AI agents reshape consensus formation in human groups

Lin Chen, Ziyi Liu, Xia Hu and Yong Li · arXiv preprint 2 Sep

A collaborative description-game study reports three regimes: low agent presence supports human-led consensus, intermediate presence disrupts convergence and high presence restores stronger, agent-led consensus. Practical value: make agent proportion and disclosure explicit in group systems. Limits: narrow task, author-reported findings and no independent reproduction. Paper