Upgrade retrieval to Agentic RAG and adopt context-aware emotive narration to reduce cognitive load¶

Strategy · All strategies
Evidence · no claims cited

Description¶

The authors' stated future directions for the pipeline. First, the retrieval layer is to move "from a linear, fixed-chunk setup to an Agentic RAG architecture" to handle complex, multi-part student queries, with large-context embedding models enabling multi-step reasoning, query decomposition, and cross-chapter concept synthesis. Second, standard text-to-speech is to be replaced with generative audio models producing emotive, context-aware narration that dynamically adjusts pacing based on instructional complexity to minimize learners' cognitive load.

Design Implications¶

Context¶

Requirements¶

• Modern, large-context embedding models for multi-step reasoning and query decomposition
• Generative audio models replacing standard text-to-speech

Constraints¶

• Both directions are stated as future work; neither is implemented or evaluated in the current system

Target Learners¶

• students submitting complex, multi-part queries on textbook content

Target Learning Goals¶

• multi-step reasoning and cross-chapter concept synthesis; reduced cognitive load during narrated explanations

Related Strategies¶

• Domain-adapt RAG pipelines for mathematical education via entity recognition, notation-aware retrieval, and pedagogical re-ranking
• Design LLM feedback systems to retain encouraging tone, constrain behavioral gaming, and reduce overwhelming cognitive load
• Planned enhancements: multi-turn conversational follow-up, RAG-based personalization, and feedback analytics for AI-assisted instruction
• Design AI-enhanced learning management systems that explicitly target engagement across cognitive, emotional, behavioral, and social domains

Examples¶

-

Key Sources¶

• Siddhanth Sridhar, Shreya Chaurasia, Baddela Sai Yaswantha Reddy, Deepak Parmar, and Dr. Shylaja S S. (2026). Dynamic Learning Solutions: A System for Personalized Educational Video Generation. arXiv preprint. https://arxiv.org/abs/2609.14408