This research explores integrating human cognitive signals (particularly eye-tracking data) into language models to create more human-aligned AI systems.

Key Insights:

• Cognitive signals help address data scarcity challenges in language model training
• Integration improves model performance while potentially reducing environmental costs
• Eye-tracking and other cognitive signals create more user-centric language models
• Applications span multiple fields including education, linguistics, and medical domains

Medical Relevance: By incorporating human cognitive processing patterns, these models could better understand medical information processing, potentially improving healthcare interfaces, patient education systems, and clinical decision support tools.

Integrating Cognitive Processing Signals into Language Models: A Review of Advances, Applications and Future Directions