Emergent Wisdom: Empowering Constructivism by Proxying Human Reasoning with LLM Thought Traces

Despite wide adoption of AI in education, fostering critical thinking skills in learners remain a significant challenge. Pedagogical strategies aligned with constructivism, especially inquiry-based techniques like Socratic Learning, excel at promoting reflective inquiry but are difficult to scale without expert facilitation. This paper introduces Emergent Wisdom, a novel multi-agent system that leverages LLM-generated thought traces as proxies for human thinking skills to develop interpretable, feedback sensitive and actionable prompt guidelines. These guidelines inform instructional plans and delivers adaptive dialectic inquiry-based learning. Feasibility studies across diverse domains—including mathematics, reading comprehension, and science—demonstrate that this approach effectively captures the nuances of reasoning progressions in learners to facilitate real-time adaptive scaffolding through generative AI modules. By formulating instructional strategies aimed at developing capabilities informed by LLM thought traces, our framework positions AI as a mirror for human reasoning and exploits its self-reflection capabilities to drive a symbiotic constructivist learning experience.