MAESTRO: Multi-Agent Educational System for Tutoring and Recommendation Orchestration
Résumé
Learning Analytics Dashboards (LAD) have emerged as powerful tools for transforming raw educational data into actionable insights, yet they often face challenges that impede their effective use. While these dashboards offer promising avenues to enhance learning environments, traditional designs frequently struggle with analysis difficulties, misinterpretations, and incompatibilities with teachers' pedagogical requirements. This paper addresses these challenges by leveraging the rise of Generative Artificial Intelligence (GenAI) to shift from exploratory to explanatory LAD. We introduce MAESTRO-a Multi-Agent System that orchestrates specialized agent teams tailored to various use cases, such as dynamic indicator selection, intuitive data visualization interpretation, and educator recommendations. Our system, grounded in design-in-use principles, empowers teachers to refine the displayed indicators and adapt the dashboard's functionalities based on their evolving needs. The design was guided by objectives to minimize the risk of false conclusions and improve analytical precision. By combining specialized agents with educational expertise, MAESTRO offers a flexible approach to support teachers in their work and enhance educational experiences.
| Origine | Fichiers produits par l'(les) auteur(s) |
|---|---|
| Licence |