Collaborative Metacognitive Scaffolding for Interdisciplinary Project-Based Learning: A Transformer-Augmented Framework with Real-Time Disciplinary Convergence Analysis

Abstract

We propose a novel framework for interdisciplinary project-based learning that integrates real-time collaborative metacognitive scaffolding with transformer-based analysis of disciplinary convergence. The conventional approach to interdisciplinary collaboration often lacks structured mechanisms for metacognitive reflection and dynamic feedback, leading to fragmented knowledge integration and suboptimal team performance. Our system addresses this gap through three interconnected components: a metacognitive annotation layer that captures disciplinary perspectives and confidence levels, a cross-disciplinary integration analyzer that quantifies conceptual synthesis using a novel Disciplinary Convergence Index (DCI), and a collaborative workflow optimizer that generates adaptive prompts and peer-matching recommendations. The framework employs a fine-tuned GPT-4 model to process student reflections and produce integrative insights, which are visualized in a shared digital workspace alongside real-time DCI metrics. Furthermore, the system dynamically adjusts project guidelines and mentorship interventions based on detected integration barriers, thereby fostering deeper interdisciplinary synthesis. Experiments demonstrate that the proposed method significantly improves team cohesion and conceptual integration compared to traditional project-based learning environments. The contributions include a scalable architecture for metacognitive scaffolding, a quantitative metric for interdisciplinary convergence, and a practical implementation combining collaborative tools with AI-driven reflection analysis. This work advances the design of learning systems by bridging the gap between metacognitive awareness and interdisciplinary collaboration in project-based settings.