LLMs eager to facilitate online discussions, humans more cautious

A new study published on arXiv explores the tendencies of humans and Large Language Models (LLMs) in facilitating online discussions. Researchers created the PEFK corpus to standardize facilitation datasets and conducted a survey using expert participants and LLM-as-a-judge models. The findings indicate that LLMs are overly eager to facilitate, whereas humans are more cautious, though both are more certain when determining that facilitation is unnecessary. Attempts to correct LLM behavior and training ModernBERT classifiers showed that the classifiers performed more reliably, but current datasets limit their performance ceiling. AI

IMPACT LLM behavior in online moderation may require adjustment to align with human caution.

RANK_REASON The cluster contains an academic paper detailing research findings on LLM behavior. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• alphaXiv
• arXiv
• CatalyzeX
• DagsHub
• Dimitrios Tsirmpas
• Gotit.pub
• Hugging Face
• LLMs
• ModernBERT
• ScienceCast

• paper
• other

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