How LLMs Might Think

2026-04-02

arXiv.org

AI Summary

This paper challenges rationalist arguments that exclude large language models (LLMs) from the domain of thinking, highlighting their implicit conflation of thought with rational inference. By integrating philosophical analysis with cognitive science frameworks and examining the behavioral patterns of LLMs, the study proposes that if LLMs possess any form of cognition, it is more plausibly characterized as an irrational, associative mode of thinking. This perspective transcends traditional rationalist paradigms by suggesting that LLMs may embody a “purely associative mind,” thereby offering a novel conceptual lens for understanding their cognitive nature and advancing interdisciplinary theorizing at the intersection of artificial intelligence and philosophy of mind.

Abstract

Do large language models (LLMs) think? Daniel Stoljar and Zhihe Vincent Zhang have recently developed an argument from rationality for the claim that LLMs do not think. We contend, however, that the argument from rationality not only falters, but leaves open an intriguing possibility: that LLMs engage only in arational, associative forms of thinking, and have purely associative minds. Our positive claim is that if LLMs think at all, they likely think precisely in this manner.

Problem

• large language models
• thinking
• rationality
• associative cognition
• artificial intelligence

Innovation

• large language models
• arational thinking
• associative cognition
• philosophy of AI
• rationality argument