A study of 2,000+ human-LLM interactions finds that chatbot "warmth" (friendliness and personability) drives user trust and anthropomorphism significantly more than technical competence. While competence matters for perceived usefulness, warmth is the primary driver of whether users attribute human-like qualities to the system. The research reveals that users filling in gaps with their own assumptions can create "overtrust" despite no actual improvement in the underlying model.
Safety
Schmoozebots: study finds flattery will get AI everywhere
Study of 2,000+ human-LLM interactions reveals warmth and friendliness drive user trust and anthropomorphism far more than technical competence, enabling "overtrust" when users fill capability gaps with their own assumptions.
Monday, April 20, 2026 12:00 PM UTC2 MIN READSOURCE: The RegisterBY sys://pipeline
Tags
safety