Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Fuzzy, Symbolic, and Contextual: Enhancing LLM Instruction via Cognitive Scaffolding

Vanessa Figueiredo

TL;DR

This work tackles how prompt-level cognitive scaffolds shape LLM instruction in dynamic dialogues. It proposes a three-layer symbolic framework—boundary prompts, fuzzy scaffolding, and a short-term memory schema—that enables adaptive, interpretable tutoring without model retraining. Through ablation studies in Socratic tutoring across two domains, the full system consistently outperforms baselines, with memory and symbolic structure proving critical for coherence, abstraction, and adaptive probing. The approach advances trustworthy, cognitively grounded language agents and opens paths for broader educational and explanatory applications. The evaluation strategy, while using synthetic users, demonstrates a scalable framework for operationalizing cognitive control in LLMs.

Abstract

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.

Fuzzy, Symbolic, and Contextual: Enhancing LLM Instruction via Cognitive Scaffolding

TL;DR

This work tackles how prompt-level cognitive scaffolds shape LLM instruction in dynamic dialogues. It proposes a three-layer symbolic framework—boundary prompts, fuzzy scaffolding, and a short-term memory schema—that enables adaptive, interpretable tutoring without model retraining. Through ablation studies in Socratic tutoring across two domains, the full system consistently outperforms baselines, with memory and symbolic structure proving critical for coherence, abstraction, and adaptive probing. The approach advances trustworthy, cognitively grounded language agents and opens paths for broader educational and explanatory applications. The evaluation strategy, while using synthetic users, demonstrates a scalable framework for operationalizing cognitive control in LLMs.

Abstract

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.

Paper Structure

This paper contains 17 sections, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Problem Definition
  4. Experimental Setup
  5. Prompting Conditions
  6. Evaluation Protocol
  7. Results
  8. Planned Experiments and Applicability
  9. Discussion and Conclusion
  10. Prompting Framework Details
  11. Symbolic Prompt Architecture
  12. Evaluation Rubric
  13. Ablation Mapping and Component Impact
  14. Results Summary
  15. Evaluation Metrics (Bar Colors):
  16. ...and 2 more sections

Figures (2)

  • Figure 1: Runtime prompting loop with symbolic scaffolding and memory. Each turn integrates boundary prompts, fuzzy heuristics, and symbolic memory updates to produce adaptive, context-aware behavior.
  • Figure 2: Bar chart of average evaluation scores across experimental conditions.