Polarization Dynamics under Algorithmic Recommendations in LLM-based Online Social Networks

DOI10.48550/arXiv.2610.13351MaRDI QIDQ7394776arXivFDO

Authors Mohammad Shirzadi, Ali Safarpoor Dehkordi, Mehwish Nasim, Ahad N. Zehmakan

Publication date 9 October 2026

Published in arXiv

Tldr We compare two types of algorithmic recommendations in an LLM-based online social network simulation, finding that both Edge Recommendation and Content-Exposure Recommendation increase polarization and extremization of opinions, the former more so than the latter.

Full work available at URL https://doi.org/10.48550/arXiv.2610.13351

arXiv classification

cs.SI

Cites work

• Exposure to ideologically diverse news and opinion on Facebook
• Out of One, Many: Using Language Models to Simulate Human Samples
• Reaching a Consensus
• Social influence and opinions

This page was built for publication: Polarization Dynamics under Algorithmic Recommendations in LLM-based Online Social Networks

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7394776)