The TER-AI project has achieved significant breakthroughs in competency prediction through the development of an innovative AI-driven system that goes well beyond traditional assessment methods. The system successfully implemented a comprehensive hierarchical competency framework that automatically maps learning activities to structures hierarchy: category, skill and sub-skill, demonstrating superior performance compared to conventional human-based evaluation approaches.
The breakthrough lies in the system’s ability to process educational content at scale using large language models, achieving evaluation quality that matches or exceeds human expert assessments while operating at dramatically improved speed, efficiency and cost. Unlike traditional competency systems that rely on static skill inventories or manual classification, our AI-powered approach dynamically generates and maintains a comprehensive skill taxonomy that evolves with educational content and learning patterns, taking into account the learning events of all users within the system.
The system represents a significant advance over existing educational technology solutions by providing real-time, explainable competency tracking with recency mechanisms. This enables personalized learning pathways that adapt to individual student progress and knowledge gaps, moving beyond the one-size-fits-all approach of traditional educational platforms to deliver truly individualized learning experiences at scale.
The scalable architecture positions the platform to serve millions of learners while maintaining personalized experiences, representing a paradigm shift in online education delivery.
To scale and sustain the solution, further research into prompt robustness and evaluation metrics is needed, along with clearer processes for human-in-the loop quality control mechanisms for dynamic competency taxonomies. Demonstrating real-world impact through broader pilot studies, creating automated quality- and impact-tracking systems, and establishing a secure review mechanism for over-generation are critical. For commercialisation, alignment with educational standards, explainability requirements, and data privacy regulations (e.g. GDPR) must be maintained.
