This research introduces Generative Causal Testing (GCT), a novel framework that bridges deep learning models and neuroscience to explain language processing in the brain.

• Demonstrates how LLM representations can effectively predict brain responses to language stimuli
• Develops a method to generate testable hypotheses about what specific language features drive activity in different brain regions
• Uses AI-generated test stimuli to validate these hypotheses in follow-up experiments
• Reveals that different brain regions process distinct aspects of language (e.g., syntax vs. semantics)

This breakthrough matters for medical research by providing more interpretable models of brain function, potentially enhancing our understanding of language disorders and improving clinical interventions for neurological conditions.

Generative causal testing to bridge data-driven models and scientific theories in language neuroscience