How Linguistics Learned to Stop Worrying and Love the Language Models

2025-01-28

AI Summary

This paper addresses the polarized debates surrounding large language models (LMs) in linguistics—namely, that LMs are either irrelevant to linguistic competence or fully replace linguistic theory. To resolve this, we propose a “theory-driven model interpretation” paradigm. Methodologically, we conduct fine-grained Transformer analyses, cross-linguistic syntactic generalization tests, psycholinguistically motivated controlled generation, and diagnostic probing experiments. Our results demonstrate that LMs exhibit systematic sensitivity to hierarchical syntax, constructional acquisition trajectories, and semantic compositionality boundaries. Crucially, we provide the first empirical evidence that LMs serve not only as *empirical instantiations* of linguistic competence but also as *critical interlocutors* for linguistic theory. This dual role bridges the gap between theoretical linguistics and computational modeling, enabling a bidirectional, interdisciplinary research framework wherein formal linguistic insights inform deep learning architectures—and vice versa. The work thus advances a principled integration of generative grammar and neural language modeling. (149 words)

Technology Category

Natural Language Processing: (Large) Language ModelsMachine Learning: Large Multimodal Models (LMMs)Cognitive Modeling & Cognitive Systems: (Computational) Cognitive Architectures

Application Category

Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactionsUser Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendationSocial Networks and Social Media: Generative AI / large language models and their impact on social systems

Abstract

Language models can produce fluent, grammatical text. Nonetheless, some maintain that language models don't really learn language and also that, even if they did, that would not be informative for the study of human learning and processing. On the other side, there have been claims that the success of LMs obviates the need for studying linguistic theory and structure. We argue that both extremes are wrong. LMs can contribute to fundamental questions about linguistic structure, language processing, and learning. They force us to rethink arguments about learning and are informative for major questions in linguistic theory. But they do not replace linguistic structure and theory. We offer an optimistic take on the relationship between language models and linguistics.

Problem

• Language Models
• Linguistic Research
• Human Language Understanding

Innovation

• Linguistic Models
• Linguistic Theory
• Enhanced Understanding