Ariel Flint, Luca Maria Aiello, Romualdo Pastor-Satorras, and Andrea Baronchelli
PNAS August 18, 2026 123 (34) e2531697123
Large language models (LLMs) are increasingly deployed in large numbers, and their interactions make collective behavior harder to anticipate than that of a single model. While most studies compare one model with a collective of fixed size, we ask a key yet overlooked question: What is the role of group size? We show that interaction among LLMs can magnify individual biases, generate new ones, or even overturn individual preferences, and that, crucially, these effects scale in unexpected, nonlinear ways with group size. Our results demonstrate that more is different for LLM populations: The number of interacting agents is a key driver of the dynamics, with implications for the design and governance of multi-agent AI systems.
Read the full article at: www.pnas.org
Complexity Digest 