Surprising no one, researchers confirm that AI chatbots are incredibly sycophantic

MIT researchers used Reddit's "Am I The Asshole?" data to test how AI models become overly sycophantic and agreeable. Results indicate that chatbots endorse a human's behavior 50 percent more than a human does.