Shared Meaning, Separate Styles: Asymmetric Human–AI Accommodation at Scale

• In Event: Communication, Information Technologies, and Media Sociology Section Roundtables
  In Refereed Roundtable (60 minute): 4. General AI, II

Abstract

Large language models (LLMs) are increasingly embedded in daily life, raising a question: how do people interact with artificial intelligence (AI), and does it resemble human–human talk? Using 148,484 human-AI conversations from mainland China and Hong Kong, we track alignment from conversation-level convergence to message-level accommodation and symmetry from linguistic and semantic perspectives. We find pervasive semantic convergence (~77%) but frequent linguistic divergence (~57%); accommodation is asymmetric and AI-led. Convergence is shaped by topic tightness, conversational context (language and location), conversation style (longer sentences, emoji), psychological motivations (achievement, tentativeness), and social focus (affiliation/friends), revealing AI-divide patterns.

Authors

• Jiayu Zheng, Johns Hopkins University

• Yuanhao (Colin) Liu, Johns Hopkins University

• Feng Wei, Johns Hopkins University