Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
September 5, 2025Cognitive ScienceOpen Access

Can Large Language Models Simulate Spoken Human Conversations?

View Full Paper
Ask AI
Bookmark
Share

Authors

EMEric MayorUniversity of BaselLBLucas BiettiNorwegian University of Science and TechnologyABAdrian BangerterUniversity of Neuchâtel

Discussion

Loading...

Member takes

Overview

Comparative analysis reveals limitations of large language models in simulating spoken conversations.

Key Points

  • LLMs are proficient in chat‐based conversation but struggle with spoken conversation fidelity.
  • In Study 1, transcripts from LLMs showed exaggerated alignment compared to human conversations.
  • Study 2 found that humans could distinguish LLM transcripts from those of actual conversations.
  • Spoken conversations generated by LLMs differ qualitatively and quantitatively from human interactions.

Cite This Study

Mayor et al. (2025) studied this question.

synapsesocial.com/papers/68bb4d2d6d6d5674bcd01482https://doi.org/10.1111/cogs.70106
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 4 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1Okay as a marker for coordinating transitions in joint actions: Effects of participant role and age in Swiss German and Swiss French interviews2024 · 1 citations
  2. 2AI and the transformation of social science research2023 · 357 citations
  3. 3The Measurement of Observer Agreement for Categorical Data1977 · 81,449 citations
  4. 4Alignment Is a Function of Conversational Dynamics2013 · 21 citations