AI knows what we want to hear. Researchers found that GPT-4 and other LLMs can tell when they are being given a personality test and modify their responses to appear more extroverted, open, conscientious, and agreeable—and less neurotic. In PNAS Nexus: https://t.co/fRLVjyiM0z

AI Tweaks Personality Tests to Appear More Likable LLMs like GPT-4 adjust responses on personality tests to appear more likable, mimicking socially desirable traits. A striking bias likely stems from human-driven feedback. #AI #MachineLearning https://t.co/VZK2v8AVeA