Large Language Models as Model Organisms for Human Associative Learning

2025-10-24

AI Summary

This study investigates the representational dynamics underlying human-like associative learning in large language models (LLMs), specifically testing the biologically unverifiable hypothesis of non-monotonic plasticity. Method: Adapting paradigms from cognitive neuroscience, we conduct controlled vocabulary-interference experiments and representation-space analyses across six state-of-the-art LLMs to simulate in-context associative learning. Contribution/Results: We demonstrate that novel association formation triggers non-monotonic internal representational reconfiguration—characterized by transient divergence followed by stabilization. This process is jointly modulated by item-level semantic similarity and global knowledge competition: moderately similar items exhibit significant representational differentiation under high interference. Our work provides the first empirical evidence for non-monotonic plasticity in LLMs and proposes a computationally grounded “lexical interference–representational differentiation” regulatory mechanism. This advances theoretical understanding of memory reorganization and knowledge competition in artificial systems, offering a testable framework for future studies on neural-symbolic integration and learning dynamics.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Knowledge Representation and Reasoning: Nonmonotonic Reasoning
• Cognitive Modeling & Cognitive Systems: Analogy

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

Associative learning--forming links between co-occurring items--is fundamental to human cognition, reshaping internal representations in complex ways. Testing hypotheses on how representational changes occur in biological systems is challenging, but large language models (LLMs) offer a scalable alternative. Building on LLMs' in-context learning, we adapt a cognitive neuroscience associative learning paradigm and investigate how representations evolve across six models. Our initial findings reveal a non-monotonic pattern consistent with the Non-Monotonic Plasticity Hypothesis, with moderately similar items differentiating after learning. Leveraging the controllability of LLMs, we further show that this differentiation is modulated by the overlap of associated items with the broader vocabulary--a factor we term vocabulary interference, capturing how new associations compete with prior knowledge. We find that higher vocabulary interference amplifies differentiation, suggesting that representational change is influenced by both item similarity and global competition. Our findings position LLMs not only as powerful tools for studying representational dynamics in human-like learning systems, but also as accessible and general computational models for generating new hypotheses about the principles underlying memory reorganization in the brain.

Problem

• Studying representational changes in human associative learning
• Investigating vocabulary interference effects on memory reorganization
• Using LLMs as computational models for cognitive neuroscience

Innovation

• Using LLMs as model organisms for associative learning
• Adapting cognitive neuroscience paradigms to LLMs
• Investigating vocabulary interference in representational change