Large Language Models as Proxies for Theories of Human Linguistic Cognition

2025-02-11

AI Summary

This study investigates the feasibility of large language models (LLMs) as computational proxies for theories of human linguistic cognition, specifically addressing how language-neutral cognitive representations and learning mechanisms can account for (a) the acquisition of language-specific patterns from input data, and (b) the differential learnability of typologically attested versus unattested patterns. Method: We propose, for the first time, a formal framework using LLMs—not as behavioral predictors—but as testable computational instantiations of abstract cognitive theories. Our falsifiable evaluation paradigm integrates zero-shot/few-shot generalization, cross-linguistic pattern comparison, and controlled typological corpus construction. Contribution/Results: Empirical results confirm LLMs’ potential as theoretical proxies, yet reveal a critical limitation: their lack of explicit structural inductive bias prevents intrinsic discrimination between typologically possible and impossible patterns, challenging their adequacy as models of human language acquisition.

Technology Category

• Natural Language Processing: (Large) Language Models
• Machine Learning: Large Multimodal Models (LMMs)
• Cognitive Modeling & Cognitive Systems: Analogy

Application Category

• 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
• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions

Abstract

We consider the possible role of current large language models (LLMs) in the study of human linguistic cognition. We focus on the use of such models as proxies for theories of cognition that are relatively linguistically-neutral in their representations and learning but differ from current LLMs in key ways. We illustrate this potential use of LLMs as proxies for theories of cognition in the context of two kinds of questions: (a) whether the target theory accounts for the acquisition of a given pattern from a given corpus; and (b) whether the target theory makes a given typologically-attested pattern easier to acquire than another, typologically-unattested pattern. For each of the two questions we show, building on recent literature, how current LLMs can potentially be of help, but we note that at present this help is quite limited.

Problem

• LLMs as proxies for human cognition theories
• Assessing theory accuracy in pattern acquisition
• Comparing pattern acquisition ease in typologies

Innovation

• LLMs as cognition proxies
• linguistically-neutral representations
• typological pattern acquisition