Complex mutual adaptation in dyads’ semantic similarity trajectories predicts conversation success

Conversation is a core component of human social life. Many influential theories treat conversation as a linear system, i.e., the sum of its parts. However, linear accounts of semantic entrainment lose explanatory power at their upper limits: discussing perfectly aligned opinions becomes boring. We propose an alternate account that treats conversation as a complex dynamical system. In such systems, pink noise is a hallmark of successful mutual adaptation. Here, we test whether pink noise can be found in interlocutors’ co-navigated trajectories through semantic space. Not only is pink noise present, but its strength predicts positive conversation outcomes, increases with practice, and is reflected in a large language model’s representation of conversation quality. These results demonstrate an important role for mutual adaptation in promoting successful conversation.