LLM politeness and language complexity do not affect agent trust in social deduction games

A new study published on arXiv investigated how linguistic variations in large language model (LLM) agents affect their social interactions. Researchers tested whether changes in language complexity or politeness strategies altered how other agents judged or acted upon a speaker's messages. The experiments, conducted within the LURK Werewolf framework, involved rewriting only the natural language realization of messages while preserving the core game content. Across numerous games and replays, neither linguistic range nor politeness strategy significantly impacted the trust other agents placed in the speaker, nor did they affect voting or night actions after statistical corrections. AI

IMPACT Suggests that current LLM agents may not be as susceptible to nuanced linguistic manipulation as human players in social deduction scenarios.

RANK_REASON The cluster contains a research paper published on arXiv detailing experimental findings on LLM behavior. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• alphaXiv
• arXiv
• CatalyzeX Code Finder for Papers
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• Influence Flower
• LURK Werewolf
• ScienceCast

• paper
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →