This paper tests whether AI can realistically guess what a specific social media user would comment when they see a new post.
Large language models are great at words, but they struggle to predict what will happen after they act in a changing world.