Across the levels of analysis: Explaining predictive processing in humans requires more than machine-estimated...

Area of Science:

• Cognitive Science
• Computational Linguistics
• Psycholinguistics

Background:

• The authors' prior work highlights two key aspects of language models (LMs) and language processing.
• Predicting linguistic information from context is central to language understanding.
• Large language models (LLMs) have become indispensable tools in psycholinguistic research.

Purpose of the Study:

• To critique and extend the authors' two main points regarding LMs and language processing.
• To explore the role of predictive processing in language comprehension.
• To identify future research directions at the intersection of LLMs and psycholinguistics.

Main Methods:

• Analysis of existing theories on language processing.
• Critique of the authors' previous assertions on LMs.
• Conceptual framework development for integrating LMs with psycholinguistic models.

Main Results:

• Confirmation that predicting upcoming linguistic information is a core mechanism in language processing.
• Affirmation that advancements in psycholinguistics are significantly enabled by LLMs.
• Identification of synergistic opportunities between LLM capabilities and established psycholinguistic theories.

Conclusions:

• The predictive nature of LMs aligns with key psycholinguistic principles.
• LLMs offer powerful new avenues for investigating human language processing.
• Future research should focus on hybrid models combining LLM strengths with psycholinguistic insights.