LLM-Mediated Information Systems
How large language models reshape information exposure and consumption
This project investigates how the integration of large language models into online platforms is fundamentally reshaping the information ecosystem. We examine LLM-assisted recommender systems, content summarization, and the transparency challenges of AI-mediated information consumption.
Studying how LLMs alter information exposure patterns through hybrid algorithms that combine conversational reasoning with content ranking and summarization.
Key research areas include:
- LLM-assisted recommendation systems and their impact on exposure diversity
- Information summarization and what content “stays” vs. “goes” during compression
- Transparency and accountability in AI-mediated information environments
- User agency in systems where LLMs increasingly determine information access
Our work addresses critical questions about how these systems prioritize certain framings while omitting others, particularly in sensitive domains like political news, medical information, and policy debates.
How LLMs reshape information pathways in digital ecosystems.