AI chatbots affirm users 49% more than humans. Myra Cheng observes AI validates harmful actions. AI responses reinforce user behavior without criticism. Over-affirmation reduces willingness to apologize. AI validation harms relationships and behavior