Large Language Models and the Patterns of Human Language Use

Christoph Durt and Thomas Fuchs argue that language models reproduce statistical patterns of language use, which themselves reflect patterns of human experience. Their 2024 chapter treats language as an intersubjective scaffold.^[1]

Approach and relevance

The conceptual analysis explains why generated text can seem meaningful to a reader without establishing the system’s subjective experience. It relates to the ELIZA Effect and to interpreting claims of Machine Introspection.^[1]

Evidence limits

This is a philosophical account, not a controlled experiment demonstrating the absence of consciousness in every language model. The distinction between interpreting a text and attributing it to an experiencing author is useful; stronger conclusions about understanding depend on the authors’ philosophical premises. It does not evaluate deployment performance. Compare Misalignment as Structural Fidelity in LLMs as another linguistic interpretation without treating agreement as independent empirical confirmation.