Large Language Models as Proxies for Theories of Human Linguistic Cognition

• DOI: 10.48550/arxiv.2502.07687
• Publication Date: 2025-02-11
• Authors: Imry Ziv; Nur Lan; Emmanuel Chemla; Roni Katzir
• Keywords: Human language

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