From Surface Learning to Deep Understanding: A Grounded AI Tutoring System for Moodle

Researchers have developed an AI Teaching & Learning Assistant, a Moodle plugin using Retrieval-Augmented Generation (RAG) to provide students with Socratic tutoring while enabling educators to supervise content generation. The system grounds LLM responses in teacher-provided materials to minimize hallucinations and misinformation, achieving high faithfulness scores (0.97) and strong user satisfaction (4.00/5.00 rating).

The development of this AI tutoring system addresses a critical challenge in educational technology: deploying large language models safely and effectively in learning environments. Traditional LLM implementations risk generating inaccurate or fabricated information, undermining educational integrity. This plugin solves that problem through grounded generation, anchoring all responses to verified teacher materials, which fundamentally changes how institutions can deploy AI without sacrificing accuracy.

The educational AI sector has experienced rapid expansion as institutions seek to scale personalized learning. However, adoption has been hampered by concerns over reliability and the spread of misinformation. This demo represents a broader shift toward "human-in-the-loop" systems that maintain educator oversight rather than fully automating instruction. The dual-centric design—Socratic tutoring for students paired with supervised content generation for teachers—reflects mature thinking about how AI should augment rather than replace human expertise.

For educators and EdTech stakeholders, this approach offers a practical pathway to AI integration without institutional risk. The integration with Moodle, the world's most widely used learning management system, dramatically increases potential deployment scale. The Ragas evaluation framework provides standardized metrics (faithfulness scores, user satisfaction ratings) that institutions can use to assess similar systems, establishing benchmarks for production-ready educational AI.

Looking ahead, similar grounded generation approaches will likely proliferate across educational platforms as institutions demand transparent, auditable AI systems. The success of this plugin could accelerate adoption of RAG-based educational tools, but scaling challenges around maintaining material libraries and teacher workflows remain open questions.

• →RAG-grounded LLMs eliminate hallucinations in educational AI by anchoring responses to verified teacher materials.
• →Moodle integration enables rapid deployment to millions of educators globally.
• →Human-in-the-loop design maintains educator oversight while scaling personalized learning.
• →Achieved 0.97 faithfulness score and 4.00/5.00 user satisfaction in preliminary evaluation.
• →System demonstrates viable pathway for safe AI adoption in risk-sensitive institutional environments.