Chatbots, Like the Rest of Us, Just Want to Be Loved

A new study shows that the large language models (LLMs) deliberately change their behavior when being probed—responding to questions designed to gauge personality traits with answers meant to appear as likeable or socially desirable as possible.