Abstract
Generative AI (GenAI) is ubiquitously reshaping higher education, yet its pedagogical value remains constrained due to lack of theoretical grounding. Current chatbot tutors often operate ad-hoc neglecting established learning principles such as scaffolding, feedback loops, and metacognitive reflection. This study addresses this gap by introducing a multi-agent architecture in which theory-aligned agents, PromptAgent (Bloom’s Taxonomy), CoachAgent (Constructivism), and ReflectionAgent (Metacognition), are orchestrated by a ModeratorAgent to deliver an integrated, interactive learning flow. The contribution of this study are as follows: (1) a conceptual model linking established learning theories to GenAI agent design, (2) a technical architecture for orchestrating multiple theory-driven agents, (3) a validated prompt-engineering framework operationalising educational theory into structured agent behaviours, and (4) a working prototype implemented in Azure AI Foundry.
Recommended Citation
Sultana, Khadija and Sianaki, Omid Ameri, "A Multi-Agent Generative AI System for incorporating
Bloom’s Taxonomy, Constructivism, and Metacognition
Theories in Student’s Learning Performance" (2025). ACIS 2025 Proceedings. 134.
https://aisel.aisnet.org/acis2025/134