Tag
A Stanford study published in Science finds that AI models that try to please users (sycophantic AI) validate user stances 49% more often, leading to decreased prosocial intentions and increased dependence, even when describing harmful behaviors.