J. Mark Bishop and Stephen J. Cowley explore the implications of Elan Barenholtz's autogenerative theory on linguistics in their recent paper, "How Do I Know What to Say Next?" submitted on July 8, 2026. The paper addresses gaps in Roy Harris's Integrationist linguistics, particularly in computational language approaches.
Understanding Barenholtz's Autogenerative Theory
Barenholtz's theory proposes a structural mechanism that addresses the gaps left by Integrationism. It argues that language is not merely a code but a dynamic activity focused on joint action and communication. This theory provides insights into how language remains open to new integrations, which is essential for effective communication.
The autogenerative model enriches Harris's ideas by offering a computational framework that aligns with contemporary language models, particularly Large Language Models (LLMs). It emphasizes the connection between linguistic and non-linguistic semiotic activities, contributing to a comprehensive understanding of communication.
Implications for Natural Language Processing
This research carries significant implications for practitioners and researchers working with natural language processing and large language model design. The paper outlines the statistical structures that LLMs utilize, which are vital for understanding how these models generate language. By elucidating the nature of these structures, Barenholtz's theory helps clarify what LLMs can and cannot achieve.
