actions that improve learning.

AI that understands learning

A first innovation is making AI pedagogically intelligent. Instead of relying only on patterns in data, XCEL builds its Learning Analytics Agents on established theories from learning sciences and educational measurement. The system learns to recognise meaningful indicators of learning behaviour, particularly those related to self-regulated learning, the ability of learners to plan, monitor and evaluate their own learning. By grounding AI in educational theory, the project aims to provide feedback that is both scientifically sound and practically useful.

A second innovation is explainable AI. Rather than functioning as a black box, the system explains why it reaches certain conclusions or recommendations. Teachers can understand why particular learners may need additional support, while learners receive personalised explanations that help them reflect on their own progress.

Third, XCEL develops conversational AI that transforms learning data into meaningful dialogue. Instead of passively displaying information, the Learning Analytics Agent engages teachers and learners in reflective conversations. It asks questions, offers personalised feedback, suggests next steps, and intervenes when it detects learning difficulties or reduced engagement.

A fourth innovation addresses privacy. XCEL develops privacy-preserving methods based on federated learning, allowing AI models to learn from distributed data without centralising personal information. This enables organisations to benefit from shared intelligence while complying with GDPR and other privacy regulations.

Demonstrating AI-powered learning support

The XCEL project will develop and evaluate a demonstrator of its Learning Analytics Agent in two representative learning environments.

The first use case focuses on K–12 education, where the system supports teachers and pupils in digital STEM learning environments. The second targets corporate training, helping trainers and employees make better use of learning data in professional development programmes. Across both settings, the project will validate how conversational AI can provide personalised feedback, stimulate reflective learning, and support data-informed teaching.

Beyond demonstrating the technology itself, XCEL will evaluate how well users understand and trust AI-generated explanations, whether teachers can make better instructional decisions, and whether learners improve their self-regulated learning skills through personalised guidance.

“XCEL demonstrates how explainable and conversational AI can transform learning analytics from static dashboards into intelligent learning companions. By combining educational science, explainable AI, conversational interfaces and privacy-preserving technologies, the project enables teachers and learners to make better-informed decisions, strengthens trust in AI, and improves learning outcomes. In doing so, XCEL helps shape the next generation of digital learning for schools, training providers, and lifelong learning.”