Fuzzy, Symbolic, and Contextual: Enhancing LLM Instruction via Cognitive Scaffolding
This work addresses the challenge of enhancing LLM instruction for educational applications, but it is incremental as it builds on existing scaffolding methods.
The study tackled the problem of how prompt-level inductive biases affect large language models' cognitive behavior in instructional dialogue by introducing a symbolic scaffolding method with short-term memory, and found that the full system consistently outperformed baseline variants, with analysis showing that removing memory or symbolic structure degraded key cognitive behaviors like abstraction and adaptive probing.
We study how prompt-level inductive biases influence the cognitive behavior of large language models (LLMs) in instructional dialogue. We introduce a symbolic scaffolding method paired with a short-term memory schema designed to promote adaptive, structured reasoning in Socratic tutoring. Using controlled ablation across five system variants, we evaluate model outputs via expert-designed rubrics covering scaffolding, responsiveness, symbolic reasoning, and conversational memory. We present preliminary results using an LLM-based evaluation framework aligned to a cognitively grounded rubric. This enables scalable, systematic comparisons across architectural variants in early-stage experimentation. The preliminary results show that our full system consistently outperforms baseline variants. Analysis reveals that removing memory or symbolic structure degrades key cognitive behaviors, including abstraction, adaptive probing, and conceptual continuity. These findings support a processing-level account in which prompt-level cognitive scaffolds can reliably shape emergent instructional strategies in LLMs.