Federal Project Grant Award Summary

The Office of Multidisciplinary Activities within the National Science Foundation (NSF) awarded The Ohio State University a $459,039 Project Grant (CFDA 47.075: Social, Behavioral, and Economic Sciences) effective August 1, 2026, with a completion date of July 31, 2029. This research project investigates how memory retrieval and prediction mechanisms in human language processing can be explained through the computational attention mechanisms underlying modern large language models (LLMs). The deliverables include systematic comparative analyses of attention-based and prediction-derived metrics extracted from LLMs, behavioral datasets documenting human language comprehension across syntactic, semantic, and interference phenomena, and theoretical insights evaluating memory-based versus expectation-based accounts of language processing difficulty.

The project addresses a fundamental gap in cognitive science by testing whether attention mechanisms in contemporary artificial intelligence systems can jointly explain dual human cognitive processes—memory retrieval and probabilistic prediction—during sentence comprehension. Outputs will encompass empirical findings, model-derived computational metrics, and comparative behavioral signatures that clarify when and how LLM attention patterns successfully capture human processing patterns. This research will advance public understanding of AI system capabilities and limitations relative to human cognition, contributing to both cognitive science theory and practical knowledge about deployed artificial intelligence technologies.