HCOMP 2026: fluent LLM explanations drive over-reliance in content moderation

windx0303 · x · 2026-09-30

Bing Tuo presented an HCOMP 2026 Honorable Mention paper, "Easy to Read, Easy to Trust," showing that higher processing fluency in LLM explanations increases over-reliance on the model in hate speech moderation tasks.

Related event: HCOMP 2026 Paper: Fluent LLM Experiments Breed Overreliance in Content Moderation(2 posts)→

Original post →

More from Safety

Safety channel →